audio_renderer: passthrough 6 channel data

gl_shader_manager: Fix interaction between graphics and compute

README.md

@@ -1,6 +1,6 @@
 yuzu emulator
 =============
-[![Travis CI Build Status](https://travis-ci.org/yuzu-emu/yuzu.svg?branch=master)](https://travis-ci.org/yuzu-emu/yuzu)
+[![Travis CI Build Status](https://travis-ci.com/yuzu-emu/yuzu.svg?branch=master)](https://travis-ci.com/yuzu-emu/yuzu)
 [![Azure Mainline CI Build Status](https://dev.azure.com/yuzu-emu/yuzu/_apis/build/status/yuzu%20mainline?branchName=master)](https://dev.azure.com/yuzu-emu/yuzu/)
 
 yuzu is an experimental open-source emulator for the Nintendo Switch from the creators of [Citra](https://citra-emu.org/).
@@ -21,7 +21,7 @@ For development discussion, please join us on [Discord](https://discord.gg/XQV6d
 
 Most of the development happens on GitHub. It's also where [our central repository](https://github.com/yuzu-emu/yuzu) is hosted.
 
-If you want to contribute please take a look at the [Contributor's Guide](CONTRIBUTING.md) and [Developer Information](https://github.com/yuzu-emu/yuzu/wiki/Developer-Information). You should as well contact any of the developers on Discord in order to know about the current state of the emulator.
+If you want to contribute please take a look at the [Contributor's Guide](CONTRIBUTING.md) and [Developer Information](https://github.com/yuzu-emu/yuzu/wiki/Developer-Information). You should also contact any of the developers on Discord in order to know about the current state of the emulator.
 
 ### Building
 

src/audio_core/algorithm/interpolate.cpp

@@ -8,13 +8,14 @@
 #include <climits>
 #include <cmath>
 #include <vector>
+
 #include "audio_core/algorithm/interpolate.h"
 #include "common/common_types.h"
 #include "common/logging/log.h"
 
 namespace AudioCore {
 
-constexpr std::array<s16, 512> curve_lut0 = {
+constexpr std::array<s16, 512> curve_lut0{
     6600,  19426, 6722,  3,     6479,  19424, 6845,  9,     6359,  19419, 6968,  15,    6239,
     19412, 7093,  22,    6121,  19403, 7219,  28,    6004,  19391, 7345,  34,    5888,  19377,
     7472,  41,    5773,  19361, 7600,  48,    5659,  19342, 7728,  55,    5546,  19321, 7857,
@@ -56,7 +57,7 @@ constexpr std::array<s16, 512> curve_lut0 = {
     19403, 6121,  22,    7093,  19412, 6239,  15,    6968,  19419, 6359,  9,     6845,  19424,
     6479,  3,     6722,  19426, 6600};
 
-constexpr std::array<s16, 512> curve_lut1 = {
+constexpr std::array<s16, 512> curve_lut1{
     -68,   32639, 69,    -5,    -200,  32630, 212,   -15,   -328,  32613, 359,   -26,   -450,
     32586, 512,   -36,   -568,  32551, 669,   -47,   -680,  32507, 832,   -58,   -788,  32454,
     1000,  -69,   -891,  32393, 1174,  -80,   -990,  32323, 1352,  -92,   -1084, 32244, 1536,
@@ -98,7 +99,7 @@ constexpr std::array<s16, 512> curve_lut1 = {
     32551, -568,  -36,   512,   32586, -450,  -26,   359,   32613, -328,  -15,   212,   32630,
     -200,  -5,    69,    32639, -68};
 
-constexpr std::array<s16, 512> curve_lut2 = {
+constexpr std::array<s16, 512> curve_lut2{
     3195,  26287, 3329,  -32,   3064,  26281, 3467,  -34,   2936,  26270, 3608,  -38,   2811,
     26253, 3751,  -42,   2688,  26230, 3897,  -46,   2568,  26202, 4046,  -50,   2451,  26169,
     4199,  -54,   2338,  26130, 4354,  -58,   2227,  26085, 4512,  -63,   2120,  26035, 4673,
@@ -146,10 +147,10 @@ std::vector<s16> Interpolate(InterpolationState& state, std::vector<s16> input,
 
     if (ratio <= 0) {
         LOG_CRITICAL(Audio, "Nonsensical interpolation ratio {}", ratio);
-        ratio = 1.0;
+        return input;
     }
 
-    const int step = static_cast<int>(ratio * 0x8000);
+    const s32 step{static_cast<s32>(ratio * 0x8000)};
     const std::array<s16, 512>& lut = [step] {
         if (step > 0xaaaa) {
             return curve_lut0;
@@ -160,28 +161,37 @@ std::vector<s16> Interpolate(InterpolationState& state, std::vector<s16> input,
         return curve_lut2;
     }();
 
-    std::vector<s16> output(static_cast<std::size_t>(input.size() / ratio));
-    int in_offset = 0;
-    for (std::size_t out_offset = 0; out_offset < output.size(); out_offset += 2) {
-        const int lut_index = (state.fraction >> 8) * 4;
+    const std::size_t num_frames{input.size() / 2};
 
-        const int l = input[(in_offset + 0) * 2 + 0] * lut[lut_index + 0] +
-                      input[(in_offset + 1) * 2 + 0] * lut[lut_index + 1] +
-                      input[(in_offset + 2) * 2 + 0] * lut[lut_index + 2] +
-                      input[(in_offset + 3) * 2 + 0] * lut[lut_index + 3];
+    std::vector<s16> output;
+    output.reserve(static_cast<std::size_t>(input.size() / ratio + InterpolationState::taps));
 
-        const int r = input[(in_offset + 0) * 2 + 1] * lut[lut_index + 0] +
-                      input[(in_offset + 1) * 2 + 1] * lut[lut_index + 1] +
-                      input[(in_offset + 2) * 2 + 1] * lut[lut_index + 2] +
-                      input[(in_offset + 3) * 2 + 1] * lut[lut_index + 3];
+    for (std::size_t frame{}; frame < num_frames; ++frame) {
+        const std::size_t lut_index{(state.fraction >> 8) * InterpolationState::taps};
 
-        const int new_offset = state.fraction + step;
+        std::rotate(state.history.begin(), state.history.end() - 1, state.history.end());
+        state.history[0][0] = input[frame * 2 + 0];
+        state.history[0][1] = input[frame * 2 + 1];
 
-        in_offset += new_offset >> 15;
-        state.fraction = new_offset & 0x7fff;
+        while (state.position <= 1.0) {
+            const s32 left{state.history[0][0] * lut[lut_index + 0] +
+                           state.history[1][0] * lut[lut_index + 1] +
+                           state.history[2][0] * lut[lut_index + 2] +
+                           state.history[3][0] * lut[lut_index + 3]};
+            const s32 right{state.history[0][1] * lut[lut_index + 0] +
+                            state.history[1][1] * lut[lut_index + 1] +
+                            state.history[2][1] * lut[lut_index + 2] +
+                            state.history[3][1] * lut[lut_index + 3]};
+            const s32 new_offset{state.fraction + step};
 
-        output[out_offset + 0] = static_cast<s16>(std::clamp(l >> 15, SHRT_MIN, SHRT_MAX));
-        output[out_offset + 1] = static_cast<s16>(std::clamp(r >> 15, SHRT_MIN, SHRT_MAX));
+            state.fraction = new_offset & 0x7fff;
+
+            output.emplace_back(static_cast<s16>(std::clamp(left >> 15, SHRT_MIN, SHRT_MAX)));
+            output.emplace_back(static_cast<s16>(std::clamp(right >> 15, SHRT_MIN, SHRT_MAX)));
+
+            state.position += ratio;
+        }
+        state.position -= 1.0;
     }
 
     return output;

src/audio_core/algorithm/interpolate.h

@@ -6,12 +6,17 @@
 
 #include <array>
 #include <vector>
+
 #include "common/common_types.h"
 
 namespace AudioCore {
 
 struct InterpolationState {
-    int fraction = 0;
+    static constexpr std::size_t taps{4};
+    static constexpr std::size_t history_size{taps * 2 - 1};
+    std::array<std::array<s16, 2>, history_size> history{};
+    double position{};
+    s32 fraction{};
 };
 
 /// Interpolates input signal to produce output signal.

src/audio_core/audio_renderer.cpp

@@ -291,8 +291,9 @@ void AudioRenderer::VoiceState::RefreshBuffer(Memory::Memory& memory) {
             samples[index * 2 + 1] = new_samples[index];
         }
         break;
-    case 2: {
-        // 2 channel is played as is
+    case 2:
+    case 6: {
+        // 2 and 6 channel is played as is
         samples = std::move(new_samples);
         break;
     }

src/audio_core/cubeb_sink.cpp

@@ -8,6 +8,7 @@
 #include "audio_core/cubeb_sink.h"
 #include "audio_core/stream.h"
 #include "audio_core/time_stretch.h"
+#include "common/assert.h"
 #include "common/logging/log.h"
 #include "common/ring_buffer.h"
 #include "core/settings.h"
@@ -65,12 +66,25 @@ public:
     void EnqueueSamples(u32 source_num_channels, const std::vector<s16>& samples) override {
         if (source_num_channels > num_channels) {
             // Downsample 6 channels to 2
+            ASSERT_MSG(source_num_channels == 6, "Channel count must be 6");
+
             std::vector<s16> buf;
             buf.reserve(samples.size() * num_channels / source_num_channels);
             for (std::size_t i = 0; i < samples.size(); i += source_num_channels) {
-                for (std::size_t ch = 0; ch < num_channels; ch++) {
-                    buf.push_back(samples[i + ch]);
-                }
+                // Downmixing implementation taken from the ATSC standard
+                const s16 left{samples[i + 0]};
+                const s16 right{samples[i + 1]};
+                const s16 center{samples[i + 2]};
+                const s16 surround_left{samples[i + 4]};
+                const s16 surround_right{samples[i + 5]};
+                // Not used in the ATSC reference implementation
+                [[maybe_unused]] const s16 low_frequency_effects { samples[i + 3] };
+
+                constexpr s32 clev{707}; // center mixing level coefficient
+                constexpr s32 slev{707}; // surround mixing level coefficient
+
+                buf.push_back(left + (clev * center / 1000) + (slev * surround_left / 1000));
+                buf.push_back(right + (clev * center / 1000) + (slev * surround_right / 1000));
             }
             queue.Push(buf);
             return;

src/common/math_util.h

@@ -24,17 +24,29 @@ struct Rectangle {
         : left(left), top(top), right(right), bottom(bottom) {}
 
     T GetWidth() const {
-        return std::abs(static_cast<std::make_signed_t<T>>(right - left));
+        if constexpr (std::is_floating_point_v<T>) {
+            return std::abs(right - left);
+        } else {
+            return std::abs(static_cast<std::make_signed_t<T>>(right - left));
+        }
     }
+
     T GetHeight() const {
-        return std::abs(static_cast<std::make_signed_t<T>>(bottom - top));
+        if constexpr (std::is_floating_point_v<T>) {
+            return std::abs(bottom - top);
+        } else {
+            return std::abs(static_cast<std::make_signed_t<T>>(bottom - top));
+        }
     }
+
     Rectangle<T> TranslateX(const T x) const {
         return Rectangle{left + x, top, right + x, bottom};
     }
+
     Rectangle<T> TranslateY(const T y) const {
         return Rectangle{left, top + y, right, bottom + y};
     }
+
     Rectangle<T> Scale(const float s) const {
         return Rectangle{left, top, static_cast<T>(left + GetWidth() * s),
                          static_cast<T>(top + GetHeight() * s)};

src/core/CMakeLists.txt

@@ -131,8 +131,8 @@ add_library(core STATIC
     frontend/framebuffer_layout.cpp
     frontend/framebuffer_layout.h
     frontend/input.h
-    frontend/scope_acquire_window_context.cpp
-    frontend/scope_acquire_window_context.h
+    frontend/scope_acquire_context.cpp
+    frontend/scope_acquire_context.h
     gdbstub/gdbstub.cpp
     gdbstub/gdbstub.h
     hardware_interrupt_manager.cpp
@@ -595,8 +595,12 @@ endif()
 
 if (ARCHITECTURE_x86_64)
     target_sources(core PRIVATE
-        arm/dynarmic/arm_dynarmic.cpp
-        arm/dynarmic/arm_dynarmic.h
+        arm/dynarmic/arm_dynarmic_32.cpp
+        arm/dynarmic/arm_dynarmic_32.h
+        arm/dynarmic/arm_dynarmic_64.cpp
+        arm/dynarmic/arm_dynarmic_64.h
+        arm/dynarmic/arm_dynarmic_cp15.cpp
+        arm/dynarmic/arm_dynarmic_cp15.h
     )
     target_link_libraries(core PRIVATE dynarmic)
 endif()

src/core/arm/arm_interface.h

@@ -25,7 +25,20 @@ public:
     explicit ARM_Interface(System& system_) : system{system_} {}
     virtual ~ARM_Interface() = default;
 
-    struct ThreadContext {
+    struct ThreadContext32 {
+        std::array<u32, 16> cpu_registers;
+        u32 cpsr;
+        std::array<u8, 4> padding;
+        std::array<u64, 32> fprs;
+        u32 fpscr;
+        u32 fpexc;
+        u32 tpidr;
+    };
+    // Internally within the kernel, it expects the AArch32 version of the
+    // thread context to be 344 bytes in size.
+    static_assert(sizeof(ThreadContext32) == 0x158);
+
+    struct ThreadContext64 {
         std::array<u64, 31> cpu_registers;
         u64 sp;
         u64 pc;
@@ -38,7 +51,7 @@ public:
     };
     // Internally within the kernel, it expects the AArch64 version of the
     // thread context to be 800 bytes in size.
-    static_assert(sizeof(ThreadContext) == 0x320);
+    static_assert(sizeof(ThreadContext64) == 0x320);
 
     /// Runs the CPU until an event happens
     virtual void Run() = 0;
@@ -130,17 +143,10 @@ public:
      */
     virtual void SetTPIDR_EL0(u64 value) = 0;
 
-    /**
-     * Saves the current CPU context
-     * @param ctx Thread context to save
-     */
-    virtual void SaveContext(ThreadContext& ctx) = 0;
-
-    /**
-     * Loads a CPU context
-     * @param ctx Thread context to load
-     */
-    virtual void LoadContext(const ThreadContext& ctx) = 0;
+    virtual void SaveContext(ThreadContext32& ctx) = 0;
+    virtual void SaveContext(ThreadContext64& ctx) = 0;
+    virtual void LoadContext(const ThreadContext32& ctx) = 0;
+    virtual void LoadContext(const ThreadContext64& ctx) = 0;
 
     /// Clears the exclusive monitor's state.
     virtual void ClearExclusiveState() = 0;

src/core/arm/dynarmic/arm_dynarmic_32.cpp

@@ -0,0 +1,208 @@
+// Copyright 2020 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <cinttypes>
+#include <memory>
+#include <dynarmic/A32/a32.h>
+#include <dynarmic/A32/config.h>
+#include <dynarmic/A32/context.h>
+#include "common/microprofile.h"
+#include "core/arm/dynarmic/arm_dynarmic_32.h"
+#include "core/arm/dynarmic/arm_dynarmic_64.h"
+#include "core/arm/dynarmic/arm_dynarmic_cp15.h"
+#include "core/core.h"
+#include "core/core_manager.h"
+#include "core/core_timing.h"
+#include "core/hle/kernel/svc.h"
+#include "core/memory.h"
+
+namespace Core {
+
+class DynarmicCallbacks32 : public Dynarmic::A32::UserCallbacks {
+public:
+    explicit DynarmicCallbacks32(ARM_Dynarmic_32& parent) : parent(parent) {}
+
+    u8 MemoryRead8(u32 vaddr) override {
+        return parent.system.Memory().Read8(vaddr);
+    }
+    u16 MemoryRead16(u32 vaddr) override {
+        return parent.system.Memory().Read16(vaddr);
+    }
+    u32 MemoryRead32(u32 vaddr) override {
+        return parent.system.Memory().Read32(vaddr);
+    }
+    u64 MemoryRead64(u32 vaddr) override {
+        return parent.system.Memory().Read64(vaddr);
+    }
+
+    void MemoryWrite8(u32 vaddr, u8 value) override {
+        parent.system.Memory().Write8(vaddr, value);
+    }
+    void MemoryWrite16(u32 vaddr, u16 value) override {
+        parent.system.Memory().Write16(vaddr, value);
+    }
+    void MemoryWrite32(u32 vaddr, u32 value) override {
+        parent.system.Memory().Write32(vaddr, value);
+    }
+    void MemoryWrite64(u32 vaddr, u64 value) override {
+        parent.system.Memory().Write64(vaddr, value);
+    }
+
+    void InterpreterFallback(u32 pc, std::size_t num_instructions) override {
+        UNIMPLEMENTED();
+    }
+
+    void ExceptionRaised(u32 pc, Dynarmic::A32::Exception exception) override {
+        switch (exception) {
+        case Dynarmic::A32::Exception::UndefinedInstruction:
+        case Dynarmic::A32::Exception::UnpredictableInstruction:
+            break;
+        case Dynarmic::A32::Exception::Breakpoint:
+            break;
+        }
+        LOG_CRITICAL(HW_GPU, "ExceptionRaised(exception = {}, pc = {:08X}, code = {:08X})",
+                     static_cast<std::size_t>(exception), pc, MemoryReadCode(pc));
+        UNIMPLEMENTED();
+    }
+
+    void CallSVC(u32 swi) override {
+        Kernel::CallSVC(parent.system, swi);
+    }
+
+    void AddTicks(u64 ticks) override {
+        // Divide the number of ticks by the amount of CPU cores. TODO(Subv): This yields only a
+        // rough approximation of the amount of executed ticks in the system, it may be thrown off
+        // if not all cores are doing a similar amount of work. Instead of doing this, we should
+        // device a way so that timing is consistent across all cores without increasing the ticks 4
+        // times.
+        u64 amortized_ticks = (ticks - num_interpreted_instructions) / Core::NUM_CPU_CORES;
+        // Always execute at least one tick.
+        amortized_ticks = std::max<u64>(amortized_ticks, 1);
+
+        parent.system.CoreTiming().AddTicks(amortized_ticks);
+        num_interpreted_instructions = 0;
+    }
+    u64 GetTicksRemaining() override {
+        return std::max(parent.system.CoreTiming().GetDowncount(), {});
+    }
+
+    ARM_Dynarmic_32& parent;
+    std::size_t num_interpreted_instructions{};
+    u64 tpidrro_el0{};
+    u64 tpidr_el0{};
+};
+
+std::shared_ptr<Dynarmic::A32::Jit> ARM_Dynarmic_32::MakeJit(Common::PageTable& page_table,
+                                                             std::size_t address_space_bits) const {
+    Dynarmic::A32::UserConfig config;
+    config.callbacks = cb.get();
+    // TODO(bunnei): Implement page table for 32-bit
+    // config.page_table = &page_table.pointers;
+    config.coprocessors[15] = std::make_shared<DynarmicCP15>((u32*)&CP15_regs[0]);
+    config.define_unpredictable_behaviour = true;
+    return std::make_unique<Dynarmic::A32::Jit>(config);
+}
+
+MICROPROFILE_DEFINE(ARM_Jit_Dynarmic_32, "ARM JIT", "Dynarmic", MP_RGB(255, 64, 64));
+
+void ARM_Dynarmic_32::Run() {
+    MICROPROFILE_SCOPE(ARM_Jit_Dynarmic_32);
+    jit->Run();
+}
+
+void ARM_Dynarmic_32::Step() {
+    cb->InterpreterFallback(jit->Regs()[15], 1);
+}
+
+ARM_Dynarmic_32::ARM_Dynarmic_32(System& system, ExclusiveMonitor& exclusive_monitor,
+                                 std::size_t core_index)
+    : ARM_Interface{system},
+      cb(std::make_unique<DynarmicCallbacks32>(*this)), core_index{core_index},
+      exclusive_monitor{dynamic_cast<DynarmicExclusiveMonitor&>(exclusive_monitor)} {}
+
+ARM_Dynarmic_32::~ARM_Dynarmic_32() = default;
+
+void ARM_Dynarmic_32::SetPC(u64 pc) {
+    jit->Regs()[15] = static_cast<u32>(pc);
+}
+
+u64 ARM_Dynarmic_32::GetPC() const {
+    return jit->Regs()[15];
+}
+
+u64 ARM_Dynarmic_32::GetReg(int index) const {
+    return jit->Regs()[index];
+}
+
+void ARM_Dynarmic_32::SetReg(int index, u64 value) {
+    jit->Regs()[index] = static_cast<u32>(value);
+}
+
+u128 ARM_Dynarmic_32::GetVectorReg(int index) const {
+    return {};
+}
+
+void ARM_Dynarmic_32::SetVectorReg(int index, u128 value) {}
+
+u32 ARM_Dynarmic_32::GetPSTATE() const {
+    return jit->Cpsr();
+}
+
+void ARM_Dynarmic_32::SetPSTATE(u32 cpsr) {
+    jit->SetCpsr(cpsr);
+}
+
+u64 ARM_Dynarmic_32::GetTlsAddress() const {
+    return CP15_regs[static_cast<std::size_t>(CP15Register::CP15_THREAD_URO)];
+}
+
+void ARM_Dynarmic_32::SetTlsAddress(VAddr address) {
+    CP15_regs[static_cast<std::size_t>(CP15Register::CP15_THREAD_URO)] = static_cast<u32>(address);
+}
+
+u64 ARM_Dynarmic_32::GetTPIDR_EL0() const {
+    return cb->tpidr_el0;
+}
+
+void ARM_Dynarmic_32::SetTPIDR_EL0(u64 value) {
+    cb->tpidr_el0 = value;
+}
+
+void ARM_Dynarmic_32::SaveContext(ThreadContext32& ctx) {
+    Dynarmic::A32::Context context;
+    jit->SaveContext(context);
+    ctx.cpu_registers = context.Regs();
+    ctx.cpsr = context.Cpsr();
+}
+
+void ARM_Dynarmic_32::LoadContext(const ThreadContext32& ctx) {
+    Dynarmic::A32::Context context;
+    context.Regs() = ctx.cpu_registers;
+    context.SetCpsr(ctx.cpsr);
+    jit->LoadContext(context);
+}
+
+void ARM_Dynarmic_32::PrepareReschedule() {
+    jit->HaltExecution();
+}
+
+void ARM_Dynarmic_32::ClearInstructionCache() {
+    jit->ClearCache();
+}
+
+void ARM_Dynarmic_32::ClearExclusiveState() {}
+
+void ARM_Dynarmic_32::PageTableChanged(Common::PageTable& page_table,
+                                       std::size_t new_address_space_size_in_bits) {
+    auto key = std::make_pair(&page_table, new_address_space_size_in_bits);
+    auto iter = jit_cache.find(key);
+    if (iter != jit_cache.end()) {
+        jit = iter->second;
+        return;
+    }
+    jit = MakeJit(page_table, new_address_space_size_in_bits);
+    jit_cache.emplace(key, jit);
+}
+
+} // namespace Core

src/core/arm/dynarmic/arm_dynarmic_32.h

@@ -0,0 +1,77 @@
+// Copyright 2020 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include <unordered_map>
+
+#include <dynarmic/A32/a32.h>
+#include <dynarmic/A64/a64.h>
+#include <dynarmic/A64/exclusive_monitor.h>
+#include "common/common_types.h"
+#include "common/hash.h"
+#include "core/arm/arm_interface.h"
+#include "core/arm/exclusive_monitor.h"
+
+namespace Memory {
+class Memory;
+}
+
+namespace Core {
+
+class DynarmicCallbacks32;
+class DynarmicExclusiveMonitor;
+class System;
+
+class ARM_Dynarmic_32 final : public ARM_Interface {
+public:
+    ARM_Dynarmic_32(System& system, ExclusiveMonitor& exclusive_monitor, std::size_t core_index);
+    ~ARM_Dynarmic_32() override;
+
+    void SetPC(u64 pc) override;
+    u64 GetPC() const override;
+    u64 GetReg(int index) const override;
+    void SetReg(int index, u64 value) override;
+    u128 GetVectorReg(int index) const override;
+    void SetVectorReg(int index, u128 value) override;
+    u32 GetPSTATE() const override;
+    void SetPSTATE(u32 pstate) override;
+    void Run() override;
+    void Step() override;
+    VAddr GetTlsAddress() const override;
+    void SetTlsAddress(VAddr address) override;
+    void SetTPIDR_EL0(u64 value) override;
+    u64 GetTPIDR_EL0() const override;
+
+    void SaveContext(ThreadContext32& ctx) override;
+    void SaveContext(ThreadContext64& ctx) override {}
+    void LoadContext(const ThreadContext32& ctx) override;
+    void LoadContext(const ThreadContext64& ctx) override {}
+
+    void PrepareReschedule() override;
+    void ClearExclusiveState() override;
+
+    void ClearInstructionCache() override;
+    void PageTableChanged(Common::PageTable& new_page_table,
+                          std::size_t new_address_space_size_in_bits) override;
+
+private:
+    std::shared_ptr<Dynarmic::A32::Jit> MakeJit(Common::PageTable& page_table,
+                                                std::size_t address_space_bits) const;
+
+    using JitCacheKey = std::pair<Common::PageTable*, std::size_t>;
+    using JitCacheType =
+        std::unordered_map<JitCacheKey, std::shared_ptr<Dynarmic::A32::Jit>, Common::PairHash>;
+
+    friend class DynarmicCallbacks32;
+    std::unique_ptr<DynarmicCallbacks32> cb;
+    JitCacheType jit_cache;
+    std::shared_ptr<Dynarmic::A32::Jit> jit;
+    std::size_t core_index;
+    DynarmicExclusiveMonitor& exclusive_monitor;
+    std::array<u32, 84> CP15_regs{};
+};
+
+} // namespace Core

src/core/arm/dynarmic/arm_dynarmic_64.cpp

@@ -8,7 +8,7 @@
 #include <dynarmic/A64/config.h>
 #include "common/logging/log.h"
 #include "common/microprofile.h"
-#include "core/arm/dynarmic/arm_dynarmic.h"
+#include "core/arm/dynarmic/arm_dynarmic_64.h"
 #include "core/core.h"
 #include "core/core_manager.h"
 #include "core/core_timing.h"
@@ -25,9 +25,9 @@ namespace Core {
 
 using Vector = Dynarmic::A64::Vector;
 
-class ARM_Dynarmic_Callbacks : public Dynarmic::A64::UserCallbacks {
+class DynarmicCallbacks64 : public Dynarmic::A64::UserCallbacks {
 public:
-    explicit ARM_Dynarmic_Callbacks(ARM_Dynarmic& parent) : parent(parent) {}
+    explicit DynarmicCallbacks64(ARM_Dynarmic_64& parent) : parent(parent) {}
 
     u8 MemoryRead8(u64 vaddr) override {
         return parent.system.Memory().Read8(vaddr);
@@ -68,7 +68,7 @@ public:
         LOG_INFO(Core_ARM, "Unicorn fallback @ 0x{:X} for {} instructions (instr = {:08X})", pc,
                  num_instructions, MemoryReadCode(pc));
 
-        ARM_Interface::ThreadContext ctx;
+        ARM_Interface::ThreadContext64 ctx;
         parent.SaveContext(ctx);
         parent.inner_unicorn.LoadContext(ctx);
         parent.inner_unicorn.ExecuteInstructions(num_instructions);
@@ -90,7 +90,7 @@ public:
                 parent.jit->HaltExecution();
                 parent.SetPC(pc);
                 Kernel::Thread* const thread = parent.system.CurrentScheduler().GetCurrentThread();
-                parent.SaveContext(thread->GetContext());
+                parent.SaveContext(thread->GetContext64());
                 GDBStub::Break();
                 GDBStub::SendTrap(thread, 5);
                 return;
@@ -126,14 +126,14 @@ public:
         return Timing::CpuCyclesToClockCycles(parent.system.CoreTiming().GetTicks());
     }
 
-    ARM_Dynarmic& parent;
+    ARM_Dynarmic_64& parent;
     std::size_t num_interpreted_instructions = 0;
     u64 tpidrro_el0 = 0;
     u64 tpidr_el0 = 0;
 };
 
-std::unique_ptr<Dynarmic::A64::Jit> ARM_Dynarmic::MakeJit(Common::PageTable& page_table,
-                                                          std::size_t address_space_bits) const {
+std::shared_ptr<Dynarmic::A64::Jit> ARM_Dynarmic_64::MakeJit(Common::PageTable& page_table,
+                                                             std::size_t address_space_bits) const {
     Dynarmic::A64::UserConfig config;
 
     // Callbacks
@@ -159,79 +159,79 @@ std::unique_ptr<Dynarmic::A64::Jit> ARM_Dynarmic::MakeJit(Common::PageTable& pag
     // Unpredictable instructions
     config.define_unpredictable_behaviour = true;
 
-    return std::make_unique<Dynarmic::A64::Jit>(config);
+    return std::make_shared<Dynarmic::A64::Jit>(config);
 }
 
-MICROPROFILE_DEFINE(ARM_Jit_Dynarmic, "ARM JIT", "Dynarmic", MP_RGB(255, 64, 64));
+MICROPROFILE_DEFINE(ARM_Jit_Dynarmic_64, "ARM JIT", "Dynarmic", MP_RGB(255, 64, 64));
 
-void ARM_Dynarmic::Run() {
-    MICROPROFILE_SCOPE(ARM_Jit_Dynarmic);
+void ARM_Dynarmic_64::Run() {
+    MICROPROFILE_SCOPE(ARM_Jit_Dynarmic_64);
 
     jit->Run();
 }
 
-void ARM_Dynarmic::Step() {
+void ARM_Dynarmic_64::Step() {
     cb->InterpreterFallback(jit->GetPC(), 1);
 }
 
-ARM_Dynarmic::ARM_Dynarmic(System& system, ExclusiveMonitor& exclusive_monitor,
-                           std::size_t core_index)
+ARM_Dynarmic_64::ARM_Dynarmic_64(System& system, ExclusiveMonitor& exclusive_monitor,
+                                 std::size_t core_index)
     : ARM_Interface{system},
-      cb(std::make_unique<ARM_Dynarmic_Callbacks>(*this)), inner_unicorn{system},
+      cb(std::make_unique<DynarmicCallbacks64>(*this)), inner_unicorn{system},
       core_index{core_index}, exclusive_monitor{
                                   dynamic_cast<DynarmicExclusiveMonitor&>(exclusive_monitor)} {}
 
-ARM_Dynarmic::~ARM_Dynarmic() = default;
+ARM_Dynarmic_64::~ARM_Dynarmic_64() = default;
 
-void ARM_Dynarmic::SetPC(u64 pc) {
+void ARM_Dynarmic_64::SetPC(u64 pc) {
     jit->SetPC(pc);
 }
 
-u64 ARM_Dynarmic::GetPC() const {
+u64 ARM_Dynarmic_64::GetPC() const {
     return jit->GetPC();
 }
 
-u64 ARM_Dynarmic::GetReg(int index) const {
+u64 ARM_Dynarmic_64::GetReg(int index) const {
     return jit->GetRegister(index);
 }
 
-void ARM_Dynarmic::SetReg(int index, u64 value) {
+void ARM_Dynarmic_64::SetReg(int index, u64 value) {
     jit->SetRegister(index, value);
 }
 
-u128 ARM_Dynarmic::GetVectorReg(int index) const {
+u128 ARM_Dynarmic_64::GetVectorReg(int index) const {
     return jit->GetVector(index);
 }
 
-void ARM_Dynarmic::SetVectorReg(int index, u128 value) {
+void ARM_Dynarmic_64::SetVectorReg(int index, u128 value) {
     jit->SetVector(index, value);
 }
 
-u32 ARM_Dynarmic::GetPSTATE() const {
+u32 ARM_Dynarmic_64::GetPSTATE() const {
     return jit->GetPstate();
 }
 
-void ARM_Dynarmic::SetPSTATE(u32 pstate) {
+void ARM_Dynarmic_64::SetPSTATE(u32 pstate) {
     jit->SetPstate(pstate);
 }
 
-u64 ARM_Dynarmic::GetTlsAddress() const {
+u64 ARM_Dynarmic_64::GetTlsAddress() const {
     return cb->tpidrro_el0;
 }
 
-void ARM_Dynarmic::SetTlsAddress(VAddr address) {
+void ARM_Dynarmic_64::SetTlsAddress(VAddr address) {
     cb->tpidrro_el0 = address;
 }
 
-u64 ARM_Dynarmic::GetTPIDR_EL0() const {
+u64 ARM_Dynarmic_64::GetTPIDR_EL0() const {
     return cb->tpidr_el0;
 }
 
-void ARM_Dynarmic::SetTPIDR_EL0(u64 value) {
+void ARM_Dynarmic_64::SetTPIDR_EL0(u64 value) {
     cb->tpidr_el0 = value;
 }
 
-void ARM_Dynarmic::SaveContext(ThreadContext& ctx) {
+void ARM_Dynarmic_64::SaveContext(ThreadContext64& ctx) {
     ctx.cpu_registers = jit->GetRegisters();
     ctx.sp = jit->GetSP();
     ctx.pc = jit->GetPC();
@@ -242,7 +242,7 @@ void ARM_Dynarmic::SaveContext(ThreadContext& ctx) {
     ctx.tpidr = cb->tpidr_el0;
 }
 
-void ARM_Dynarmic::LoadContext(const ThreadContext& ctx) {
+void ARM_Dynarmic_64::LoadContext(const ThreadContext64& ctx) {
     jit->SetRegisters(ctx.cpu_registers);
     jit->SetSP(ctx.sp);
     jit->SetPC(ctx.pc);
@@ -253,25 +253,32 @@ void ARM_Dynarmic::LoadContext(const ThreadContext& ctx) {
     SetTPIDR_EL0(ctx.tpidr);
 }
 
-void ARM_Dynarmic::PrepareReschedule() {
+void ARM_Dynarmic_64::PrepareReschedule() {
     jit->HaltExecution();
 }
 
-void ARM_Dynarmic::ClearInstructionCache() {
+void ARM_Dynarmic_64::ClearInstructionCache() {
     jit->ClearCache();
 }
 
-void ARM_Dynarmic::ClearExclusiveState() {
+void ARM_Dynarmic_64::ClearExclusiveState() {
     jit->ClearExclusiveState();
 }
 
-void ARM_Dynarmic::PageTableChanged(Common::PageTable& page_table,
-                                    std::size_t new_address_space_size_in_bits) {
+void ARM_Dynarmic_64::PageTableChanged(Common::PageTable& page_table,
+                                       std::size_t new_address_space_size_in_bits) {
+    auto key = std::make_pair(&page_table, new_address_space_size_in_bits);
+    auto iter = jit_cache.find(key);
+    if (iter != jit_cache.end()) {
+        jit = iter->second;
+        return;
+    }
     jit = MakeJit(page_table, new_address_space_size_in_bits);
+    jit_cache.emplace(key, jit);
 }
 
-DynarmicExclusiveMonitor::DynarmicExclusiveMonitor(Memory::Memory& memory_, std::size_t core_count)
-    : monitor(core_count), memory{memory_} {}
+DynarmicExclusiveMonitor::DynarmicExclusiveMonitor(Memory::Memory& memory, std::size_t core_count)
+    : monitor(core_count), memory{memory} {}
 
 DynarmicExclusiveMonitor::~DynarmicExclusiveMonitor() = default;
 

src/core/arm/dynarmic/arm_dynarmic_64.h

@@ -5,9 +5,12 @@
 #pragma once
 
 #include <memory>
+#include <unordered_map>
+
 #include <dynarmic/A64/a64.h>
 #include <dynarmic/A64/exclusive_monitor.h>
 #include "common/common_types.h"
+#include "common/hash.h"
 #include "core/arm/arm_interface.h"
 #include "core/arm/exclusive_monitor.h"
 #include "core/arm/unicorn/arm_unicorn.h"
@@ -18,14 +21,14 @@ class Memory;
 
 namespace Core {
 
-class ARM_Dynarmic_Callbacks;
+class DynarmicCallbacks64;
 class DynarmicExclusiveMonitor;
 class System;
 
-class ARM_Dynarmic final : public ARM_Interface {
+class ARM_Dynarmic_64 final : public ARM_Interface {
 public:
-    ARM_Dynarmic(System& system, ExclusiveMonitor& exclusive_monitor, std::size_t core_index);
-    ~ARM_Dynarmic() override;
+    ARM_Dynarmic_64(System& system, ExclusiveMonitor& exclusive_monitor, std::size_t core_index);
+    ~ARM_Dynarmic_64() override;
 
     void SetPC(u64 pc) override;
     u64 GetPC() const override;
@@ -42,8 +45,10 @@ public:
     void SetTPIDR_EL0(u64 value) override;
     u64 GetTPIDR_EL0() const override;
 
-    void SaveContext(ThreadContext& ctx) override;
-    void LoadContext(const ThreadContext& ctx) override;
+    void SaveContext(ThreadContext32& ctx) override {}
+    void SaveContext(ThreadContext64& ctx) override;
+    void LoadContext(const ThreadContext32& ctx) override {}
+    void LoadContext(const ThreadContext64& ctx) override;
 
     void PrepareReschedule() override;
     void ClearExclusiveState() override;
@@ -53,12 +58,17 @@ public:
                           std::size_t new_address_space_size_in_bits) override;
 
 private:
-    std::unique_ptr<Dynarmic::A64::Jit> MakeJit(Common::PageTable& page_table,
+    std::shared_ptr<Dynarmic::A64::Jit> MakeJit(Common::PageTable& page_table,
                                                 std::size_t address_space_bits) const;
 
-    friend class ARM_Dynarmic_Callbacks;
-    std::unique_ptr<ARM_Dynarmic_Callbacks> cb;
-    std::unique_ptr<Dynarmic::A64::Jit> jit;
+    using JitCacheKey = std::pair<Common::PageTable*, std::size_t>;
+    using JitCacheType =
+        std::unordered_map<JitCacheKey, std::shared_ptr<Dynarmic::A64::Jit>, Common::PairHash>;
+
+    friend class DynarmicCallbacks64;
+    std::unique_ptr<DynarmicCallbacks64> cb;
+    JitCacheType jit_cache;
+    std::shared_ptr<Dynarmic::A64::Jit> jit;
     ARM_Unicorn inner_unicorn;
 
     std::size_t core_index;
@@ -67,7 +77,7 @@ private:
 
 class DynarmicExclusiveMonitor final : public ExclusiveMonitor {
 public:
-    explicit DynarmicExclusiveMonitor(Memory::Memory& memory_, std::size_t core_count);
+    explicit DynarmicExclusiveMonitor(Memory::Memory& memory, std::size_t core_count);
     ~DynarmicExclusiveMonitor() override;
 
     void SetExclusive(std::size_t core_index, VAddr addr) override;
@@ -80,7 +90,7 @@ public:
     bool ExclusiveWrite128(std::size_t core_index, VAddr vaddr, u128 value) override;
 
 private:
-    friend class ARM_Dynarmic;
+    friend class ARM_Dynarmic_64;
     Dynarmic::A64::ExclusiveMonitor monitor;
     Memory::Memory& memory;
 };

src/core/arm/dynarmic/arm_dynarmic_cp15.cpp

@@ -0,0 +1,80 @@
+// Copyright 2017 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "core/arm/dynarmic/arm_dynarmic_cp15.h"
+
+using Callback = Dynarmic::A32::Coprocessor::Callback;
+using CallbackOrAccessOneWord = Dynarmic::A32::Coprocessor::CallbackOrAccessOneWord;
+using CallbackOrAccessTwoWords = Dynarmic::A32::Coprocessor::CallbackOrAccessTwoWords;
+
+std::optional<Callback> DynarmicCP15::CompileInternalOperation(bool two, unsigned opc1,
+                                                               CoprocReg CRd, CoprocReg CRn,
+                                                               CoprocReg CRm, unsigned opc2) {
+    return {};
+}
+
+CallbackOrAccessOneWord DynarmicCP15::CompileSendOneWord(bool two, unsigned opc1, CoprocReg CRn,
+                                                         CoprocReg CRm, unsigned opc2) {
+    // TODO(merry): Privileged CP15 registers
+
+    if (!two && CRn == CoprocReg::C7 && opc1 == 0 && CRm == CoprocReg::C5 && opc2 == 4) {
+        // This is a dummy write, we ignore the value written here.
+        return &CP15[static_cast<std::size_t>(CP15Register::CP15_FLUSH_PREFETCH_BUFFER)];
+    }
+
+    if (!two && CRn == CoprocReg::C7 && opc1 == 0 && CRm == CoprocReg::C10) {
+        switch (opc2) {
+        case 4:
+            // This is a dummy write, we ignore the value written here.
+            return &CP15[static_cast<std::size_t>(CP15Register::CP15_DATA_SYNC_BARRIER)];
+        case 5:
+            // This is a dummy write, we ignore the value written here.
+            return &CP15[static_cast<std::size_t>(CP15Register::CP15_DATA_MEMORY_BARRIER)];
+        default:
+            return {};
+        }
+    }
+
+    if (!two && CRn == CoprocReg::C13 && opc1 == 0 && CRm == CoprocReg::C0 && opc2 == 2) {
+        return &CP15[static_cast<std::size_t>(CP15Register::CP15_THREAD_UPRW)];
+    }
+
+    return {};
+}
+
+CallbackOrAccessTwoWords DynarmicCP15::CompileSendTwoWords(bool two, unsigned opc, CoprocReg CRm) {
+    return {};
+}
+
+CallbackOrAccessOneWord DynarmicCP15::CompileGetOneWord(bool two, unsigned opc1, CoprocReg CRn,
+                                                        CoprocReg CRm, unsigned opc2) {
+    // TODO(merry): Privileged CP15 registers
+
+    if (!two && CRn == CoprocReg::C13 && opc1 == 0 && CRm == CoprocReg::C0) {
+        switch (opc2) {
+        case 2:
+            return &CP15[static_cast<std::size_t>(CP15Register::CP15_THREAD_UPRW)];
+        case 3:
+            return &CP15[static_cast<std::size_t>(CP15Register::CP15_THREAD_URO)];
+        default:
+            return {};
+        }
+    }
+
+    return {};
+}
+
+CallbackOrAccessTwoWords DynarmicCP15::CompileGetTwoWords(bool two, unsigned opc, CoprocReg CRm) {
+    return {};
+}
+
+std::optional<Callback> DynarmicCP15::CompileLoadWords(bool two, bool long_transfer, CoprocReg CRd,
+                                                       std::optional<u8> option) {
+    return {};
+}
+
+std::optional<Callback> DynarmicCP15::CompileStoreWords(bool two, bool long_transfer, CoprocReg CRd,
+                                                        std::optional<u8> option) {
+    return {};
+}

src/core/arm/dynarmic/arm_dynarmic_cp15.h

@@ -0,0 +1,152 @@
+// Copyright 2017 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include <optional>
+
+#include <dynarmic/A32/coprocessor.h>
+#include "common/common_types.h"
+
+enum class CP15Register {
+    // c0 - Information registers
+    CP15_MAIN_ID,
+    CP15_CACHE_TYPE,
+    CP15_TCM_STATUS,
+    CP15_TLB_TYPE,
+    CP15_CPU_ID,
+    CP15_PROCESSOR_FEATURE_0,
+    CP15_PROCESSOR_FEATURE_1,
+    CP15_DEBUG_FEATURE_0,
+    CP15_AUXILIARY_FEATURE_0,
+    CP15_MEMORY_MODEL_FEATURE_0,
+    CP15_MEMORY_MODEL_FEATURE_1,
+    CP15_MEMORY_MODEL_FEATURE_2,
+    CP15_MEMORY_MODEL_FEATURE_3,
+    CP15_ISA_FEATURE_0,
+    CP15_ISA_FEATURE_1,
+    CP15_ISA_FEATURE_2,
+    CP15_ISA_FEATURE_3,
+    CP15_ISA_FEATURE_4,
+
+    // c1 - Control registers
+    CP15_CONTROL,
+    CP15_AUXILIARY_CONTROL,
+    CP15_COPROCESSOR_ACCESS_CONTROL,
+
+    // c2 - Translation table registers
+    CP15_TRANSLATION_BASE_TABLE_0,
+    CP15_TRANSLATION_BASE_TABLE_1,
+    CP15_TRANSLATION_BASE_CONTROL,
+    CP15_DOMAIN_ACCESS_CONTROL,
+    CP15_RESERVED,
+
+    // c5 - Fault status registers
+    CP15_FAULT_STATUS,
+    CP15_INSTR_FAULT_STATUS,
+    CP15_COMBINED_DATA_FSR = CP15_FAULT_STATUS,
+    CP15_INST_FSR,
+
+    // c6 - Fault Address registers
+    CP15_FAULT_ADDRESS,
+    CP15_COMBINED_DATA_FAR = CP15_FAULT_ADDRESS,
+    CP15_WFAR,
+    CP15_IFAR,
+
+    // c7 - Cache operation registers
+    CP15_WAIT_FOR_INTERRUPT,
+    CP15_PHYS_ADDRESS,
+    CP15_INVALIDATE_INSTR_CACHE,
+    CP15_INVALIDATE_INSTR_CACHE_USING_MVA,
+    CP15_INVALIDATE_INSTR_CACHE_USING_INDEX,
+    CP15_FLUSH_PREFETCH_BUFFER,
+    CP15_FLUSH_BRANCH_TARGET_CACHE,
+    CP15_FLUSH_BRANCH_TARGET_CACHE_ENTRY,
+    CP15_INVALIDATE_DATA_CACHE,
+    CP15_INVALIDATE_DATA_CACHE_LINE_USING_MVA,
+    CP15_INVALIDATE_DATA_CACHE_LINE_USING_INDEX,
+    CP15_INVALIDATE_DATA_AND_INSTR_CACHE,
+    CP15_CLEAN_DATA_CACHE,
+    CP15_CLEAN_DATA_CACHE_LINE_USING_MVA,
+    CP15_CLEAN_DATA_CACHE_LINE_USING_INDEX,
+    CP15_DATA_SYNC_BARRIER,
+    CP15_DATA_MEMORY_BARRIER,
+    CP15_CLEAN_AND_INVALIDATE_DATA_CACHE,
+    CP15_CLEAN_AND_INVALIDATE_DATA_CACHE_LINE_USING_MVA,
+    CP15_CLEAN_AND_INVALIDATE_DATA_CACHE_LINE_USING_INDEX,
+
+    // c8 - TLB operations
+    CP15_INVALIDATE_ITLB,
+    CP15_INVALIDATE_ITLB_SINGLE_ENTRY,
+    CP15_INVALIDATE_ITLB_ENTRY_ON_ASID_MATCH,
+    CP15_INVALIDATE_ITLB_ENTRY_ON_MVA,
+    CP15_INVALIDATE_DTLB,
+    CP15_INVALIDATE_DTLB_SINGLE_ENTRY,
+    CP15_INVALIDATE_DTLB_ENTRY_ON_ASID_MATCH,
+    CP15_INVALIDATE_DTLB_ENTRY_ON_MVA,
+    CP15_INVALIDATE_UTLB,
+    CP15_INVALIDATE_UTLB_SINGLE_ENTRY,
+    CP15_INVALIDATE_UTLB_ENTRY_ON_ASID_MATCH,
+    CP15_INVALIDATE_UTLB_ENTRY_ON_MVA,
+
+    // c9 - Data cache lockdown register
+    CP15_DATA_CACHE_LOCKDOWN,
+
+    // c10 - TLB/Memory map registers
+    CP15_TLB_LOCKDOWN,
+    CP15_PRIMARY_REGION_REMAP,
+    CP15_NORMAL_REGION_REMAP,
+
+    // c13 - Thread related registers
+    CP15_PID,
+    CP15_CONTEXT_ID,
+    CP15_THREAD_UPRW, // Thread ID register - User/Privileged Read/Write
+    CP15_THREAD_URO,  // Thread ID register - User Read Only (Privileged R/W)
+    CP15_THREAD_PRW,  // Thread ID register - Privileged R/W only.
+
+    // c15 - Performance and TLB lockdown registers
+    CP15_PERFORMANCE_MONITOR_CONTROL,
+    CP15_CYCLE_COUNTER,
+    CP15_COUNT_0,
+    CP15_COUNT_1,
+    CP15_READ_MAIN_TLB_LOCKDOWN_ENTRY,
+    CP15_WRITE_MAIN_TLB_LOCKDOWN_ENTRY,
+    CP15_MAIN_TLB_LOCKDOWN_VIRT_ADDRESS,
+    CP15_MAIN_TLB_LOCKDOWN_PHYS_ADDRESS,
+    CP15_MAIN_TLB_LOCKDOWN_ATTRIBUTE,
+    CP15_TLB_DEBUG_CONTROL,
+
+    // Skyeye defined
+    CP15_TLB_FAULT_ADDR,
+    CP15_TLB_FAULT_STATUS,
+
+    // Not an actual register.
+    // All registers should be defined above this.
+    CP15_REGISTER_COUNT,
+};
+
+class DynarmicCP15 final : public Dynarmic::A32::Coprocessor {
+public:
+    using CoprocReg = Dynarmic::A32::CoprocReg;
+
+    explicit DynarmicCP15(u32* cp15) : CP15(cp15){};
+
+    std::optional<Callback> CompileInternalOperation(bool two, unsigned opc1, CoprocReg CRd,
+                                                     CoprocReg CRn, CoprocReg CRm,
+                                                     unsigned opc2) override;
+    CallbackOrAccessOneWord CompileSendOneWord(bool two, unsigned opc1, CoprocReg CRn,
+                                               CoprocReg CRm, unsigned opc2) override;
+    CallbackOrAccessTwoWords CompileSendTwoWords(bool two, unsigned opc, CoprocReg CRm) override;
+    CallbackOrAccessOneWord CompileGetOneWord(bool two, unsigned opc1, CoprocReg CRn, CoprocReg CRm,
+                                              unsigned opc2) override;
+    CallbackOrAccessTwoWords CompileGetTwoWords(bool two, unsigned opc, CoprocReg CRm) override;
+    std::optional<Callback> CompileLoadWords(bool two, bool long_transfer, CoprocReg CRd,
+                                             std::optional<u8> option) override;
+    std::optional<Callback> CompileStoreWords(bool two, bool long_transfer, CoprocReg CRd,
+                                              std::optional<u8> option) override;
+
+private:
+    u32* CP15{};
+};

src/core/arm/exclusive_monitor.cpp

@@ -3,7 +3,7 @@
 // Refer to the license.txt file included.
 
 #ifdef ARCHITECTURE_x86_64
-#include "core/arm/dynarmic/arm_dynarmic.h"
+#include "core/arm/dynarmic/arm_dynarmic_64.h"
 #endif
 #include "core/arm/exclusive_monitor.h"
 #include "core/memory.h"

src/core/arm/unicorn/arm_unicorn.cpp

@@ -53,7 +53,7 @@ static bool UnmappedMemoryHook(uc_engine* uc, uc_mem_type type, u64 addr, int si
                                void* user_data) {
     auto* const system = static_cast<System*>(user_data);
 
-    ARM_Interface::ThreadContext ctx{};
+    ARM_Interface::ThreadContext64 ctx{};
     system->CurrentArmInterface().SaveContext(ctx);
     ASSERT_MSG(false, "Attempted to read from unmapped memory: 0x{:X}, pc=0x{:X}, lr=0x{:X}", addr,
                ctx.pc, ctx.cpu_registers[30]);
@@ -179,7 +179,7 @@ void ARM_Unicorn::ExecuteInstructions(std::size_t num_instructions) {
         }
 
         Kernel::Thread* const thread = system.CurrentScheduler().GetCurrentThread();
-        SaveContext(thread->GetContext());
+        SaveContext(thread->GetContext64());
         if (last_bkpt_hit || GDBStub::IsMemoryBreak() || GDBStub::GetCpuStepFlag()) {
             last_bkpt_hit = false;
             GDBStub::Break();
@@ -188,7 +188,7 @@ void ARM_Unicorn::ExecuteInstructions(std::size_t num_instructions) {
     }
 }
 
-void ARM_Unicorn::SaveContext(ThreadContext& ctx) {
+void ARM_Unicorn::SaveContext(ThreadContext64& ctx) {
     int uregs[32];
     void* tregs[32];
 
@@ -215,7 +215,7 @@ void ARM_Unicorn::SaveContext(ThreadContext& ctx) {
     CHECKED(uc_reg_read_batch(uc, uregs, tregs, 32));
 }
 
-void ARM_Unicorn::LoadContext(const ThreadContext& ctx) {
+void ARM_Unicorn::LoadContext(const ThreadContext64& ctx) {
     int uregs[32];
     void* tregs[32];
 

src/core/arm/unicorn/arm_unicorn.h

@@ -30,8 +30,6 @@ public:
     void SetTlsAddress(VAddr address) override;
     void SetTPIDR_EL0(u64 value) override;
     u64 GetTPIDR_EL0() const override;
-    void SaveContext(ThreadContext& ctx) override;
-    void LoadContext(const ThreadContext& ctx) override;
     void PrepareReschedule() override;
     void ClearExclusiveState() override;
     void ExecuteInstructions(std::size_t num_instructions);
@@ -41,6 +39,11 @@ public:
     void PageTableChanged(Common::PageTable&, std::size_t) override {}
     void RecordBreak(GDBStub::BreakpointAddress bkpt);
 
+    void SaveContext(ThreadContext32& ctx) override {}
+    void SaveContext(ThreadContext64& ctx) override;
+    void LoadContext(const ThreadContext32& ctx) override {}
+    void LoadContext(const ThreadContext64& ctx) override;
+
 private:
     static void InterruptHook(uc_engine* uc, u32 int_no, void* user_data);
 

src/core/core.cpp

@@ -24,6 +24,7 @@
 #include "core/file_sys/sdmc_factory.h"
 #include "core/file_sys/vfs_concat.h"
 #include "core/file_sys/vfs_real.h"
+#include "core/frontend/scope_acquire_context.h"
 #include "core/gdbstub/gdbstub.h"
 #include "core/hardware_interrupt_manager.h"
 #include "core/hle/kernel/client_port.h"
@@ -173,6 +174,7 @@ struct System::Impl {
         }
         interrupt_manager = std::make_unique<Core::Hardware::InterruptManager>(system);
         gpu_core = VideoCore::CreateGPU(system);
+        renderer->Rasterizer().SetupDirtyFlags();
 
         is_powered_on = true;
         exit_lock = false;
@@ -184,6 +186,8 @@ struct System::Impl {
 
     ResultStatus Load(System& system, Frontend::EmuWindow& emu_window,
                       const std::string& filepath) {
+        Core::Frontend::ScopeAcquireContext acquire_context{emu_window};
+
         app_loader = Loader::GetLoader(GetGameFileFromPath(virtual_filesystem, filepath));
         if (!app_loader) {
             LOG_CRITICAL(Core, "Failed to obtain loader for {}!", filepath);

src/core/core_manager.cpp

@@ -6,9 +6,6 @@
 #include <mutex>
 
 #include "common/logging/log.h"
-#ifdef ARCHITECTURE_x86_64
-#include "core/arm/dynarmic/arm_dynarmic.h"
-#endif
 #include "core/arm/exclusive_monitor.h"
 #include "core/arm/unicorn/arm_unicorn.h"
 #include "core/core.h"

src/core/frontend/emu_window.h

@@ -26,9 +26,6 @@ public:
 
     /// Releases (dunno if this is the "right" word) the context from the caller thread
     virtual void DoneCurrent() = 0;
-
-    /// Swap buffers to display the next frame
-    virtual void SwapBuffers() = 0;
 };
 
 /**

src/core/frontend/framebuffer_layout.h

@@ -29,6 +29,7 @@ enum class AspectRatio {
 struct FramebufferLayout {
     u32 width{ScreenUndocked::Width};
     u32 height{ScreenUndocked::Height};
+    bool is_srgb{};
 
     Common::Rectangle<u32> screen;
 

src/core/frontend/scope_acquire_context.cpp

@@ -0,0 +1,18 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "core/frontend/emu_window.h"
+#include "core/frontend/scope_acquire_context.h"
+
+namespace Core::Frontend {
+
+ScopeAcquireContext::ScopeAcquireContext(Core::Frontend::GraphicsContext& context)
+    : context{context} {
+    context.MakeCurrent();
+}
+ScopeAcquireContext::~ScopeAcquireContext() {
+    context.DoneCurrent();
+}
+
+} // namespace Core::Frontend

src/core/frontend/scope_acquire_context.h

@@ -8,16 +8,16 @@
 
 namespace Core::Frontend {
 
-class EmuWindow;
+class GraphicsContext;
 
 /// Helper class to acquire/release window context within a given scope
-class ScopeAcquireWindowContext : NonCopyable {
+class ScopeAcquireContext : NonCopyable {
 public:
-    explicit ScopeAcquireWindowContext(Core::Frontend::EmuWindow& window);
-    ~ScopeAcquireWindowContext();
+    explicit ScopeAcquireContext(Core::Frontend::GraphicsContext& context);
+    ~ScopeAcquireContext();
 
 private:
-    Core::Frontend::EmuWindow& emu_window;
+    Core::Frontend::GraphicsContext& context;
 };
 
 } // namespace Core::Frontend

src/core/frontend/scope_acquire_window_context.cpp

@@ -1,18 +0,0 @@
-// Copyright 2019 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#include "core/frontend/emu_window.h"
-#include "core/frontend/scope_acquire_window_context.h"
-
-namespace Core::Frontend {
-
-ScopeAcquireWindowContext::ScopeAcquireWindowContext(Core::Frontend::EmuWindow& emu_window_)
-    : emu_window{emu_window_} {
-    emu_window.MakeCurrent();
-}
-ScopeAcquireWindowContext::~ScopeAcquireWindowContext() {
-    emu_window.DoneCurrent();
-}
-
-} // namespace Core::Frontend

src/core/gdbstub/gdbstub.cpp

@@ -217,7 +217,7 @@ static u64 RegRead(std::size_t id, Kernel::Thread* thread = nullptr) {
         return 0;
     }
 
-    const auto& thread_context = thread->GetContext();
+    const auto& thread_context = thread->GetContext64();
 
     if (id < SP_REGISTER) {
         return thread_context.cpu_registers[id];
@@ -239,7 +239,7 @@ static void RegWrite(std::size_t id, u64 val, Kernel::Thread* thread = nullptr)
         return;
     }
 
-    auto& thread_context = thread->GetContext();
+    auto& thread_context = thread->GetContext64();
 
     if (id < SP_REGISTER) {
         thread_context.cpu_registers[id] = val;
@@ -259,7 +259,7 @@ static u128 FpuRead(std::size_t id, Kernel::Thread* thread = nullptr) {
         return u128{0};
     }
 
-    auto& thread_context = thread->GetContext();
+    auto& thread_context = thread->GetContext64();
 
     if (id >= UC_ARM64_REG_Q0 && id < FPCR_REGISTER) {
         return thread_context.vector_registers[id - UC_ARM64_REG_Q0];
@@ -275,7 +275,7 @@ static void FpuWrite(std::size_t id, u128 val, Kernel::Thread* thread = nullptr)
         return;
     }
 
-    auto& thread_context = thread->GetContext();
+    auto& thread_context = thread->GetContext64();
 
     if (id >= UC_ARM64_REG_Q0 && id < FPCR_REGISTER) {
         thread_context.vector_registers[id - UC_ARM64_REG_Q0] = val;
@@ -916,7 +916,7 @@ static void WriteRegister() {
     // Update ARM context, skipping scheduler - no running threads at this point
     Core::System::GetInstance()
         .ArmInterface(current_core)
-        .LoadContext(current_thread->GetContext());
+        .LoadContext(current_thread->GetContext64());
 
     SendReply("OK");
 }
@@ -947,7 +947,7 @@ static void WriteRegisters() {
     // Update ARM context, skipping scheduler - no running threads at this point
     Core::System::GetInstance()
         .ArmInterface(current_core)
-        .LoadContext(current_thread->GetContext());
+        .LoadContext(current_thread->GetContext64());
 
     SendReply("OK");
 }
@@ -1019,7 +1019,7 @@ static void Step() {
         // Update ARM context, skipping scheduler - no running threads at this point
         Core::System::GetInstance()
             .ArmInterface(current_core)
-            .LoadContext(current_thread->GetContext());
+            .LoadContext(current_thread->GetContext64());
     }
     step_loop = true;
     halt_loop = true;

src/core/hle/kernel/kernel.cpp

@@ -186,6 +186,10 @@ struct KernelCore::Impl {
             return;
         }
 
+        for (auto& core : cores) {
+            core.SetIs64Bit(process->Is64BitProcess());
+        }
+
         system.Memory().SetCurrentPageTable(*process);
     }
 

src/core/hle/kernel/physical_core.cpp

@@ -5,7 +5,8 @@
 #include "common/logging/log.h"
 #include "core/arm/arm_interface.h"
 #ifdef ARCHITECTURE_x86_64
-#include "core/arm/dynarmic/arm_dynarmic.h"
+#include "core/arm/dynarmic/arm_dynarmic_32.h"
+#include "core/arm/dynarmic/arm_dynarmic_64.h"
 #endif
 #include "core/arm/exclusive_monitor.h"
 #include "core/arm/unicorn/arm_unicorn.h"
@@ -20,13 +21,17 @@ PhysicalCore::PhysicalCore(Core::System& system, std::size_t id,
                            Core::ExclusiveMonitor& exclusive_monitor)
     : core_index{id} {
 #ifdef ARCHITECTURE_x86_64
-    arm_interface = std::make_unique<Core::ARM_Dynarmic>(system, exclusive_monitor, core_index);
+    arm_interface_32 =
+        std::make_unique<Core::ARM_Dynarmic_32>(system, exclusive_monitor, core_index);
+    arm_interface_64 =
+        std::make_unique<Core::ARM_Dynarmic_64>(system, exclusive_monitor, core_index);
+
 #else
     arm_interface = std::make_shared<Core::ARM_Unicorn>(system);
     LOG_WARNING(Core, "CPU JIT requested, but Dynarmic not available");
 #endif
 
-    scheduler = std::make_unique<Kernel::Scheduler>(system, *arm_interface, core_index);
+    scheduler = std::make_unique<Kernel::Scheduler>(system, core_index);
 }
 
 PhysicalCore::~PhysicalCore() = default;
@@ -48,4 +53,12 @@ void PhysicalCore::Shutdown() {
     scheduler->Shutdown();
 }
 
+void PhysicalCore::SetIs64Bit(bool is_64_bit) {
+    if (is_64_bit) {
+        arm_interface = arm_interface_64.get();
+    } else {
+        arm_interface = arm_interface_32.get();
+    }
+}
+
 } // namespace Kernel

src/core/hle/kernel/physical_core.h

@@ -68,10 +68,14 @@ public:
         return *scheduler;
     }
 
+    void SetIs64Bit(bool is_64_bit);
+
 private:
     std::size_t core_index;
-    std::unique_ptr<Core::ARM_Interface> arm_interface;
+    std::unique_ptr<Core::ARM_Interface> arm_interface_32;
+    std::unique_ptr<Core::ARM_Interface> arm_interface_64;
     std::unique_ptr<Kernel::Scheduler> scheduler;
+    Core::ARM_Interface* arm_interface{};
 };
 
 } // namespace Kernel

src/core/hle/kernel/process.cpp

@@ -42,7 +42,8 @@ void SetupMainThread(Process& owner_process, KernelCore& kernel, u32 priority) {
 
     // Register 1 must be a handle to the main thread
     const Handle thread_handle = owner_process.GetHandleTable().Create(thread).Unwrap();
-    thread->GetContext().cpu_registers[1] = thread_handle;
+    thread->GetContext32().cpu_registers[1] = thread_handle;
+    thread->GetContext64().cpu_registers[1] = thread_handle;
 
     // Threads by default are dormant, wake up the main thread so it runs when the scheduler fires
     thread->ResumeFromWait();

src/core/hle/kernel/scheduler.cpp

@@ -383,8 +383,8 @@ void GlobalScheduler::Unlock() {
     // TODO(Blinkhawk): Setup the interrupts and change context on current core.
 }
 
-Scheduler::Scheduler(Core::System& system, Core::ARM_Interface& cpu_core, std::size_t core_id)
-    : system(system), cpu_core(cpu_core), core_id(core_id) {}
+Scheduler::Scheduler(Core::System& system, std::size_t core_id)
+    : system{system}, core_id{core_id} {}
 
 Scheduler::~Scheduler() = default;
 
@@ -422,9 +422,10 @@ void Scheduler::UnloadThread() {
 
     // Save context for previous thread
     if (previous_thread) {
-        cpu_core.SaveContext(previous_thread->GetContext());
+        system.ArmInterface(core_id).SaveContext(previous_thread->GetContext32());
+        system.ArmInterface(core_id).SaveContext(previous_thread->GetContext64());
         // Save the TPIDR_EL0 system register in case it was modified.
-        previous_thread->SetTPIDR_EL0(cpu_core.GetTPIDR_EL0());
+        previous_thread->SetTPIDR_EL0(system.ArmInterface(core_id).GetTPIDR_EL0());
 
         if (previous_thread->GetStatus() == ThreadStatus::Running) {
             // This is only the case when a reschedule is triggered without the current thread
@@ -451,9 +452,10 @@ void Scheduler::SwitchContext() {
 
     // Save context for previous thread
     if (previous_thread) {
-        cpu_core.SaveContext(previous_thread->GetContext());
+        system.ArmInterface(core_id).SaveContext(previous_thread->GetContext32());
+        system.ArmInterface(core_id).SaveContext(previous_thread->GetContext64());
         // Save the TPIDR_EL0 system register in case it was modified.
-        previous_thread->SetTPIDR_EL0(cpu_core.GetTPIDR_EL0());
+        previous_thread->SetTPIDR_EL0(system.ArmInterface(core_id).GetTPIDR_EL0());
 
         if (previous_thread->GetStatus() == ThreadStatus::Running) {
             // This is only the case when a reschedule is triggered without the current thread
@@ -481,9 +483,10 @@ void Scheduler::SwitchContext() {
             system.Kernel().MakeCurrentProcess(thread_owner_process);
         }
 
-        cpu_core.LoadContext(new_thread->GetContext());
-        cpu_core.SetTlsAddress(new_thread->GetTLSAddress());
-        cpu_core.SetTPIDR_EL0(new_thread->GetTPIDR_EL0());
+        system.ArmInterface(core_id).LoadContext(new_thread->GetContext32());
+        system.ArmInterface(core_id).LoadContext(new_thread->GetContext64());
+        system.ArmInterface(core_id).SetTlsAddress(new_thread->GetTLSAddress());
+        system.ArmInterface(core_id).SetTPIDR_EL0(new_thread->GetTPIDR_EL0());
     } else {
         current_thread = nullptr;
         // Note: We do not reset the current process and current page table when idling because

src/core/hle/kernel/scheduler.h

@@ -181,7 +181,7 @@ private:
 
 class Scheduler final {
 public:
-    explicit Scheduler(Core::System& system, Core::ARM_Interface& cpu_core, std::size_t core_id);
+    explicit Scheduler(Core::System& system, std::size_t core_id);
     ~Scheduler();
 
     /// Returns whether there are any threads that are ready to run.
@@ -235,7 +235,6 @@ private:
     std::shared_ptr<Thread> selected_thread = nullptr;
 
     Core::System& system;
-    Core::ARM_Interface& cpu_core;
     u64 last_context_switch_time = 0;
     u64 idle_selection_count = 0;
     const std::size_t core_id;

src/core/hle/kernel/svc.cpp

@@ -187,6 +187,13 @@ static ResultCode SetHeapSize(Core::System& system, VAddr* heap_addr, u64 heap_s
     return RESULT_SUCCESS;
 }
 
+static ResultCode SetHeapSize32(Core::System& system, u32* heap_addr, u32 heap_size) {
+    VAddr temp_heap_addr{};
+    const ResultCode result{SetHeapSize(system, &temp_heap_addr, heap_size)};
+    *heap_addr = static_cast<u32>(temp_heap_addr);
+    return result;
+}
+
 static ResultCode SetMemoryPermission(Core::System& system, VAddr addr, u64 size, u32 prot) {
     LOG_TRACE(Kernel_SVC, "called, addr=0x{:X}, size=0x{:X}, prot=0x{:X}", addr, size, prot);
 
@@ -371,6 +378,12 @@ static ResultCode ConnectToNamedPort(Core::System& system, Handle* out_handle,
     return RESULT_SUCCESS;
 }
 
+static ResultCode ConnectToNamedPort32(Core::System& system, Handle* out_handle,
+                                       u32 port_name_address) {
+
+    return ConnectToNamedPort(system, out_handle, port_name_address);
+}
+
 /// Makes a blocking IPC call to an OS service.
 static ResultCode SendSyncRequest(Core::System& system, Handle handle) {
     const auto& handle_table = system.Kernel().CurrentProcess()->GetHandleTable();
@@ -390,6 +403,10 @@ static ResultCode SendSyncRequest(Core::System& system, Handle handle) {
     return session->SendSyncRequest(SharedFrom(thread), system.Memory());
 }
 
+static ResultCode SendSyncRequest32(Core::System& system, Handle handle) {
+    return SendSyncRequest(system, handle);
+}
+
 /// Get the ID for the specified thread.
 static ResultCode GetThreadId(Core::System& system, u64* thread_id, Handle thread_handle) {
     LOG_TRACE(Kernel_SVC, "called thread=0x{:08X}", thread_handle);
@@ -405,6 +422,17 @@ static ResultCode GetThreadId(Core::System& system, u64* thread_id, Handle threa
     return RESULT_SUCCESS;
 }
 
+static ResultCode GetThreadId32(Core::System& system, u32* thread_id_low, u32* thread_id_high,
+                                Handle thread_handle) {
+    u64 thread_id{};
+    const ResultCode result{GetThreadId(system, &thread_id, thread_handle)};
+
+    *thread_id_low = static_cast<u32>(thread_id >> 32);
+    *thread_id_high = static_cast<u32>(thread_id & std::numeric_limits<u32>::max());
+
+    return result;
+}
+
 /// Gets the ID of the specified process or a specified thread's owning process.
 static ResultCode GetProcessId(Core::System& system, u64* process_id, Handle handle) {
     LOG_DEBUG(Kernel_SVC, "called handle=0x{:08X}", handle);
@@ -479,6 +507,12 @@ static ResultCode WaitSynchronization(Core::System& system, Handle* index, VAddr
     return result;
 }
 
+static ResultCode WaitSynchronization32(Core::System& system, u32 timeout_low, u32 handles_address,
+                                        s32 handle_count, u32 timeout_high, Handle* index) {
+    const s64 nano_seconds{(static_cast<s64>(timeout_high) << 32) | static_cast<s64>(timeout_low)};
+    return WaitSynchronization(system, index, handles_address, handle_count, nano_seconds);
+}
+
 /// Resumes a thread waiting on WaitSynchronization
 static ResultCode CancelSynchronization(Core::System& system, Handle thread_handle) {
     LOG_TRACE(Kernel_SVC, "called thread=0x{:X}", thread_handle);
@@ -917,6 +951,18 @@ static ResultCode GetInfo(Core::System& system, u64* result, u64 info_id, u64 ha
     }
 }
 
+static ResultCode GetInfo32(Core::System& system, u32* result_low, u32* result_high, u32 sub_id_low,
+                            u32 info_id, u32 handle, u32 sub_id_high) {
+    const u64 sub_id{static_cast<u64>(sub_id_low | (static_cast<u64>(sub_id_high) << 32))};
+    u64 res_value{};
+
+    const ResultCode result{GetInfo(system, &res_value, info_id, handle, sub_id)};
+    *result_high = static_cast<u32>(res_value >> 32);
+    *result_low = static_cast<u32>(res_value & std::numeric_limits<u32>::max());
+
+    return result;
+}
+
 /// Maps memory at a desired address
 static ResultCode MapPhysicalMemory(Core::System& system, VAddr addr, u64 size) {
     LOG_DEBUG(Kernel_SVC, "called, addr=0x{:016X}, size=0x{:X}", addr, size);
@@ -1058,7 +1104,7 @@ static ResultCode GetThreadContext(Core::System& system, VAddr thread_context, H
         return ERR_BUSY;
     }
 
-    Core::ARM_Interface::ThreadContext ctx = thread->GetContext();
+    Core::ARM_Interface::ThreadContext64 ctx = thread->GetContext64();
     // Mask away mode bits, interrupt bits, IL bit, and other reserved bits.
     ctx.pstate &= 0xFF0FFE20;
 
@@ -1088,6 +1134,10 @@ static ResultCode GetThreadPriority(Core::System& system, u32* priority, Handle
     return RESULT_SUCCESS;
 }
 
+static ResultCode GetThreadPriority32(Core::System& system, u32* priority, Handle handle) {
+    return GetThreadPriority(system, priority, handle);
+}
+
 /// Sets the priority for the specified thread
 static ResultCode SetThreadPriority(Core::System& system, Handle handle, u32 priority) {
     LOG_TRACE(Kernel_SVC, "called");
@@ -1259,6 +1309,11 @@ static ResultCode QueryMemory(Core::System& system, VAddr memory_info_address,
                               query_address);
 }
 
+static ResultCode QueryMemory32(Core::System& system, u32 memory_info_address,
+                                u32 page_info_address, u32 query_address) {
+    return QueryMemory(system, memory_info_address, page_info_address, query_address);
+}
+
 static ResultCode MapProcessCodeMemory(Core::System& system, Handle process_handle, u64 dst_address,
                                        u64 src_address, u64 size) {
     LOG_DEBUG(Kernel_SVC,
@@ -1675,6 +1730,10 @@ static void SignalProcessWideKey(Core::System& system, VAddr condition_variable_
     }
 }
 
+static void SignalProcessWideKey32(Core::System& system, u32 condition_variable_addr, s32 target) {
+    SignalProcessWideKey(system, condition_variable_addr, target);
+}
+
 // Wait for an address (via Address Arbiter)
 static ResultCode WaitForAddress(Core::System& system, VAddr address, u32 type, s32 value,
                                  s64 timeout) {
@@ -1760,6 +1819,10 @@ static ResultCode CloseHandle(Core::System& system, Handle handle) {
     return handle_table.Close(handle);
 }
 
+static ResultCode CloseHandle32(Core::System& system, Handle handle) {
+    return CloseHandle(system, handle);
+}
+
 /// Clears the signaled state of an event or process.
 static ResultCode ResetSignal(Core::System& system, Handle handle) {
     LOG_DEBUG(Kernel_SVC, "called handle 0x{:08X}", handle);
@@ -2317,69 +2380,196 @@ struct FunctionDef {
 };
 } // namespace
 
-static const FunctionDef SVC_Table[] = {
+static const FunctionDef SVC_Table_32[] = {
     {0x00, nullptr, "Unknown"},
-    {0x01, SvcWrap<SetHeapSize>, "SetHeapSize"},
-    {0x02, SvcWrap<SetMemoryPermission>, "SetMemoryPermission"},
-    {0x03, SvcWrap<SetMemoryAttribute>, "SetMemoryAttribute"},
-    {0x04, SvcWrap<MapMemory>, "MapMemory"},
-    {0x05, SvcWrap<UnmapMemory>, "UnmapMemory"},
-    {0x06, SvcWrap<QueryMemory>, "QueryMemory"},
-    {0x07, SvcWrap<ExitProcess>, "ExitProcess"},
-    {0x08, SvcWrap<CreateThread>, "CreateThread"},
-    {0x09, SvcWrap<StartThread>, "StartThread"},
-    {0x0A, SvcWrap<ExitThread>, "ExitThread"},
-    {0x0B, SvcWrap<SleepThread>, "SleepThread"},
-    {0x0C, SvcWrap<GetThreadPriority>, "GetThreadPriority"},
-    {0x0D, SvcWrap<SetThreadPriority>, "SetThreadPriority"},
-    {0x0E, SvcWrap<GetThreadCoreMask>, "GetThreadCoreMask"},
-    {0x0F, SvcWrap<SetThreadCoreMask>, "SetThreadCoreMask"},
-    {0x10, SvcWrap<GetCurrentProcessorNumber>, "GetCurrentProcessorNumber"},
-    {0x11, SvcWrap<SignalEvent>, "SignalEvent"},
-    {0x12, SvcWrap<ClearEvent>, "ClearEvent"},
-    {0x13, SvcWrap<MapSharedMemory>, "MapSharedMemory"},
-    {0x14, SvcWrap<UnmapSharedMemory>, "UnmapSharedMemory"},
-    {0x15, SvcWrap<CreateTransferMemory>, "CreateTransferMemory"},
-    {0x16, SvcWrap<CloseHandle>, "CloseHandle"},
-    {0x17, SvcWrap<ResetSignal>, "ResetSignal"},
-    {0x18, SvcWrap<WaitSynchronization>, "WaitSynchronization"},
-    {0x19, SvcWrap<CancelSynchronization>, "CancelSynchronization"},
-    {0x1A, SvcWrap<ArbitrateLock>, "ArbitrateLock"},
-    {0x1B, SvcWrap<ArbitrateUnlock>, "ArbitrateUnlock"},
-    {0x1C, SvcWrap<WaitProcessWideKeyAtomic>, "WaitProcessWideKeyAtomic"},
-    {0x1D, SvcWrap<SignalProcessWideKey>, "SignalProcessWideKey"},
-    {0x1E, SvcWrap<GetSystemTick>, "GetSystemTick"},
-    {0x1F, SvcWrap<ConnectToNamedPort>, "ConnectToNamedPort"},
+    {0x01, SvcWrap32<SetHeapSize32>, "SetHeapSize32"},
+    {0x02, nullptr, "Unknown"},
+    {0x03, nullptr, "SetMemoryAttribute32"},
+    {0x04, nullptr, "MapMemory32"},
+    {0x05, nullptr, "UnmapMemory32"},
+    {0x06, SvcWrap32<QueryMemory32>, "QueryMemory32"},
+    {0x07, nullptr, "ExitProcess32"},
+    {0x08, nullptr, "CreateThread32"},
+    {0x09, nullptr, "StartThread32"},
+    {0x0a, nullptr, "ExitThread32"},
+    {0x0b, nullptr, "SleepThread32"},
+    {0x0c, SvcWrap32<GetThreadPriority32>, "GetThreadPriority32"},
+    {0x0d, nullptr, "SetThreadPriority32"},
+    {0x0e, nullptr, "GetThreadCoreMask32"},
+    {0x0f, nullptr, "SetThreadCoreMask32"},
+    {0x10, nullptr, "GetCurrentProcessorNumber32"},
+    {0x11, nullptr, "SignalEvent32"},
+    {0x12, nullptr, "ClearEvent32"},
+    {0x13, nullptr, "MapSharedMemory32"},
+    {0x14, nullptr, "UnmapSharedMemory32"},
+    {0x15, nullptr, "CreateTransferMemory32"},
+    {0x16, SvcWrap32<CloseHandle32>, "CloseHandle32"},
+    {0x17, nullptr, "ResetSignal32"},
+    {0x18, SvcWrap32<WaitSynchronization32>, "WaitSynchronization32"},
+    {0x19, nullptr, "CancelSynchronization32"},
+    {0x1a, nullptr, "ArbitrateLock32"},
+    {0x1b, nullptr, "ArbitrateUnlock32"},
+    {0x1c, nullptr, "WaitProcessWideKeyAtomic32"},
+    {0x1d, SvcWrap32<SignalProcessWideKey32>, "SignalProcessWideKey32"},
+    {0x1e, nullptr, "GetSystemTick32"},
+    {0x1f, SvcWrap32<ConnectToNamedPort32>, "ConnectToNamedPort32"},
+    {0x20, nullptr, "Unknown"},
+    {0x21, SvcWrap32<SendSyncRequest32>, "SendSyncRequest32"},
+    {0x22, nullptr, "SendSyncRequestWithUserBuffer32"},
+    {0x23, nullptr, "Unknown"},
+    {0x24, nullptr, "GetProcessId32"},
+    {0x25, SvcWrap32<GetThreadId32>, "GetThreadId32"},
+    {0x26, nullptr, "Break32"},
+    {0x27, nullptr, "OutputDebugString32"},
+    {0x28, nullptr, "Unknown"},
+    {0x29, SvcWrap32<GetInfo32>, "GetInfo32"},
+    {0x2a, nullptr, "Unknown"},
+    {0x2b, nullptr, "Unknown"},
+    {0x2c, nullptr, "MapPhysicalMemory32"},
+    {0x2d, nullptr, "UnmapPhysicalMemory32"},
+    {0x2e, nullptr, "Unknown"},
+    {0x2f, nullptr, "Unknown"},
+    {0x30, nullptr, "Unknown"},
+    {0x31, nullptr, "Unknown"},
+    {0x32, nullptr, "SetThreadActivity32"},
+    {0x33, nullptr, "GetThreadContext32"},
+    {0x34, nullptr, "WaitForAddress32"},
+    {0x35, nullptr, "SignalToAddress32"},
+    {0x36, nullptr, "Unknown"},
+    {0x37, nullptr, "Unknown"},
+    {0x38, nullptr, "Unknown"},
+    {0x39, nullptr, "Unknown"},
+    {0x3a, nullptr, "Unknown"},
+    {0x3b, nullptr, "Unknown"},
+    {0x3c, nullptr, "Unknown"},
+    {0x3d, nullptr, "Unknown"},
+    {0x3e, nullptr, "Unknown"},
+    {0x3f, nullptr, "Unknown"},
+    {0x40, nullptr, "CreateSession32"},
+    {0x41, nullptr, "AcceptSession32"},
+    {0x42, nullptr, "Unknown"},
+    {0x43, nullptr, "ReplyAndReceive32"},
+    {0x44, nullptr, "Unknown"},
+    {0x45, nullptr, "CreateEvent32"},
+    {0x46, nullptr, "Unknown"},
+    {0x47, nullptr, "Unknown"},
+    {0x48, nullptr, "Unknown"},
+    {0x49, nullptr, "Unknown"},
+    {0x4a, nullptr, "Unknown"},
+    {0x4b, nullptr, "Unknown"},
+    {0x4c, nullptr, "Unknown"},
+    {0x4d, nullptr, "Unknown"},
+    {0x4e, nullptr, "Unknown"},
+    {0x4f, nullptr, "Unknown"},
+    {0x50, nullptr, "Unknown"},
+    {0x51, nullptr, "Unknown"},
+    {0x52, nullptr, "Unknown"},
+    {0x53, nullptr, "Unknown"},
+    {0x54, nullptr, "Unknown"},
+    {0x55, nullptr, "Unknown"},
+    {0x56, nullptr, "Unknown"},
+    {0x57, nullptr, "Unknown"},
+    {0x58, nullptr, "Unknown"},
+    {0x59, nullptr, "Unknown"},
+    {0x5a, nullptr, "Unknown"},
+    {0x5b, nullptr, "Unknown"},
+    {0x5c, nullptr, "Unknown"},
+    {0x5d, nullptr, "Unknown"},
+    {0x5e, nullptr, "Unknown"},
+    {0x5F, nullptr, "FlushProcessDataCache32"},
+    {0x60, nullptr, "Unknown"},
+    {0x61, nullptr, "Unknown"},
+    {0x62, nullptr, "Unknown"},
+    {0x63, nullptr, "Unknown"},
+    {0x64, nullptr, "Unknown"},
+    {0x65, nullptr, "GetProcessList32"},
+    {0x66, nullptr, "Unknown"},
+    {0x67, nullptr, "Unknown"},
+    {0x68, nullptr, "Unknown"},
+    {0x69, nullptr, "Unknown"},
+    {0x6A, nullptr, "Unknown"},
+    {0x6B, nullptr, "Unknown"},
+    {0x6C, nullptr, "Unknown"},
+    {0x6D, nullptr, "Unknown"},
+    {0x6E, nullptr, "Unknown"},
+    {0x6f, nullptr, "GetSystemInfo32"},
+    {0x70, nullptr, "CreatePort32"},
+    {0x71, nullptr, "ManageNamedPort32"},
+    {0x72, nullptr, "ConnectToPort32"},
+    {0x73, nullptr, "SetProcessMemoryPermission32"},
+    {0x74, nullptr, "Unknown"},
+    {0x75, nullptr, "Unknown"},
+    {0x76, nullptr, "Unknown"},
+    {0x77, nullptr, "MapProcessCodeMemory32"},
+    {0x78, nullptr, "UnmapProcessCodeMemory32"},
+    {0x79, nullptr, "Unknown"},
+    {0x7A, nullptr, "Unknown"},
+    {0x7B, nullptr, "TerminateProcess32"},
+};
+
+static const FunctionDef SVC_Table_64[] = {
+    {0x00, nullptr, "Unknown"},
+    {0x01, SvcWrap64<SetHeapSize>, "SetHeapSize"},
+    {0x02, SvcWrap64<SetMemoryPermission>, "SetMemoryPermission"},
+    {0x03, SvcWrap64<SetMemoryAttribute>, "SetMemoryAttribute"},
+    {0x04, SvcWrap64<MapMemory>, "MapMemory"},
+    {0x05, SvcWrap64<UnmapMemory>, "UnmapMemory"},
+    {0x06, SvcWrap64<QueryMemory>, "QueryMemory"},
+    {0x07, SvcWrap64<ExitProcess>, "ExitProcess"},
+    {0x08, SvcWrap64<CreateThread>, "CreateThread"},
+    {0x09, SvcWrap64<StartThread>, "StartThread"},
+    {0x0A, SvcWrap64<ExitThread>, "ExitThread"},
+    {0x0B, SvcWrap64<SleepThread>, "SleepThread"},
+    {0x0C, SvcWrap64<GetThreadPriority>, "GetThreadPriority"},
+    {0x0D, SvcWrap64<SetThreadPriority>, "SetThreadPriority"},
+    {0x0E, SvcWrap64<GetThreadCoreMask>, "GetThreadCoreMask"},
+    {0x0F, SvcWrap64<SetThreadCoreMask>, "SetThreadCoreMask"},
+    {0x10, SvcWrap64<GetCurrentProcessorNumber>, "GetCurrentProcessorNumber"},
+    {0x11, SvcWrap64<SignalEvent>, "SignalEvent"},
+    {0x12, SvcWrap64<ClearEvent>, "ClearEvent"},
+    {0x13, SvcWrap64<MapSharedMemory>, "MapSharedMemory"},
+    {0x14, SvcWrap64<UnmapSharedMemory>, "UnmapSharedMemory"},
+    {0x15, SvcWrap64<CreateTransferMemory>, "CreateTransferMemory"},
+    {0x16, SvcWrap64<CloseHandle>, "CloseHandle"},
+    {0x17, SvcWrap64<ResetSignal>, "ResetSignal"},
+    {0x18, SvcWrap64<WaitSynchronization>, "WaitSynchronization"},
+    {0x19, SvcWrap64<CancelSynchronization>, "CancelSynchronization"},
+    {0x1A, SvcWrap64<ArbitrateLock>, "ArbitrateLock"},
+    {0x1B, SvcWrap64<ArbitrateUnlock>, "ArbitrateUnlock"},
+    {0x1C, SvcWrap64<WaitProcessWideKeyAtomic>, "WaitProcessWideKeyAtomic"},
+    {0x1D, SvcWrap64<SignalProcessWideKey>, "SignalProcessWideKey"},
+    {0x1E, SvcWrap64<GetSystemTick>, "GetSystemTick"},
+    {0x1F, SvcWrap64<ConnectToNamedPort>, "ConnectToNamedPort"},
     {0x20, nullptr, "SendSyncRequestLight"},
-    {0x21, SvcWrap<SendSyncRequest>, "SendSyncRequest"},
+    {0x21, SvcWrap64<SendSyncRequest>, "SendSyncRequest"},
     {0x22, nullptr, "SendSyncRequestWithUserBuffer"},
     {0x23, nullptr, "SendAsyncRequestWithUserBuffer"},
-    {0x24, SvcWrap<GetProcessId>, "GetProcessId"},
-    {0x25, SvcWrap<GetThreadId>, "GetThreadId"},
-    {0x26, SvcWrap<Break>, "Break"},
-    {0x27, SvcWrap<OutputDebugString>, "OutputDebugString"},
+    {0x24, SvcWrap64<GetProcessId>, "GetProcessId"},
+    {0x25, SvcWrap64<GetThreadId>, "GetThreadId"},
+    {0x26, SvcWrap64<Break>, "Break"},
+    {0x27, SvcWrap64<OutputDebugString>, "OutputDebugString"},
     {0x28, nullptr, "ReturnFromException"},
-    {0x29, SvcWrap<GetInfo>, "GetInfo"},
+    {0x29, SvcWrap64<GetInfo>, "GetInfo"},
     {0x2A, nullptr, "FlushEntireDataCache"},
     {0x2B, nullptr, "FlushDataCache"},
-    {0x2C, SvcWrap<MapPhysicalMemory>, "MapPhysicalMemory"},
-    {0x2D, SvcWrap<UnmapPhysicalMemory>, "UnmapPhysicalMemory"},
+    {0x2C, SvcWrap64<MapPhysicalMemory>, "MapPhysicalMemory"},
+    {0x2D, SvcWrap64<UnmapPhysicalMemory>, "UnmapPhysicalMemory"},
     {0x2E, nullptr, "GetFutureThreadInfo"},
     {0x2F, nullptr, "GetLastThreadInfo"},
-    {0x30, SvcWrap<GetResourceLimitLimitValue>, "GetResourceLimitLimitValue"},
-    {0x31, SvcWrap<GetResourceLimitCurrentValue>, "GetResourceLimitCurrentValue"},
-    {0x32, SvcWrap<SetThreadActivity>, "SetThreadActivity"},
-    {0x33, SvcWrap<GetThreadContext>, "GetThreadContext"},
-    {0x34, SvcWrap<WaitForAddress>, "WaitForAddress"},
-    {0x35, SvcWrap<SignalToAddress>, "SignalToAddress"},
+    {0x30, SvcWrap64<GetResourceLimitLimitValue>, "GetResourceLimitLimitValue"},
+    {0x31, SvcWrap64<GetResourceLimitCurrentValue>, "GetResourceLimitCurrentValue"},
+    {0x32, SvcWrap64<SetThreadActivity>, "SetThreadActivity"},
+    {0x33, SvcWrap64<GetThreadContext>, "GetThreadContext"},
+    {0x34, SvcWrap64<WaitForAddress>, "WaitForAddress"},
+    {0x35, SvcWrap64<SignalToAddress>, "SignalToAddress"},
     {0x36, nullptr, "SynchronizePreemptionState"},
     {0x37, nullptr, "Unknown"},
     {0x38, nullptr, "Unknown"},
     {0x39, nullptr, "Unknown"},
     {0x3A, nullptr, "Unknown"},
     {0x3B, nullptr, "Unknown"},
-    {0x3C, SvcWrap<KernelDebug>, "KernelDebug"},
-    {0x3D, SvcWrap<ChangeKernelTraceState>, "ChangeKernelTraceState"},
+    {0x3C, SvcWrap64<KernelDebug>, "KernelDebug"},
+    {0x3D, SvcWrap64<ChangeKernelTraceState>, "ChangeKernelTraceState"},
     {0x3E, nullptr, "Unknown"},
     {0x3F, nullptr, "Unknown"},
     {0x40, nullptr, "CreateSession"},
@@ -2387,7 +2577,7 @@ static const FunctionDef SVC_Table[] = {
     {0x42, nullptr, "ReplyAndReceiveLight"},
     {0x43, nullptr, "ReplyAndReceive"},
     {0x44, nullptr, "ReplyAndReceiveWithUserBuffer"},
-    {0x45, SvcWrap<CreateEvent>, "CreateEvent"},
+    {0x45, SvcWrap64<CreateEvent>, "CreateEvent"},
     {0x46, nullptr, "Unknown"},
     {0x47, nullptr, "Unknown"},
     {0x48, nullptr, "MapPhysicalMemoryUnsafe"},
@@ -2398,9 +2588,9 @@ static const FunctionDef SVC_Table[] = {
     {0x4D, nullptr, "SleepSystem"},
     {0x4E, nullptr, "ReadWriteRegister"},
     {0x4F, nullptr, "SetProcessActivity"},
-    {0x50, SvcWrap<CreateSharedMemory>, "CreateSharedMemory"},
-    {0x51, SvcWrap<MapTransferMemory>, "MapTransferMemory"},
-    {0x52, SvcWrap<UnmapTransferMemory>, "UnmapTransferMemory"},
+    {0x50, SvcWrap64<CreateSharedMemory>, "CreateSharedMemory"},
+    {0x51, SvcWrap64<MapTransferMemory>, "MapTransferMemory"},
+    {0x52, SvcWrap64<UnmapTransferMemory>, "UnmapTransferMemory"},
     {0x53, nullptr, "CreateInterruptEvent"},
     {0x54, nullptr, "QueryPhysicalAddress"},
     {0x55, nullptr, "QueryIoMapping"},
@@ -2419,8 +2609,8 @@ static const FunctionDef SVC_Table[] = {
     {0x62, nullptr, "TerminateDebugProcess"},
     {0x63, nullptr, "GetDebugEvent"},
     {0x64, nullptr, "ContinueDebugEvent"},
-    {0x65, SvcWrap<GetProcessList>, "GetProcessList"},
-    {0x66, SvcWrap<GetThreadList>, "GetThreadList"},
+    {0x65, SvcWrap64<GetProcessList>, "GetProcessList"},
+    {0x66, SvcWrap64<GetThreadList>, "GetThreadList"},
     {0x67, nullptr, "GetDebugThreadContext"},
     {0x68, nullptr, "SetDebugThreadContext"},
     {0x69, nullptr, "QueryDebugProcessMemory"},
@@ -2436,24 +2626,32 @@ static const FunctionDef SVC_Table[] = {
     {0x73, nullptr, "SetProcessMemoryPermission"},
     {0x74, nullptr, "MapProcessMemory"},
     {0x75, nullptr, "UnmapProcessMemory"},
-    {0x76, SvcWrap<QueryProcessMemory>, "QueryProcessMemory"},
-    {0x77, SvcWrap<MapProcessCodeMemory>, "MapProcessCodeMemory"},
-    {0x78, SvcWrap<UnmapProcessCodeMemory>, "UnmapProcessCodeMemory"},
+    {0x76, SvcWrap64<QueryProcessMemory>, "QueryProcessMemory"},
+    {0x77, SvcWrap64<MapProcessCodeMemory>, "MapProcessCodeMemory"},
+    {0x78, SvcWrap64<UnmapProcessCodeMemory>, "UnmapProcessCodeMemory"},
     {0x79, nullptr, "CreateProcess"},
     {0x7A, nullptr, "StartProcess"},
     {0x7B, nullptr, "TerminateProcess"},
-    {0x7C, SvcWrap<GetProcessInfo>, "GetProcessInfo"},
-    {0x7D, SvcWrap<CreateResourceLimit>, "CreateResourceLimit"},
-    {0x7E, SvcWrap<SetResourceLimitLimitValue>, "SetResourceLimitLimitValue"},
+    {0x7C, SvcWrap64<GetProcessInfo>, "GetProcessInfo"},
+    {0x7D, SvcWrap64<CreateResourceLimit>, "CreateResourceLimit"},
+    {0x7E, SvcWrap64<SetResourceLimitLimitValue>, "SetResourceLimitLimitValue"},
     {0x7F, nullptr, "CallSecureMonitor"},
 };
 
-static const FunctionDef* GetSVCInfo(u32 func_num) {
-    if (func_num >= std::size(SVC_Table)) {
+static const FunctionDef* GetSVCInfo32(u32 func_num) {
+    if (func_num >= std::size(SVC_Table_32)) {
         LOG_ERROR(Kernel_SVC, "Unknown svc=0x{:02X}", func_num);
         return nullptr;
     }
-    return &SVC_Table[func_num];
+    return &SVC_Table_32[func_num];
+}
+
+static const FunctionDef* GetSVCInfo64(u32 func_num) {
+    if (func_num >= std::size(SVC_Table_64)) {
+        LOG_ERROR(Kernel_SVC, "Unknown svc=0x{:02X}", func_num);
+        return nullptr;
+    }
+    return &SVC_Table_64[func_num];
 }
 
 MICROPROFILE_DEFINE(Kernel_SVC, "Kernel", "SVC", MP_RGB(70, 200, 70));
@@ -2464,7 +2662,8 @@ void CallSVC(Core::System& system, u32 immediate) {
     // Lock the global kernel mutex when we enter the kernel HLE.
     std::lock_guard lock{HLE::g_hle_lock};
 
-    const FunctionDef* info = GetSVCInfo(immediate);
+    const FunctionDef* info = system.CurrentProcess()->Is64BitProcess() ? GetSVCInfo64(immediate)
+                                                                        : GetSVCInfo32(immediate);
     if (info) {
         if (info->func) {
             info->func(system);

src/core/hle/kernel/svc_wrap.h

@@ -15,6 +15,10 @@ static inline u64 Param(const Core::System& system, int n) {
     return system.CurrentArmInterface().GetReg(n);
 }
 
+static inline u32 Param32(const Core::System& system, int n) {
+    return static_cast<u32>(system.CurrentArmInterface().GetReg(n));
+}
+
 /**
  * HLE a function return from the current ARM userland process
  * @param system System context
@@ -24,40 +28,44 @@ static inline void FuncReturn(Core::System& system, u64 result) {
     system.CurrentArmInterface().SetReg(0, result);
 }
 
+static inline void FuncReturn32(Core::System& system, u32 result) {
+    system.CurrentArmInterface().SetReg(0, (u64)result);
+}
+
 ////////////////////////////////////////////////////////////////////////////////////////////////////
 // Function wrappers that return type ResultCode
 
 template <ResultCode func(Core::System&, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     FuncReturn(system, func(system, Param(system, 0)).raw);
 }
 
 template <ResultCode func(Core::System&, u64, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     FuncReturn(system, func(system, Param(system, 0), Param(system, 1)).raw);
 }
 
 template <ResultCode func(Core::System&, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     FuncReturn(system, func(system, static_cast<u32>(Param(system, 0))).raw);
 }
 
 template <ResultCode func(Core::System&, u32, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     FuncReturn(
         system,
         func(system, static_cast<u32>(Param(system, 0)), static_cast<u32>(Param(system, 1))).raw);
 }
 
 template <ResultCode func(Core::System&, u32, u64, u64, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     FuncReturn(system, func(system, static_cast<u32>(Param(system, 0)), Param(system, 1),
                             Param(system, 2), Param(system, 3))
                            .raw);
 }
 
 template <ResultCode func(Core::System&, u32*)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     u32 param = 0;
     const u32 retval = func(system, &param).raw;
     system.CurrentArmInterface().SetReg(1, param);
@@ -65,7 +73,7 @@ void SvcWrap(Core::System& system) {
 }
 
 template <ResultCode func(Core::System&, u32*, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     u32 param_1 = 0;
     const u32 retval = func(system, &param_1, static_cast<u32>(Param(system, 1))).raw;
     system.CurrentArmInterface().SetReg(1, param_1);
@@ -73,7 +81,7 @@ void SvcWrap(Core::System& system) {
 }
 
 template <ResultCode func(Core::System&, u32*, u32*)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     u32 param_1 = 0;
     u32 param_2 = 0;
     const u32 retval = func(system, &param_1, &param_2).raw;
@@ -86,7 +94,7 @@ void SvcWrap(Core::System& system) {
 }
 
 template <ResultCode func(Core::System&, u32*, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     u32 param_1 = 0;
     const u32 retval = func(system, &param_1, Param(system, 1)).raw;
     system.CurrentArmInterface().SetReg(1, param_1);
@@ -94,7 +102,7 @@ void SvcWrap(Core::System& system) {
 }
 
 template <ResultCode func(Core::System&, u32*, u64, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     u32 param_1 = 0;
     const u32 retval =
         func(system, &param_1, Param(system, 1), static_cast<u32>(Param(system, 2))).raw;
@@ -104,7 +112,7 @@ void SvcWrap(Core::System& system) {
 }
 
 template <ResultCode func(Core::System&, u64*, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     u64 param_1 = 0;
     const u32 retval = func(system, &param_1, static_cast<u32>(Param(system, 1))).raw;
 
@@ -113,12 +121,12 @@ void SvcWrap(Core::System& system) {
 }
 
 template <ResultCode func(Core::System&, u64, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     FuncReturn(system, func(system, Param(system, 0), static_cast<u32>(Param(system, 1))).raw);
 }
 
 template <ResultCode func(Core::System&, u64*, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     u64 param_1 = 0;
     const u32 retval = func(system, &param_1, Param(system, 1)).raw;
 
@@ -127,7 +135,7 @@ void SvcWrap(Core::System& system) {
 }
 
 template <ResultCode func(Core::System&, u64*, u32, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     u64 param_1 = 0;
     const u32 retval = func(system, &param_1, static_cast<u32>(Param(system, 1)),
                             static_cast<u32>(Param(system, 2)))
@@ -138,19 +146,19 @@ void SvcWrap(Core::System& system) {
 }
 
 template <ResultCode func(Core::System&, u32, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     FuncReturn(system, func(system, static_cast<u32>(Param(system, 0)), Param(system, 1)).raw);
 }
 
 template <ResultCode func(Core::System&, u32, u32, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     FuncReturn(system, func(system, static_cast<u32>(Param(system, 0)),
                             static_cast<u32>(Param(system, 1)), Param(system, 2))
                            .raw);
 }
 
 template <ResultCode func(Core::System&, u32, u32*, u64*)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     u32 param_1 = 0;
     u64 param_2 = 0;
     const ResultCode retval = func(system, static_cast<u32>(Param(system, 2)), &param_1, &param_2);
@@ -161,54 +169,54 @@ void SvcWrap(Core::System& system) {
 }
 
 template <ResultCode func(Core::System&, u64, u64, u32, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     FuncReturn(system, func(system, Param(system, 0), Param(system, 1),
                             static_cast<u32>(Param(system, 2)), static_cast<u32>(Param(system, 3)))
                            .raw);
 }
 
 template <ResultCode func(Core::System&, u64, u64, u32, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     FuncReturn(system, func(system, Param(system, 0), Param(system, 1),
                             static_cast<u32>(Param(system, 2)), Param(system, 3))
                            .raw);
 }
 
 template <ResultCode func(Core::System&, u32, u64, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     FuncReturn(system, func(system, static_cast<u32>(Param(system, 0)), Param(system, 1),
                             static_cast<u32>(Param(system, 2)))
                            .raw);
 }
 
 template <ResultCode func(Core::System&, u64, u64, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     FuncReturn(system, func(system, Param(system, 0), Param(system, 1), Param(system, 2)).raw);
 }
 
 template <ResultCode func(Core::System&, u64, u64, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     FuncReturn(
         system,
         func(system, Param(system, 0), Param(system, 1), static_cast<u32>(Param(system, 2))).raw);
 }
 
 template <ResultCode func(Core::System&, u32, u64, u64, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     FuncReturn(system, func(system, static_cast<u32>(Param(system, 0)), Param(system, 1),
                             Param(system, 2), static_cast<u32>(Param(system, 3)))
                            .raw);
 }
 
 template <ResultCode func(Core::System&, u32, u64, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     FuncReturn(
         system,
         func(system, static_cast<u32>(Param(system, 0)), Param(system, 1), Param(system, 2)).raw);
 }
 
 template <ResultCode func(Core::System&, u32*, u64, u64, s64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     u32 param_1 = 0;
     const u32 retval = func(system, &param_1, Param(system, 1), static_cast<u32>(Param(system, 2)),
                             static_cast<s64>(Param(system, 3)))
@@ -219,14 +227,14 @@ void SvcWrap(Core::System& system) {
 }
 
 template <ResultCode func(Core::System&, u64, u64, u32, s64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     FuncReturn(system, func(system, Param(system, 0), Param(system, 1),
                             static_cast<u32>(Param(system, 2)), static_cast<s64>(Param(system, 3)))
                            .raw);
 }
 
 template <ResultCode func(Core::System&, u64*, u64, u64, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     u64 param_1 = 0;
     const u32 retval =
         func(system, &param_1, Param(system, 1), Param(system, 2), Param(system, 3)).raw;
@@ -236,7 +244,7 @@ void SvcWrap(Core::System& system) {
 }
 
 template <ResultCode func(Core::System&, u32*, u64, u64, u64, u32, s32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     u32 param_1 = 0;
     const u32 retval = func(system, &param_1, Param(system, 1), Param(system, 2), Param(system, 3),
                             static_cast<u32>(Param(system, 4)), static_cast<s32>(Param(system, 5)))
@@ -247,7 +255,7 @@ void SvcWrap(Core::System& system) {
 }
 
 template <ResultCode func(Core::System&, u32*, u64, u64, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     u32 param_1 = 0;
     const u32 retval = func(system, &param_1, Param(system, 1), Param(system, 2),
                             static_cast<u32>(Param(system, 3)))
@@ -258,7 +266,7 @@ void SvcWrap(Core::System& system) {
 }
 
 template <ResultCode func(Core::System&, Handle*, u64, u32, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     u32 param_1 = 0;
     const u32 retval = func(system, &param_1, Param(system, 1), static_cast<u32>(Param(system, 2)),
                             static_cast<u32>(Param(system, 3)))
@@ -269,14 +277,14 @@ void SvcWrap(Core::System& system) {
 }
 
 template <ResultCode func(Core::System&, u64, u32, s32, s64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     FuncReturn(system, func(system, Param(system, 0), static_cast<u32>(Param(system, 1)),
                             static_cast<s32>(Param(system, 2)), static_cast<s64>(Param(system, 3)))
                            .raw);
 }
 
 template <ResultCode func(Core::System&, u64, u32, s32, s32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     FuncReturn(system, func(system, Param(system, 0), static_cast<u32>(Param(system, 1)),
                             static_cast<s32>(Param(system, 2)), static_cast<s32>(Param(system, 3)))
                            .raw);
@@ -286,7 +294,7 @@ void SvcWrap(Core::System& system) {
 // Function wrappers that return type u32
 
 template <u32 func(Core::System&)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     FuncReturn(system, func(system));
 }
 
@@ -294,7 +302,7 @@ void SvcWrap(Core::System& system) {
 // Function wrappers that return type u64
 
 template <u64 func(Core::System&)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     FuncReturn(system, func(system));
 }
 
@@ -302,44 +310,110 @@ void SvcWrap(Core::System& system) {
 /// Function wrappers that return type void
 
 template <void func(Core::System&)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     func(system);
 }
 
 template <void func(Core::System&, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     func(system, static_cast<u32>(Param(system, 0)));
 }
 
 template <void func(Core::System&, u32, u64, u64, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     func(system, static_cast<u32>(Param(system, 0)), Param(system, 1), Param(system, 2),
          Param(system, 3));
 }
 
 template <void func(Core::System&, s64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     func(system, static_cast<s64>(Param(system, 0)));
 }
 
 template <void func(Core::System&, u64, s32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     func(system, Param(system, 0), static_cast<s32>(Param(system, 1)));
 }
 
 template <void func(Core::System&, u64, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     func(system, Param(system, 0), Param(system, 1));
 }
 
 template <void func(Core::System&, u64, u64, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     func(system, Param(system, 0), Param(system, 1), Param(system, 2));
 }
 
 template <void func(Core::System&, u32, u64, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     func(system, static_cast<u32>(Param(system, 0)), Param(system, 1), Param(system, 2));
 }
 
+// Used by QueryMemory32
+template <ResultCode func(Core::System&, u32, u32, u32)>
+void SvcWrap32(Core::System& system) {
+    FuncReturn32(system,
+                 func(system, Param32(system, 0), Param32(system, 1), Param32(system, 2)).raw);
+}
+
+// Used by GetInfo32
+template <ResultCode func(Core::System&, u32*, u32*, u32, u32, u32, u32)>
+void SvcWrap32(Core::System& system) {
+    u32 param_1 = 0;
+    u32 param_2 = 0;
+
+    const u32 retval = func(system, &param_1, &param_2, Param32(system, 0), Param32(system, 1),
+                            Param32(system, 2), Param32(system, 3))
+                           .raw;
+
+    system.CurrentArmInterface().SetReg(1, param_1);
+    system.CurrentArmInterface().SetReg(2, param_2);
+    FuncReturn(system, retval);
+}
+
+// Used by GetThreadPriority32, ConnectToNamedPort32
+template <ResultCode func(Core::System&, u32*, u32)>
+void SvcWrap32(Core::System& system) {
+    u32 param_1 = 0;
+    const u32 retval = func(system, &param_1, Param32(system, 1)).raw;
+    system.CurrentArmInterface().SetReg(1, param_1);
+    FuncReturn(system, retval);
+}
+
+// Used by GetThreadId32
+template <ResultCode func(Core::System&, u32*, u32*, u32)>
+void SvcWrap32(Core::System& system) {
+    u32 param_1 = 0;
+    u32 param_2 = 0;
+
+    const u32 retval = func(system, &param_1, &param_2, Param32(system, 1)).raw;
+    system.CurrentArmInterface().SetReg(1, param_1);
+    system.CurrentArmInterface().SetReg(2, param_2);
+    FuncReturn(system, retval);
+}
+
+// Used by SignalProcessWideKey32
+template <void func(Core::System&, u32, s32)>
+void SvcWrap32(Core::System& system) {
+    func(system, static_cast<u32>(Param(system, 0)), static_cast<s32>(Param(system, 1)));
+}
+
+// Used by SendSyncRequest32
+template <ResultCode func(Core::System&, u32)>
+void SvcWrap32(Core::System& system) {
+    FuncReturn(system, func(system, static_cast<u32>(Param(system, 0))).raw);
+}
+
+// Used by WaitSynchronization32
+template <ResultCode func(Core::System&, u32, u32, s32, u32, Handle*)>
+void SvcWrap32(Core::System& system) {
+    u32 param_1 = 0;
+    const u32 retval = func(system, Param32(system, 0), Param32(system, 1), Param32(system, 2),
+                            Param32(system, 3), &param_1)
+                           .raw;
+    system.CurrentArmInterface().SetReg(1, param_1);
+    FuncReturn(system, retval);
+}
+
 } // namespace Kernel

src/core/hle/kernel/thread.cpp

@@ -133,15 +133,16 @@ void Thread::CancelWait() {
     ResumeFromWait();
 }
 
-/**
- * Resets a thread context, making it ready to be scheduled and run by the CPU
- * @param context Thread context to reset
- * @param stack_top Address of the top of the stack
- * @param entry_point Address of entry point for execution
- * @param arg User argument for thread
- */
-static void ResetThreadContext(Core::ARM_Interface::ThreadContext& context, VAddr stack_top,
-                               VAddr entry_point, u64 arg) {
+static void ResetThreadContext32(Core::ARM_Interface::ThreadContext32& context, u32 stack_top,
+                                 u32 entry_point, u32 arg) {
+    context = {};
+    context.cpu_registers[0] = arg;
+    context.cpu_registers[15] = entry_point;
+    context.cpu_registers[13] = stack_top;
+}
+
+static void ResetThreadContext64(Core::ARM_Interface::ThreadContext64& context, VAddr stack_top,
+                                 VAddr entry_point, u64 arg) {
     context = {};
     context.cpu_registers[0] = arg;
     context.pc = entry_point;
@@ -198,9 +199,9 @@ ResultVal<std::shared_ptr<Thread>> Thread::Create(KernelCore& kernel, std::strin
 
     thread->owner_process->RegisterThread(thread.get());
 
-    // TODO(peachum): move to ScheduleThread() when scheduler is added so selected core is used
-    // to initialize the context
-    ResetThreadContext(thread->context, stack_top, entry_point, arg);
+    ResetThreadContext32(thread->context_32, static_cast<u32>(stack_top),
+                         static_cast<u32>(entry_point), static_cast<u32>(arg));
+    ResetThreadContext64(thread->context_64, stack_top, entry_point, arg);
 
     return MakeResult<std::shared_ptr<Thread>>(std::move(thread));
 }
@@ -213,11 +214,13 @@ void Thread::SetPriority(u32 priority) {
 }
 
 void Thread::SetWaitSynchronizationResult(ResultCode result) {
-    context.cpu_registers[0] = result.raw;
+    context_32.cpu_registers[0] = result.raw;
+    context_64.cpu_registers[0] = result.raw;
 }
 
 void Thread::SetWaitSynchronizationOutput(s32 output) {
-    context.cpu_registers[1] = output;
+    context_32.cpu_registers[1] = output;
+    context_64.cpu_registers[1] = output;
 }
 
 s32 Thread::GetSynchronizationObjectIndex(std::shared_ptr<SynchronizationObject> object) const {

src/core/hle/kernel/thread.h

@@ -102,7 +102,8 @@ public:
 
     using MutexWaitingThreads = std::vector<std::shared_ptr<Thread>>;
 
-    using ThreadContext = Core::ARM_Interface::ThreadContext;
+    using ThreadContext32 = Core::ARM_Interface::ThreadContext32;
+    using ThreadContext64 = Core::ARM_Interface::ThreadContext64;
 
     using ThreadSynchronizationObjects = std::vector<std::shared_ptr<SynchronizationObject>>;
 
@@ -273,12 +274,20 @@ public:
         return status == ThreadStatus::WaitSynch;
     }
 
-    ThreadContext& GetContext() {
-        return context;
+    ThreadContext32& GetContext32() {
+        return context_32;
     }
 
-    const ThreadContext& GetContext() const {
-        return context;
+    const ThreadContext32& GetContext32() const {
+        return context_32;
+    }
+
+    ThreadContext64& GetContext64() {
+        return context_64;
+    }
+
+    const ThreadContext64& GetContext64() const {
+        return context_64;
     }
 
     ThreadStatus GetStatus() const {
@@ -466,7 +475,8 @@ private:
     void AdjustSchedulingOnPriority(u32 old_priority);
     void AdjustSchedulingOnAffinity(u64 old_affinity_mask, s32 old_core);
 
-    Core::ARM_Interface::ThreadContext context{};
+    ThreadContext32 context_32{};
+    ThreadContext64 context_64{};
 
     u64 thread_id = 0;
 

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

@@ -607,7 +607,7 @@ ICommonStateGetter::ICommonStateGetter(Core::System& system,
         {40, nullptr, "GetCradleFwVersion"},
         {50, nullptr, "IsVrModeEnabled"},
         {51, nullptr, "SetVrModeEnabled"},
-        {52, nullptr, "SwitchLcdBacklight"},
+        {52, &ICommonStateGetter::SetLcdBacklighOffEnabled, "SetLcdBacklighOffEnabled"},
         {53, nullptr, "BeginVrModeEx"},
         {54, nullptr, "EndVrModeEx"},
         {55, nullptr, "IsInControllerFirmwareUpdateSection"},
@@ -636,7 +636,6 @@ void ICommonStateGetter::GetBootMode(Kernel::HLERequestContext& ctx) {
 
     IPC::ResponseBuilder rb{ctx, 3};
     rb.Push(RESULT_SUCCESS);
-
     rb.Push<u8>(static_cast<u8>(Service::PM::SystemBootMode::Normal)); // Normal boot mode
 }
 
@@ -660,6 +659,7 @@ void ICommonStateGetter::ReceiveMessage(Kernel::HLERequestContext& ctx) {
         rb.PushEnum<AppletMessageQueue::AppletMessage>(message);
         return;
     }
+
     rb.Push(RESULT_SUCCESS);
     rb.PushEnum<AppletMessageQueue::AppletMessage>(message);
 }
@@ -672,6 +672,17 @@ void ICommonStateGetter::GetCurrentFocusState(Kernel::HLERequestContext& ctx) {
     rb.Push(static_cast<u8>(FocusState::InFocus));
 }
 
+void ICommonStateGetter::SetLcdBacklighOffEnabled(Kernel::HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    const auto is_lcd_backlight_off_enabled = rp.Pop<bool>();
+
+    LOG_WARNING(Service_AM, "(STUBBED) called. is_lcd_backlight_off_enabled={}",
+                is_lcd_backlight_off_enabled);
+
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(RESULT_SUCCESS);
+}
+
 void ICommonStateGetter::GetDefaultDisplayResolutionChangeEvent(Kernel::HLERequestContext& ctx) {
     LOG_DEBUG(Service_AM, "called");
 

src/core/hle/service/am/am.h

@@ -182,6 +182,7 @@ private:
     void GetOperationMode(Kernel::HLERequestContext& ctx);
     void GetPerformanceMode(Kernel::HLERequestContext& ctx);
     void GetBootMode(Kernel::HLERequestContext& ctx);
+    void SetLcdBacklighOffEnabled(Kernel::HLERequestContext& ctx);
     void GetDefaultDisplayResolution(Kernel::HLERequestContext& ctx);
     void SetCpuBoostMode(Kernel::HLERequestContext& ctx);
 

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

@@ -287,13 +287,13 @@ void Controller_NPad::RequestPadStateUpdate(u32 npad_id) {
         analog_state[static_cast<std::size_t>(JoystickId::Joystick_Left)]->GetAnalogDirectionStatus(
             Input::AnalogDirection::DOWN));
 
-    pad_state.r_stick_up.Assign(analog_state[static_cast<std::size_t>(JoystickId::Joystick_Right)]
-                                    ->GetAnalogDirectionStatus(Input::AnalogDirection::RIGHT));
-    pad_state.r_stick_left.Assign(analog_state[static_cast<std::size_t>(JoystickId::Joystick_Right)]
-                                      ->GetAnalogDirectionStatus(Input::AnalogDirection::LEFT));
     pad_state.r_stick_right.Assign(
         analog_state[static_cast<std::size_t>(JoystickId::Joystick_Right)]
-            ->GetAnalogDirectionStatus(Input::AnalogDirection::UP));
+            ->GetAnalogDirectionStatus(Input::AnalogDirection::RIGHT));
+    pad_state.r_stick_left.Assign(analog_state[static_cast<std::size_t>(JoystickId::Joystick_Right)]
+                                      ->GetAnalogDirectionStatus(Input::AnalogDirection::LEFT));
+    pad_state.r_stick_up.Assign(analog_state[static_cast<std::size_t>(JoystickId::Joystick_Right)]
+                                    ->GetAnalogDirectionStatus(Input::AnalogDirection::UP));
     pad_state.r_stick_down.Assign(analog_state[static_cast<std::size_t>(JoystickId::Joystick_Right)]
                                       ->GetAnalogDirectionStatus(Input::AnalogDirection::DOWN));
 

src/core/loader/deconstructed_rom_directory.cpp

@@ -129,12 +129,6 @@ AppLoader_DeconstructedRomDirectory::LoadResult AppLoader_DeconstructedRomDirect
     }
     metadata.Print();
 
-    const FileSys::ProgramAddressSpaceType arch_bits{metadata.GetAddressSpaceType()};
-    if (arch_bits == FileSys::ProgramAddressSpaceType::Is32Bit ||
-        arch_bits == FileSys::ProgramAddressSpaceType::Is32BitNoMap) {
-        return {ResultStatus::Error32BitISA, {}};
-    }
-
     if (process.LoadFromMetadata(metadata).IsError()) {
         return {ResultStatus::ErrorUnableToParseKernelMetadata, {}};
     }

src/core/reporter.cpp

@@ -111,7 +111,7 @@ json GetProcessorStateDataAuto(Core::System& system) {
     const auto& vm_manager{process->VMManager()};
     auto& arm{system.CurrentArmInterface()};
 
-    Core::ARM_Interface::ThreadContext context{};
+    Core::ARM_Interface::ThreadContext64 context{};
     arm.SaveContext(context);
 
     return GetProcessorStateData(process->Is64BitProcess() ? "AArch64" : "AArch32",

src/core/settings.cpp

@@ -94,6 +94,7 @@ void LogSettings() {
     LogSetting("Renderer_UseAccurateGpuEmulation", Settings::values.use_accurate_gpu_emulation);
     LogSetting("Renderer_UseAsynchronousGpuEmulation",
                Settings::values.use_asynchronous_gpu_emulation);
+    LogSetting("Renderer_UseVsync", Settings::values.use_vsync);
     LogSetting("Audio_OutputEngine", Settings::values.sink_id);
     LogSetting("Audio_EnableAudioStretching", Settings::values.enable_audio_stretching);
     LogSetting("Audio_OutputDevice", Settings::values.audio_device_id);

src/core/settings.h

@@ -430,11 +430,13 @@ struct Values {
 
     float resolution_factor;
     int aspect_ratio;
+    int max_anisotropy;
     bool use_frame_limit;
     u16 frame_limit;
     bool use_disk_shader_cache;
     bool use_accurate_gpu_emulation;
     bool use_asynchronous_gpu_emulation;
+    bool use_vsync;
     bool force_30fps_mode;
 
     float bg_red;

src/core/telemetry_session.cpp

@@ -188,6 +188,7 @@ void TelemetrySession::AddInitialInfo(Loader::AppLoader& app_loader) {
              Settings::values.use_accurate_gpu_emulation);
     AddField(field_type, "Renderer_UseAsynchronousGpuEmulation",
              Settings::values.use_asynchronous_gpu_emulation);
+    AddField(field_type, "Renderer_UseVsync", Settings::values.use_vsync);
     AddField(field_type, "System_UseDockedMode", Settings::values.use_docked_mode);
 }
 

src/input_common/analog_from_button.cpp

@@ -34,6 +34,20 @@ public:
                                y * coef * (x == 0 ? 1.0f : SQRT_HALF));
     }
 
+    bool GetAnalogDirectionStatus(Input::AnalogDirection direction) const override {
+        switch (direction) {
+        case Input::AnalogDirection::RIGHT:
+            return right->GetStatus();
+        case Input::AnalogDirection::LEFT:
+            return left->GetStatus();
+        case Input::AnalogDirection::UP:
+            return up->GetStatus();
+        case Input::AnalogDirection::DOWN:
+            return down->GetStatus();
+        }
+        return false;
+    }
+
 private:
     Button up;
     Button down;

src/input_common/udp/client.cpp

@@ -32,8 +32,16 @@ public:
                     SocketCallback callback)
         : callback(std::move(callback)), timer(io_service),
           socket(io_service, udp::endpoint(udp::v4(), 0)), client_id(client_id),
-          pad_index(pad_index),
-          send_endpoint(udp::endpoint(boost::asio::ip::make_address_v4(host), port)) {}
+          pad_index(pad_index) {
+        boost::system::error_code ec{};
+        auto ipv4 = boost::asio::ip::make_address_v4(host, ec);
+        if (ec.failed()) {
+            LOG_ERROR(Input, "Invalid IPv4 address \"{}\" provided to socket", host);
+            ipv4 = boost::asio::ip::address_v4{};
+        }
+
+        send_endpoint = {udp::endpoint(ipv4, port)};
+    }
 
     void Stop() {
         io_service.stop();
@@ -85,17 +93,18 @@ private:
     }
 
     void HandleSend(const boost::system::error_code& error) {
+        boost::system::error_code _ignored{};
         // Send a request for getting port info for the pad
         Request::PortInfo port_info{1, {pad_index, 0, 0, 0}};
         const auto port_message = Request::Create(port_info, client_id);
         std::memcpy(&send_buffer1, &port_message, PORT_INFO_SIZE);
-        socket.send_to(boost::asio::buffer(send_buffer1), send_endpoint);
+        socket.send_to(boost::asio::buffer(send_buffer1), send_endpoint, {}, _ignored);
 
         // Send a request for getting pad data for the pad
         Request::PadData pad_data{Request::PadData::Flags::Id, pad_index, EMPTY_MAC_ADDRESS};
         const auto pad_message = Request::Create(pad_data, client_id);
         std::memcpy(send_buffer2.data(), &pad_message, PAD_DATA_SIZE);
-        socket.send_to(boost::asio::buffer(send_buffer2), send_endpoint);
+        socket.send_to(boost::asio::buffer(send_buffer2), send_endpoint, {}, _ignored);
         StartSend(timer.expiry());
     }
 

src/input_common/udp/protocol.cpp

@@ -31,7 +31,6 @@ namespace Response {
  */
 std::optional<Type> Validate(u8* data, std::size_t size) {
     if (size < sizeof(Header)) {
-        LOG_DEBUG(Input, "Invalid UDP packet received");
         return std::nullopt;
     }
     Header header{};

src/video_core/CMakeLists.txt

@@ -2,6 +2,8 @@ add_library(video_core STATIC
     buffer_cache/buffer_block.h
     buffer_cache/buffer_cache.h
     buffer_cache/map_interval.h
+    dirty_flags.cpp
+    dirty_flags.h
     dma_pusher.cpp
     dma_pusher.h
     engines/const_buffer_engine_interface.h
@@ -69,8 +71,8 @@ add_library(video_core STATIC
     renderer_opengl/gl_shader_manager.h
     renderer_opengl/gl_shader_util.cpp
     renderer_opengl/gl_shader_util.h
-    renderer_opengl/gl_state.cpp
-    renderer_opengl/gl_state.h
+    renderer_opengl/gl_state_tracker.cpp
+    renderer_opengl/gl_state_tracker.h
     renderer_opengl/gl_stream_buffer.cpp
     renderer_opengl/gl_stream_buffer.h
     renderer_opengl/gl_texture_cache.cpp
@@ -198,6 +200,8 @@ if (ENABLE_VULKAN)
         renderer_vulkan/vk_shader_util.h
         renderer_vulkan/vk_staging_buffer_pool.cpp
         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

src/video_core/dirty_flags.cpp

@@ -0,0 +1,46 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <array>
+#include <cstddef>
+
+#include "common/common_types.h"
+#include "video_core/dirty_flags.h"
+
+#define OFF(field_name) MAXWELL3D_REG_INDEX(field_name)
+#define NUM(field_name) (sizeof(::Tegra::Engines::Maxwell3D::Regs::field_name) / sizeof(u32))
+
+namespace VideoCommon::Dirty {
+
+using Tegra::Engines::Maxwell3D;
+
+void SetupCommonOnWriteStores(Tegra::Engines::Maxwell3D::DirtyState::Flags& store) {
+    store[RenderTargets] = true;
+    store[ZetaBuffer] = true;
+    for (std::size_t i = 0; i < Maxwell3D::Regs::NumRenderTargets; ++i) {
+        store[ColorBuffer0 + i] = true;
+    }
+}
+
+void SetupDirtyRenderTargets(Tegra::Engines::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;
+    for (std::size_t rt = 0; rt < Maxwell3D::Regs::NumRenderTargets; ++rt) {
+        FillBlock(tables[0], begin + rt * num_per_rt, num_per_rt, ColorBuffer0 + rt);
+    }
+    FillBlock(tables[1], begin, num, RenderTargets);
+
+    static constexpr std::array zeta_flags{ZetaBuffer, RenderTargets};
+    for (std::size_t i = 0; i < std::size(zeta_flags); ++i) {
+        const u8 flag = zeta_flags[i];
+        auto& table = tables[i];
+        table[OFF(zeta_enable)] = flag;
+        table[OFF(zeta_width)] = flag;
+        table[OFF(zeta_height)] = flag;
+        FillBlock(table, OFF(zeta), NUM(zeta), flag);
+    }
+}
+
+} // namespace VideoCommon::Dirty

src/video_core/dirty_flags.h

@@ -0,0 +1,51 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <algorithm>
+#include <cstddef>
+#include <iterator>
+
+#include "common/common_types.h"
+#include "video_core/engines/maxwell_3d.h"
+
+namespace VideoCommon::Dirty {
+
+enum : u8 {
+    NullEntry = 0,
+
+    RenderTargets,
+    ColorBuffer0,
+    ColorBuffer1,
+    ColorBuffer2,
+    ColorBuffer3,
+    ColorBuffer4,
+    ColorBuffer5,
+    ColorBuffer6,
+    ColorBuffer7,
+    ZetaBuffer,
+
+    LastCommonEntry,
+};
+
+template <typename Integer>
+void FillBlock(Tegra::Engines::Maxwell3D::DirtyState::Table& table, std::size_t begin,
+               std::size_t num, Integer dirty_index) {
+    const auto it = std::begin(table) + begin;
+    std::fill(it, it + num, static_cast<u8>(dirty_index));
+}
+
+template <typename Integer1, typename Integer2>
+void FillBlock(Tegra::Engines::Maxwell3D::DirtyState::Tables& tables, std::size_t begin,
+               std::size_t num, Integer1 index_a, Integer2 index_b) {
+    FillBlock(tables[0], begin, num, index_a);
+    FillBlock(tables[1], begin, num, index_b);
+}
+
+void SetupCommonOnWriteStores(Tegra::Engines::Maxwell3D::DirtyState::Flags& store);
+
+void SetupDirtyRenderTargets(Tegra::Engines::Maxwell3D::DirtyState::Tables& tables);
+
+} // namespace VideoCommon::Dirty

src/video_core/dma_pusher.cpp

@@ -22,7 +22,7 @@ void DmaPusher::DispatchCalls() {
     MICROPROFILE_SCOPE(DispatchCalls);
 
     // On entering GPU code, assume all memory may be touched by the ARM core.
-    gpu.Maxwell3D().dirty.OnMemoryWrite();
+    gpu.Maxwell3D().OnMemoryWrite();
 
     dma_pushbuffer_subindex = 0;
 

src/video_core/engines/kepler_compute.cpp

@@ -39,7 +39,7 @@ void KeplerCompute::CallMethod(const GPU::MethodCall& method_call) {
         const bool is_last_call = method_call.IsLastCall();
         upload_state.ProcessData(method_call.argument, is_last_call);
         if (is_last_call) {
-            system.GPU().Maxwell3D().dirty.OnMemoryWrite();
+            system.GPU().Maxwell3D().OnMemoryWrite();
         }
         break;
     }

src/video_core/engines/kepler_memory.cpp

@@ -34,7 +34,7 @@ void KeplerMemory::CallMethod(const GPU::MethodCall& method_call) {
         const bool is_last_call = method_call.IsLastCall();
         upload_state.ProcessData(method_call.argument, is_last_call);
         if (is_last_call) {
-            system.GPU().Maxwell3D().dirty.OnMemoryWrite();
+            system.GPU().Maxwell3D().OnMemoryWrite();
         }
         break;
     }

src/video_core/engines/maxwell_3d.cpp

@@ -26,7 +26,8 @@ Maxwell3D::Maxwell3D(Core::System& system, VideoCore::RasterizerInterface& raste
                      MemoryManager& memory_manager)
     : system{system}, rasterizer{rasterizer}, memory_manager{memory_manager},
       macro_interpreter{*this}, upload_state{memory_manager, regs.upload} {
-    InitDirtySettings();
+    dirty.flags.flip();
+
     InitializeRegisterDefaults();
 }
 
@@ -75,8 +76,8 @@ void Maxwell3D::InitializeRegisterDefaults() {
     regs.stencil_back_mask = 0xFFFFFFFF;
 
     regs.depth_test_func = Regs::ComparisonOp::Always;
-    regs.cull.front_face = Regs::Cull::FrontFace::CounterClockWise;
-    regs.cull.cull_face = Regs::Cull::CullFace::Back;
+    regs.front_face = Regs::FrontFace::CounterClockWise;
+    regs.cull_face = Regs::CullFace::Back;
 
     // TODO(Rodrigo): Most games do not set a point size. I think this is a case of a
     // register carrying a default value. Assume it's OpenGL's default (1).
@@ -95,7 +96,7 @@ void Maxwell3D::InitializeRegisterDefaults() {
     regs.rasterize_enable = 1;
     regs.rt_separate_frag_data = 1;
     regs.framebuffer_srgb = 1;
-    regs.cull.front_face = Maxwell3D::Regs::Cull::FrontFace::ClockWise;
+    regs.front_face = Maxwell3D::Regs::FrontFace::ClockWise;
 
     mme_inline[MAXWELL3D_REG_INDEX(draw.vertex_end_gl)] = true;
     mme_inline[MAXWELL3D_REG_INDEX(draw.vertex_begin_gl)] = true;
@@ -103,164 +104,6 @@ void Maxwell3D::InitializeRegisterDefaults() {
     mme_inline[MAXWELL3D_REG_INDEX(index_array.count)] = true;
 }
 
-#define DIRTY_REGS_POS(field_name) static_cast<u8>(offsetof(Maxwell3D::DirtyRegs, field_name))
-
-void Maxwell3D::InitDirtySettings() {
-    const auto set_block = [this](std::size_t start, std::size_t range, u8 position) {
-        const auto start_itr = dirty_pointers.begin() + start;
-        const auto end_itr = start_itr + range;
-        std::fill(start_itr, end_itr, position);
-    };
-    dirty.regs.fill(true);
-
-    // Init Render Targets
-    constexpr u32 registers_per_rt = sizeof(regs.rt[0]) / sizeof(u32);
-    constexpr u32 rt_start_reg = MAXWELL3D_REG_INDEX(rt);
-    constexpr u32 rt_end_reg = rt_start_reg + registers_per_rt * 8;
-    u8 rt_dirty_reg = DIRTY_REGS_POS(render_target);
-    for (u32 rt_reg = rt_start_reg; rt_reg < rt_end_reg; rt_reg += registers_per_rt) {
-        set_block(rt_reg, registers_per_rt, rt_dirty_reg);
-        ++rt_dirty_reg;
-    }
-    constexpr u32 depth_buffer_flag = DIRTY_REGS_POS(depth_buffer);
-    dirty_pointers[MAXWELL3D_REG_INDEX(zeta_enable)] = depth_buffer_flag;
-    dirty_pointers[MAXWELL3D_REG_INDEX(zeta_width)] = depth_buffer_flag;
-    dirty_pointers[MAXWELL3D_REG_INDEX(zeta_height)] = depth_buffer_flag;
-    constexpr u32 registers_in_zeta = sizeof(regs.zeta) / sizeof(u32);
-    constexpr u32 zeta_reg = MAXWELL3D_REG_INDEX(zeta);
-    set_block(zeta_reg, registers_in_zeta, depth_buffer_flag);
-
-    // Init Vertex Arrays
-    constexpr u32 vertex_array_start = MAXWELL3D_REG_INDEX(vertex_array);
-    constexpr u32 vertex_array_size = sizeof(regs.vertex_array[0]) / sizeof(u32);
-    constexpr u32 vertex_array_end = vertex_array_start + vertex_array_size * Regs::NumVertexArrays;
-    u8 va_dirty_reg = DIRTY_REGS_POS(vertex_array);
-    u8 vi_dirty_reg = DIRTY_REGS_POS(vertex_instance);
-    for (u32 vertex_reg = vertex_array_start; vertex_reg < vertex_array_end;
-         vertex_reg += vertex_array_size) {
-        set_block(vertex_reg, 3, va_dirty_reg);
-        // The divisor concerns vertex array instances
-        dirty_pointers[static_cast<std::size_t>(vertex_reg) + 3] = vi_dirty_reg;
-        ++va_dirty_reg;
-        ++vi_dirty_reg;
-    }
-    constexpr u32 vertex_limit_start = MAXWELL3D_REG_INDEX(vertex_array_limit);
-    constexpr u32 vertex_limit_size = sizeof(regs.vertex_array_limit[0]) / sizeof(u32);
-    constexpr u32 vertex_limit_end = vertex_limit_start + vertex_limit_size * Regs::NumVertexArrays;
-    va_dirty_reg = DIRTY_REGS_POS(vertex_array);
-    for (u32 vertex_reg = vertex_limit_start; vertex_reg < vertex_limit_end;
-         vertex_reg += vertex_limit_size) {
-        set_block(vertex_reg, vertex_limit_size, va_dirty_reg);
-        va_dirty_reg++;
-    }
-    constexpr u32 vertex_instance_start = MAXWELL3D_REG_INDEX(instanced_arrays);
-    constexpr u32 vertex_instance_size =
-        sizeof(regs.instanced_arrays.is_instanced[0]) / sizeof(u32);
-    constexpr u32 vertex_instance_end =
-        vertex_instance_start + vertex_instance_size * Regs::NumVertexArrays;
-    vi_dirty_reg = DIRTY_REGS_POS(vertex_instance);
-    for (u32 vertex_reg = vertex_instance_start; vertex_reg < vertex_instance_end;
-         vertex_reg += vertex_instance_size) {
-        set_block(vertex_reg, vertex_instance_size, vi_dirty_reg);
-        vi_dirty_reg++;
-    }
-    set_block(MAXWELL3D_REG_INDEX(vertex_attrib_format), regs.vertex_attrib_format.size(),
-              DIRTY_REGS_POS(vertex_attrib_format));
-
-    // Init Shaders
-    constexpr u32 shader_registers_count =
-        sizeof(regs.shader_config[0]) * Regs::MaxShaderProgram / sizeof(u32);
-    set_block(MAXWELL3D_REG_INDEX(shader_config[0]), shader_registers_count,
-              DIRTY_REGS_POS(shaders));
-
-    // State
-
-    // Viewport
-    constexpr u8 viewport_dirty_reg = DIRTY_REGS_POS(viewport);
-    constexpr u32 viewport_start = MAXWELL3D_REG_INDEX(viewports);
-    constexpr u32 viewport_size = sizeof(regs.viewports) / sizeof(u32);
-    set_block(viewport_start, viewport_size, viewport_dirty_reg);
-    constexpr u32 view_volume_start = MAXWELL3D_REG_INDEX(view_volume_clip_control);
-    constexpr u32 view_volume_size = sizeof(regs.view_volume_clip_control) / sizeof(u32);
-    set_block(view_volume_start, view_volume_size, viewport_dirty_reg);
-
-    // Viewport transformation
-    constexpr u32 viewport_trans_start = MAXWELL3D_REG_INDEX(viewport_transform);
-    constexpr u32 viewport_trans_size = sizeof(regs.viewport_transform) / sizeof(u32);
-    set_block(viewport_trans_start, viewport_trans_size, DIRTY_REGS_POS(viewport_transform));
-
-    // Cullmode
-    constexpr u32 cull_mode_start = MAXWELL3D_REG_INDEX(cull);
-    constexpr u32 cull_mode_size = sizeof(regs.cull) / sizeof(u32);
-    set_block(cull_mode_start, cull_mode_size, DIRTY_REGS_POS(cull_mode));
-
-    // Screen y control
-    dirty_pointers[MAXWELL3D_REG_INDEX(screen_y_control)] = DIRTY_REGS_POS(screen_y_control);
-
-    // Primitive Restart
-    constexpr u32 primitive_restart_start = MAXWELL3D_REG_INDEX(primitive_restart);
-    constexpr u32 primitive_restart_size = sizeof(regs.primitive_restart) / sizeof(u32);
-    set_block(primitive_restart_start, primitive_restart_size, DIRTY_REGS_POS(primitive_restart));
-
-    // Depth Test
-    constexpr u8 depth_test_dirty_reg = DIRTY_REGS_POS(depth_test);
-    dirty_pointers[MAXWELL3D_REG_INDEX(depth_test_enable)] = depth_test_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(depth_write_enabled)] = depth_test_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(depth_test_func)] = depth_test_dirty_reg;
-
-    // Stencil Test
-    constexpr u32 stencil_test_dirty_reg = DIRTY_REGS_POS(stencil_test);
-    dirty_pointers[MAXWELL3D_REG_INDEX(stencil_enable)] = stencil_test_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(stencil_front_func_func)] = stencil_test_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(stencil_front_func_ref)] = stencil_test_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(stencil_front_func_mask)] = stencil_test_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(stencil_front_op_fail)] = stencil_test_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(stencil_front_op_zfail)] = stencil_test_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(stencil_front_op_zpass)] = stencil_test_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(stencil_front_mask)] = stencil_test_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(stencil_two_side_enable)] = stencil_test_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(stencil_back_func_func)] = stencil_test_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(stencil_back_func_ref)] = stencil_test_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(stencil_back_func_mask)] = stencil_test_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(stencil_back_op_fail)] = stencil_test_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(stencil_back_op_zfail)] = stencil_test_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(stencil_back_op_zpass)] = stencil_test_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(stencil_back_mask)] = stencil_test_dirty_reg;
-
-    // Color Mask
-    constexpr u8 color_mask_dirty_reg = DIRTY_REGS_POS(color_mask);
-    dirty_pointers[MAXWELL3D_REG_INDEX(color_mask_common)] = color_mask_dirty_reg;
-    set_block(MAXWELL3D_REG_INDEX(color_mask), sizeof(regs.color_mask) / sizeof(u32),
-              color_mask_dirty_reg);
-    // Blend State
-    constexpr u8 blend_state_dirty_reg = DIRTY_REGS_POS(blend_state);
-    set_block(MAXWELL3D_REG_INDEX(blend_color), sizeof(regs.blend_color) / sizeof(u32),
-              blend_state_dirty_reg);
-    dirty_pointers[MAXWELL3D_REG_INDEX(independent_blend_enable)] = blend_state_dirty_reg;
-    set_block(MAXWELL3D_REG_INDEX(blend), sizeof(regs.blend) / sizeof(u32), blend_state_dirty_reg);
-    set_block(MAXWELL3D_REG_INDEX(independent_blend), sizeof(regs.independent_blend) / sizeof(u32),
-              blend_state_dirty_reg);
-
-    // Scissor State
-    constexpr u8 scissor_test_dirty_reg = DIRTY_REGS_POS(scissor_test);
-    set_block(MAXWELL3D_REG_INDEX(scissor_test), sizeof(regs.scissor_test) / sizeof(u32),
-              scissor_test_dirty_reg);
-
-    // Polygon Offset
-    constexpr u8 polygon_offset_dirty_reg = DIRTY_REGS_POS(polygon_offset);
-    dirty_pointers[MAXWELL3D_REG_INDEX(polygon_offset_fill_enable)] = polygon_offset_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(polygon_offset_line_enable)] = polygon_offset_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(polygon_offset_point_enable)] = polygon_offset_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(polygon_offset_units)] = polygon_offset_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(polygon_offset_factor)] = polygon_offset_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(polygon_offset_clamp)] = polygon_offset_dirty_reg;
-
-    // Depth bounds
-    constexpr u8 depth_bounds_values_dirty_reg = DIRTY_REGS_POS(depth_bounds_values);
-    dirty_pointers[MAXWELL3D_REG_INDEX(depth_bounds[0])] = depth_bounds_values_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(depth_bounds[1])] = depth_bounds_values_dirty_reg;
-}
-
 void Maxwell3D::CallMacroMethod(u32 method, std::size_t num_parameters, const u32* parameters) {
     // Reset the current macro.
     executing_macro = 0;
@@ -319,19 +162,9 @@ void Maxwell3D::CallMethod(const GPU::MethodCall& method_call) {
 
     if (regs.reg_array[method] != method_call.argument) {
         regs.reg_array[method] = method_call.argument;
-        const std::size_t dirty_reg = dirty_pointers[method];
-        if (dirty_reg) {
-            dirty.regs[dirty_reg] = true;
-            if (dirty_reg >= DIRTY_REGS_POS(vertex_array) &&
-                dirty_reg < DIRTY_REGS_POS(vertex_array_buffers)) {
-                dirty.vertex_array_buffers = true;
-            } else if (dirty_reg >= DIRTY_REGS_POS(vertex_instance) &&
-                       dirty_reg < DIRTY_REGS_POS(vertex_instances)) {
-                dirty.vertex_instances = true;
-            } else if (dirty_reg >= DIRTY_REGS_POS(render_target) &&
-                       dirty_reg < DIRTY_REGS_POS(render_settings)) {
-                dirty.render_settings = true;
-            }
+
+        for (const auto& table : dirty.tables) {
+            dirty.flags[table[method]] = true;
         }
     }
 
@@ -419,7 +252,7 @@ void Maxwell3D::CallMethod(const GPU::MethodCall& method_call) {
         const bool is_last_call = method_call.IsLastCall();
         upload_state.ProcessData(method_call.argument, is_last_call);
         if (is_last_call) {
-            dirty.OnMemoryWrite();
+            OnMemoryWrite();
         }
         break;
     }
@@ -727,7 +560,7 @@ void Maxwell3D::FinishCBData() {
 
     const u32 id = cb_data_state.id;
     memory_manager.WriteBlock(address, cb_data_state.buffer[id].data(), size);
-    dirty.OnMemoryWrite();
+    OnMemoryWrite();
 
     cb_data_state.id = null_cb_data;
     cb_data_state.current = null_cb_data;

src/video_core/engines/maxwell_3d.h

@@ -6,6 +6,7 @@
 
 #include <array>
 #include <bitset>
+#include <limits>
 #include <optional>
 #include <type_traits>
 #include <unordered_map>
@@ -431,21 +432,15 @@ public:
             GeneratedPrimitives = 0x1F,
         };
 
-        struct Cull {
-            enum class FrontFace : u32 {
-                ClockWise = 0x0900,
-                CounterClockWise = 0x0901,
-            };
+        enum class FrontFace : u32 {
+            ClockWise = 0x0900,
+            CounterClockWise = 0x0901,
+        };
 
-            enum class CullFace : u32 {
-                Front = 0x0404,
-                Back = 0x0405,
-                FrontAndBack = 0x0408,
-            };
-
-            u32 enabled;
-            FrontFace front_face;
-            CullFace cull_face;
+        enum class CullFace : u32 {
+            Front = 0x0404,
+            Back = 0x0405,
+            FrontAndBack = 0x0408,
         };
 
         struct Blend {
@@ -574,7 +569,7 @@ public:
             f32 translate_z;
             INSERT_UNION_PADDING_WORDS(2);
 
-            Common::Rectangle<s32> GetRect() const {
+            Common::Rectangle<f32> GetRect() const {
                 return {
                     GetX(),               // left
                     GetY() + GetHeight(), // top
@@ -583,20 +578,20 @@ public:
                 };
             };
 
-            s32 GetX() const {
-                return static_cast<s32>(std::max(0.0f, translate_x - std::fabs(scale_x)));
+            f32 GetX() const {
+                return std::max(0.0f, translate_x - std::fabs(scale_x));
             }
 
-            s32 GetY() const {
-                return static_cast<s32>(std::max(0.0f, translate_y - std::fabs(scale_y)));
+            f32 GetY() const {
+                return std::max(0.0f, translate_y - std::fabs(scale_y));
             }
 
-            s32 GetWidth() const {
-                return static_cast<s32>(translate_x + std::fabs(scale_x)) - GetX();
+            f32 GetWidth() const {
+                return translate_x + std::fabs(scale_x) - GetX();
             }
 
-            s32 GetHeight() const {
-                return static_cast<s32>(translate_y + std::fabs(scale_y)) - GetY();
+            f32 GetHeight() const {
+                return translate_y + std::fabs(scale_y) - GetY();
             }
         };
 
@@ -872,16 +867,7 @@ public:
 
                 INSERT_UNION_PADDING_WORDS(0x35);
 
-                union {
-                    BitField<0, 1, u32> c0;
-                    BitField<1, 1, u32> c1;
-                    BitField<2, 1, u32> c2;
-                    BitField<3, 1, u32> c3;
-                    BitField<4, 1, u32> c4;
-                    BitField<5, 1, u32> c5;
-                    BitField<6, 1, u32> c6;
-                    BitField<7, 1, u32> c7;
-                } clip_distance_enabled;
+                u32 clip_distance_enabled;
 
                 u32 samplecnt_enable;
 
@@ -1060,7 +1046,9 @@ public:
 
                 INSERT_UNION_PADDING_WORDS(1);
 
-                Cull cull;
+                u32 cull_test_enabled;
+                FrontFace front_face;
+                CullFace cull_face;
 
                 u32 pixel_center_integer;
 
@@ -1238,79 +1226,6 @@ public:
 
     State state{};
 
-    struct DirtyRegs {
-        static constexpr std::size_t NUM_REGS = 256;
-        static_assert(NUM_REGS - 1 <= std::numeric_limits<u8>::max());
-
-        union {
-            struct {
-                bool null_dirty;
-
-                // Vertex Attributes
-                bool vertex_attrib_format;
-
-                // Vertex Arrays
-                std::array<bool, 32> vertex_array;
-
-                bool vertex_array_buffers;
-
-                // Vertex Instances
-                std::array<bool, 32> vertex_instance;
-
-                bool vertex_instances;
-
-                // Render Targets
-                std::array<bool, 8> render_target;
-                bool depth_buffer;
-
-                bool render_settings;
-
-                // Shaders
-                bool shaders;
-
-                // Rasterizer State
-                bool viewport;
-                bool clip_coefficient;
-                bool cull_mode;
-                bool primitive_restart;
-                bool depth_test;
-                bool stencil_test;
-                bool blend_state;
-                bool scissor_test;
-                bool transform_feedback;
-                bool color_mask;
-                bool polygon_offset;
-                bool depth_bounds_values;
-
-                // Complementary
-                bool viewport_transform;
-                bool screen_y_control;
-
-                bool memory_general;
-            };
-            std::array<bool, NUM_REGS> regs;
-        };
-
-        void ResetVertexArrays() {
-            vertex_array.fill(true);
-            vertex_array_buffers = true;
-        }
-
-        void ResetRenderTargets() {
-            depth_buffer = true;
-            render_target.fill(true);
-            render_settings = true;
-        }
-
-        void OnMemoryWrite() {
-            shaders = true;
-            memory_general = true;
-            ResetRenderTargets();
-            ResetVertexArrays();
-        }
-
-    } dirty{};
-
     /// Reads a register value located at the input method address
     u32 GetRegisterValue(u32 method) const;
 
@@ -1356,6 +1271,11 @@ public:
         return execute_on;
     }
 
+    /// Notify a memory write has happened.
+    void OnMemoryWrite() {
+        dirty.flags |= dirty.on_write_stores;
+    }
+
     enum class MMEDrawMode : u32 {
         Undefined,
         Array,
@@ -1371,6 +1291,16 @@ public:
         u32 gl_end_count{};
     } mme_draw;
 
+    struct DirtyState {
+        using Flags = std::bitset<std::numeric_limits<u8>::max()>;
+        using Table = std::array<u8, Regs::NUM_REGS>;
+        using Tables = std::array<Table, 2>;
+
+        Flags flags;
+        Flags on_write_stores;
+        Tables tables{};
+    } dirty;
+
 private:
     void InitializeRegisterDefaults();
 
@@ -1417,8 +1347,6 @@ private:
     /// Retrieves information about a specific TSC entry from the TSC buffer.
     Texture::TSCEntry GetTSCEntry(u32 tsc_index) const;
 
-    void InitDirtySettings();
-
     /**
      * Call a macro on this engine.
      * @param method Method to call
@@ -1561,7 +1489,9 @@ ASSERT_REG_POSITION(index_array, 0x5F2);
 ASSERT_REG_POSITION(polygon_offset_clamp, 0x61F);
 ASSERT_REG_POSITION(instanced_arrays, 0x620);
 ASSERT_REG_POSITION(vp_point_size, 0x644);
-ASSERT_REG_POSITION(cull, 0x646);
+ASSERT_REG_POSITION(cull_test_enabled, 0x646);
+ASSERT_REG_POSITION(front_face, 0x647);
+ASSERT_REG_POSITION(cull_face, 0x648);
 ASSERT_REG_POSITION(pixel_center_integer, 0x649);
 ASSERT_REG_POSITION(viewport_transform_enabled, 0x64B);
 ASSERT_REG_POSITION(view_volume_clip_control, 0x64F);

src/video_core/engines/maxwell_dma.cpp

@@ -57,7 +57,7 @@ void MaxwellDMA::HandleCopy() {
     }
 
     // All copies here update the main memory, so mark all rasterizer states as invalid.
-    system.GPU().Maxwell3D().dirty.OnMemoryWrite();
+    system.GPU().Maxwell3D().OnMemoryWrite();
 
     if (regs.exec.is_dst_linear && regs.exec.is_src_linear) {
         // When the enable_2d bit is disabled, the copy is performed as if we were copying a 1D

src/video_core/gpu.cpp

@@ -140,71 +140,6 @@ void GPU::FlushCommands() {
     renderer.Rasterizer().FlushCommands();
 }
 
-u32 RenderTargetBytesPerPixel(RenderTargetFormat format) {
-    ASSERT(format != RenderTargetFormat::NONE);
-
-    switch (format) {
-    case RenderTargetFormat::RGBA32_FLOAT:
-    case RenderTargetFormat::RGBA32_UINT:
-        return 16;
-    case RenderTargetFormat::RGBA16_UINT:
-    case RenderTargetFormat::RGBA16_UNORM:
-    case RenderTargetFormat::RGBA16_FLOAT:
-    case RenderTargetFormat::RGBX16_FLOAT:
-    case RenderTargetFormat::RG32_FLOAT:
-    case RenderTargetFormat::RG32_UINT:
-        return 8;
-    case RenderTargetFormat::RGBA8_UNORM:
-    case RenderTargetFormat::RGBA8_SNORM:
-    case RenderTargetFormat::RGBA8_SRGB:
-    case RenderTargetFormat::RGBA8_UINT:
-    case RenderTargetFormat::RGB10_A2_UNORM:
-    case RenderTargetFormat::BGRA8_UNORM:
-    case RenderTargetFormat::BGRA8_SRGB:
-    case RenderTargetFormat::RG16_UNORM:
-    case RenderTargetFormat::RG16_SNORM:
-    case RenderTargetFormat::RG16_UINT:
-    case RenderTargetFormat::RG16_SINT:
-    case RenderTargetFormat::RG16_FLOAT:
-    case RenderTargetFormat::R32_FLOAT:
-    case RenderTargetFormat::R11G11B10_FLOAT:
-    case RenderTargetFormat::R32_UINT:
-        return 4;
-    case RenderTargetFormat::R16_UNORM:
-    case RenderTargetFormat::R16_SNORM:
-    case RenderTargetFormat::R16_UINT:
-    case RenderTargetFormat::R16_SINT:
-    case RenderTargetFormat::R16_FLOAT:
-    case RenderTargetFormat::RG8_UNORM:
-    case RenderTargetFormat::RG8_SNORM:
-        return 2;
-    case RenderTargetFormat::R8_UNORM:
-    case RenderTargetFormat::R8_UINT:
-        return 1;
-    default:
-        UNIMPLEMENTED_MSG("Unimplemented render target format {}", static_cast<u32>(format));
-        return 1;
-    }
-}
-
-u32 DepthFormatBytesPerPixel(DepthFormat format) {
-    switch (format) {
-    case DepthFormat::Z32_S8_X24_FLOAT:
-        return 8;
-    case DepthFormat::Z32_FLOAT:
-    case DepthFormat::S8_Z24_UNORM:
-    case DepthFormat::Z24_X8_UNORM:
-    case DepthFormat::Z24_S8_UNORM:
-    case DepthFormat::Z24_C8_UNORM:
-        return 4;
-    case DepthFormat::Z16_UNORM:
-        return 2;
-    default:
-        UNIMPLEMENTED_MSG("Unimplemented Depth format {}", static_cast<u32>(format));
-        return 1;
-    }
-}
-
 // Note that, traditionally, methods are treated as 4-byte addressable locations, and hence
 // their numbers are written down multiplied by 4 in Docs. Here we are not multiply by 4.
 // So the values you see in docs might be multiplied by 4.

src/video_core/gpu.h

@@ -83,12 +83,6 @@ enum class DepthFormat : u32 {
     Z32_S8_X24_FLOAT = 0x19,
 };
 
-/// Returns the number of bytes per pixel of each rendertarget format.
-u32 RenderTargetBytesPerPixel(RenderTargetFormat format);
-
-/// Returns the number of bytes per pixel of each depth format.
-u32 DepthFormatBytesPerPixel(DepthFormat format);
-
 struct CommandListHeader;
 class DebugContext;
 

src/video_core/gpu_thread.cpp

@@ -5,7 +5,7 @@
 #include "common/assert.h"
 #include "common/microprofile.h"
 #include "core/core.h"
-#include "core/frontend/scope_acquire_window_context.h"
+#include "core/frontend/scope_acquire_context.h"
 #include "video_core/dma_pusher.h"
 #include "video_core/gpu.h"
 #include "video_core/gpu_thread.h"
@@ -27,7 +27,7 @@ static void RunThread(VideoCore::RendererBase& renderer, Tegra::DmaPusher& dma_p
         return;
     }
 
-    Core::Frontend::ScopeAcquireWindowContext acquire_context{renderer.GetRenderWindow()};
+    Core::Frontend::ScopeAcquireContext acquire_context{renderer.GetRenderWindow()};
 
     CommandDataContainer next;
     while (state.is_running) {

src/video_core/rasterizer_interface.h

@@ -89,6 +89,9 @@ public:
     virtual void LoadDiskResources(const std::atomic_bool& stop_loading = false,
                                    const DiskResourceLoadCallback& callback = {}) {}
 
+    /// Initializes renderer dirty flags
+    virtual void SetupDirtyFlags() {}
+
     /// Grant access to the Guest Driver Profile for recording/obtaining info on the guest driver.
     GuestDriverProfile& AccessGuestDriverProfile() {
         return guest_driver_profile;

src/video_core/renderer_base.h

@@ -35,15 +35,19 @@ public:
     explicit RendererBase(Core::Frontend::EmuWindow& window);
     virtual ~RendererBase();
 
-    /// Swap buffers (render frame)
-    virtual void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) = 0;
-
     /// Initialize the renderer
     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;
+
+    /// Draws the latest frame to the window waiting timeout_ms for a frame to arrive (Renderer
+    /// specific implementation)
+    virtual void TryPresent(int timeout_ms) = 0;
+
     // Getter/setter functions:
     // ------------------------
 

src/video_core/renderer_opengl/gl_framebuffer_cache.cpp

@@ -11,7 +11,6 @@
 #include "common/common_types.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/renderer_opengl/gl_framebuffer_cache.h"
-#include "video_core/renderer_opengl/gl_state.h"
 
 namespace OpenGL {
 
@@ -36,8 +35,7 @@ OGLFramebuffer FramebufferCacheOpenGL::CreateFramebuffer(const FramebufferCacheK
     framebuffer.Create();
 
     // TODO(Rodrigo): Use DSA here after Nvidia fixes their framebuffer DSA bugs.
-    local_state.draw.draw_framebuffer = framebuffer.handle;
-    local_state.ApplyFramebufferState();
+    glBindFramebuffer(GL_DRAW_FRAMEBUFFER, framebuffer.handle);
 
     if (key.zeta) {
         const bool stencil = key.zeta->GetSurfaceParams().type == SurfaceType::DepthStencil;

src/video_core/renderer_opengl/gl_framebuffer_cache.h

@@ -13,7 +13,6 @@
 #include "common/common_types.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/renderer_opengl/gl_resource_manager.h"
-#include "video_core/renderer_opengl/gl_state.h"
 #include "video_core/renderer_opengl/gl_texture_cache.h"
 
 namespace OpenGL {
@@ -63,7 +62,6 @@ public:
 private:
     OGLFramebuffer CreateFramebuffer(const FramebufferCacheKey& key);
 
-    OpenGLState local_state;
     std::unordered_map<FramebufferCacheKey, OGLFramebuffer> cache;
 };
 

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -30,7 +30,7 @@
 #include "video_core/renderer_opengl/gl_shader_cache.h"
 #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.h"
+#include "video_core/renderer_opengl/gl_state_tracker.h"
 #include "video_core/renderer_opengl/gl_texture_cache.h"
 #include "video_core/renderer_opengl/utils.h"
 #include "video_core/textures/texture.h"
@@ -55,7 +55,8 @@ struct DrawParameters;
 class RasterizerOpenGL : public VideoCore::RasterizerAccelerated {
 public:
     explicit RasterizerOpenGL(Core::System& system, Core::Frontend::EmuWindow& emu_window,
-                              ScreenInfo& info);
+                              ScreenInfo& info, GLShader::ProgramManager& program_manager,
+                              StateTracker& state_tracker);
     ~RasterizerOpenGL() override;
 
     void Draw(bool is_indexed, bool is_instanced) override;
@@ -76,6 +77,7 @@ public:
                            u32 pixel_stride) override;
     void LoadDiskResources(const std::atomic_bool& stop_loading,
                            const VideoCore::DiskResourceLoadCallback& callback) override;
+    void SetupDirtyFlags() override;
 
     /// Returns true when there are commands queued to the OpenGL server.
     bool AnyCommandQueued() const {
@@ -86,8 +88,7 @@ private:
     /// Configures the color and depth framebuffer states.
     void ConfigureFramebuffers();
 
-    void ConfigureClearFramebuffer(OpenGLState& current_state, bool using_color_fb,
-                                   bool using_depth_fb, bool using_stencil_fb);
+    void ConfigureClearFramebuffer(bool using_color_fb, bool using_depth_fb, bool using_stencil_fb);
 
     /// Configures the current constbuffers to use for the draw command.
     void SetupDrawConstBuffers(std::size_t stage_index, const Shader& shader);
@@ -130,11 +131,13 @@ private:
                     const GLShader::ImageEntry& entry);
 
     /// Syncs the viewport and depth range to match the guest state
-    void SyncViewport(OpenGLState& current_state);
+    void SyncViewport();
+
+    /// Syncs the depth clamp state
+    void SyncDepthClamp();
 
     /// Syncs the clip enabled status to match the guest state
-    void SyncClipEnabled(
-        const std::array<bool, Tegra::Engines::Maxwell3D::Regs::NumClipDistances>& clip_mask);
+    void SyncClipEnabled(u32 clip_mask);
 
     /// Syncs the clip coefficients to match the guest state
     void SyncClipCoef();
@@ -164,7 +167,7 @@ private:
     void SyncMultiSampleState();
 
     /// Syncs the scissor test state to match the guest state
-    void SyncScissorTest(OpenGLState& current_state);
+    void SyncScissorTest();
 
     /// Syncs the transform feedback state to match the guest state
     void SyncTransformFeedback();
@@ -173,7 +176,7 @@ private:
     void SyncPointState();
 
     /// Syncs the rasterizer enable state to match the guest state
-    void SyncRasterizeEnable(OpenGLState& current_state);
+    void SyncRasterizeEnable();
 
     /// Syncs Color Mask
     void SyncColorMask();
@@ -184,6 +187,9 @@ private:
     /// Syncs the alpha test state to match the guest state
     void SyncAlphaTest();
 
+    /// Syncs the framebuffer sRGB state to match the guest state
+    void SyncFramebufferSRGB();
+
     /// Check for extension that are not strictly required but are needed for correct emulation
     void CheckExtensions();
 
@@ -191,18 +197,17 @@ private:
 
     std::size_t CalculateIndexBufferSize() const;
 
-    /// Updates and returns a vertex array object representing current vertex format
-    GLuint SetupVertexFormat();
+    /// Updates the current vertex format
+    void SetupVertexFormat();
 
-    void SetupVertexBuffer(GLuint vao);
-    void SetupVertexInstances(GLuint vao);
+    void SetupVertexBuffer();
+    void SetupVertexInstances();
 
     GLintptr SetupIndexBuffer();
 
     void SetupShaders(GLenum primitive_mode);
 
     const Device device;
-    OpenGLState state;
 
     TextureCacheOpenGL texture_cache;
     ShaderCacheOpenGL shader_cache;
@@ -212,22 +217,20 @@ private:
 
     Core::System& system;
     ScreenInfo& screen_info;
-
-    std::unique_ptr<GLShader::ProgramManager> shader_program_manager;
-    std::map<std::array<Tegra::Engines::Maxwell3D::Regs::VertexAttribute,
-                        Tegra::Engines::Maxwell3D::Regs::NumVertexAttributes>,
-             OGLVertexArray>
-        vertex_array_cache;
+    GLShader::ProgramManager& program_manager;
+    StateTracker& state_tracker;
 
     static constexpr std::size_t STREAM_BUFFER_SIZE = 128 * 1024 * 1024;
     OGLBufferCache buffer_cache;
 
-    VertexArrayPushBuffer vertex_array_pushbuffer;
+    VertexArrayPushBuffer vertex_array_pushbuffer{state_tracker};
     BindBuffersRangePushBuffer bind_ubo_pushbuffer{GL_UNIFORM_BUFFER};
     BindBuffersRangePushBuffer bind_ssbo_pushbuffer{GL_SHADER_STORAGE_BUFFER};
 
     /// Number of commands queued to the OpenGL driver. Reseted on flush.
     std::size_t num_queued_commands = 0;
+
+    u32 last_clip_distance_mask = 0;
 };
 
 } // namespace OpenGL

src/video_core/renderer_opengl/gl_resource_manager.cpp

@@ -8,13 +8,29 @@
 #include "common/microprofile.h"
 #include "video_core/renderer_opengl/gl_resource_manager.h"
 #include "video_core/renderer_opengl/gl_shader_util.h"
-#include "video_core/renderer_opengl/gl_state.h"
 
 MICROPROFILE_DEFINE(OpenGL_ResourceCreation, "OpenGL", "Resource Creation", MP_RGB(128, 128, 192));
 MICROPROFILE_DEFINE(OpenGL_ResourceDeletion, "OpenGL", "Resource Deletion", MP_RGB(128, 128, 192));
 
 namespace OpenGL {
 
+void OGLRenderbuffer::Create() {
+    if (handle != 0)
+        return;
+
+    MICROPROFILE_SCOPE(OpenGL_ResourceCreation);
+    glCreateRenderbuffers(1, &handle);
+}
+
+void OGLRenderbuffer::Release() {
+    if (handle == 0)
+        return;
+
+    MICROPROFILE_SCOPE(OpenGL_ResourceDeletion);
+    glDeleteRenderbuffers(1, &handle);
+    handle = 0;
+}
+
 void OGLTexture::Create(GLenum target) {
     if (handle != 0)
         return;
@@ -29,7 +45,6 @@ void OGLTexture::Release() {
 
     MICROPROFILE_SCOPE(OpenGL_ResourceDeletion);
     glDeleteTextures(1, &handle);
-    OpenGLState::GetCurState().UnbindTexture(handle).Apply();
     handle = 0;
 }
 
@@ -47,7 +62,6 @@ void OGLTextureView::Release() {
 
     MICROPROFILE_SCOPE(OpenGL_ResourceDeletion);
     glDeleteTextures(1, &handle);
-    OpenGLState::GetCurState().UnbindTexture(handle).Apply();
     handle = 0;
 }
 
@@ -65,7 +79,6 @@ void OGLSampler::Release() {
 
     MICROPROFILE_SCOPE(OpenGL_ResourceDeletion);
     glDeleteSamplers(1, &handle);
-    OpenGLState::GetCurState().ResetSampler(handle).Apply();
     handle = 0;
 }
 
@@ -109,7 +122,6 @@ void OGLProgram::Release() {
 
     MICROPROFILE_SCOPE(OpenGL_ResourceDeletion);
     glDeleteProgram(handle);
-    OpenGLState::GetCurState().ResetProgram(handle).Apply();
     handle = 0;
 }
 
@@ -127,7 +139,6 @@ void OGLPipeline::Release() {
 
     MICROPROFILE_SCOPE(OpenGL_ResourceDeletion);
     glDeleteProgramPipelines(1, &handle);
-    OpenGLState::GetCurState().ResetPipeline(handle).Apply();
     handle = 0;
 }
 
@@ -171,24 +182,6 @@ void OGLSync::Release() {
     handle = 0;
 }
 
-void OGLVertexArray::Create() {
-    if (handle != 0)
-        return;
-
-    MICROPROFILE_SCOPE(OpenGL_ResourceCreation);
-    glCreateVertexArrays(1, &handle);
-}
-
-void OGLVertexArray::Release() {
-    if (handle == 0)
-        return;
-
-    MICROPROFILE_SCOPE(OpenGL_ResourceDeletion);
-    glDeleteVertexArrays(1, &handle);
-    OpenGLState::GetCurState().ResetVertexArray(handle).Apply();
-    handle = 0;
-}
-
 void OGLFramebuffer::Create() {
     if (handle != 0)
         return;
@@ -203,7 +196,6 @@ void OGLFramebuffer::Release() {
 
     MICROPROFILE_SCOPE(OpenGL_ResourceDeletion);
     glDeleteFramebuffers(1, &handle);
-    OpenGLState::GetCurState().ResetFramebuffer(handle).Apply();
     handle = 0;
 }
 

src/video_core/renderer_opengl/gl_resource_manager.h

@@ -11,6 +11,31 @@
 
 namespace OpenGL {
 
+class OGLRenderbuffer : private NonCopyable {
+public:
+    OGLRenderbuffer() = default;
+
+    OGLRenderbuffer(OGLRenderbuffer&& o) noexcept : handle(std::exchange(o.handle, 0)) {}
+
+    ~OGLRenderbuffer() {
+        Release();
+    }
+
+    OGLRenderbuffer& operator=(OGLRenderbuffer&& o) noexcept {
+        Release();
+        handle = std::exchange(o.handle, 0);
+        return *this;
+    }
+
+    /// Creates a new internal OpenGL resource and stores the handle
+    void Create();
+
+    /// Deletes the internal OpenGL resource
+    void Release();
+
+    GLuint handle = 0;
+};
+
 class OGLTexture : private NonCopyable {
 public:
     OGLTexture() = default;
@@ -216,31 +241,6 @@ public:
     GLsync handle = 0;
 };
 
-class OGLVertexArray : private NonCopyable {
-public:
-    OGLVertexArray() = default;
-
-    OGLVertexArray(OGLVertexArray&& o) noexcept : handle(std::exchange(o.handle, 0)) {}
-
-    ~OGLVertexArray() {
-        Release();
-    }
-
-    OGLVertexArray& operator=(OGLVertexArray&& o) noexcept {
-        Release();
-        handle = std::exchange(o.handle, 0);
-        return *this;
-    }
-
-    /// Creates a new internal OpenGL resource and stores the handle
-    void Create();
-
-    /// Deletes the internal OpenGL resource
-    void Release();
-
-    GLuint handle = 0;
-};
-
 class OGLFramebuffer : private NonCopyable {
 public:
     OGLFramebuffer() = default;

src/video_core/renderer_opengl/gl_sampler_cache.cpp

@@ -38,7 +38,7 @@ OGLSampler SamplerCacheOpenGL::CreateSampler(const Tegra::Texture::TSCEntry& tsc
         glSamplerParameterf(sampler_id, GL_TEXTURE_MAX_ANISOTROPY, tsc.GetMaxAnisotropy());
     } else if (GLAD_GL_EXT_texture_filter_anisotropic) {
         glSamplerParameterf(sampler_id, GL_TEXTURE_MAX_ANISOTROPY_EXT, tsc.GetMaxAnisotropy());
-    } else if (tsc.GetMaxAnisotropy() != 1) {
+    } else {
         LOG_WARNING(Render_OpenGL, "Anisotropy not supported by host GPU driver");
     }
 

src/video_core/renderer_opengl/gl_shader_cache.cpp

@@ -22,6 +22,7 @@
 #include "video_core/renderer_opengl/gl_shader_cache.h"
 #include "video_core/renderer_opengl/gl_shader_decompiler.h"
 #include "video_core/renderer_opengl/gl_shader_disk_cache.h"
+#include "video_core/renderer_opengl/gl_state_tracker.h"
 #include "video_core/renderer_opengl/utils.h"
 #include "video_core/shader/shader_ir.h"
 
@@ -623,7 +624,7 @@ bool ShaderCacheOpenGL::GenerateUnspecializedShaders(
 }
 
 Shader ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program) {
-    if (!system.GPU().Maxwell3D().dirty.shaders) {
+    if (!system.GPU().Maxwell3D().dirty.flags[Dirty::Shaders]) {
         return last_shaders[static_cast<std::size_t>(program)];
     }
 

src/video_core/renderer_opengl/gl_shader_decompiler.cpp

@@ -2547,7 +2547,10 @@ ShaderEntries GetEntries(const VideoCommon::Shader::ShaderIR& ir) {
     for (const auto& image : ir.GetImages()) {
         entries.images.emplace_back(image);
     }
-    entries.clip_distances = ir.GetClipDistances();
+    const auto clip_distances = ir.GetClipDistances();
+    for (std::size_t i = 0; i < std::size(clip_distances); ++i) {
+        entries.clip_distances = (clip_distances[i] ? 1U : 0U) << i;
+    }
     entries.shader_length = ir.GetLength();
     return entries;
 }

src/video_core/renderer_opengl/gl_shader_decompiler.h

@@ -74,7 +74,7 @@ struct ShaderEntries {
     std::vector<GlobalMemoryEntry> global_memory_entries;
     std::vector<SamplerEntry> samplers;
     std::vector<ImageEntry> images;
-    std::array<bool, Maxwell::NumClipDistances> clip_distances{};
+    u32 clip_distances{};
     std::size_t shader_length{};
 };
 

src/video_core/renderer_opengl/gl_shader_manager.cpp

@@ -2,45 +2,52 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <glad/glad.h>
+
 #include "common/common_types.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/renderer_opengl/gl_shader_manager.h"
 
 namespace OpenGL::GLShader {
 
-using Tegra::Engines::Maxwell3D;
-
-ProgramManager::ProgramManager() {
-    pipeline.Create();
-}
+ProgramManager::ProgramManager() = default;
 
 ProgramManager::~ProgramManager() = default;
 
-void ProgramManager::ApplyTo(OpenGLState& state) {
-    UpdatePipeline();
-    state.draw.shader_program = 0;
-    state.draw.program_pipeline = pipeline.handle;
+void ProgramManager::Create() {
+    graphics_pipeline.Create();
+    glBindProgramPipeline(graphics_pipeline.handle);
 }
 
-void ProgramManager::UpdatePipeline() {
+void ProgramManager::BindGraphicsPipeline() {
+    if (!is_graphics_bound) {
+        is_graphics_bound = true;
+        glUseProgram(0);
+    }
+
     // Avoid updating the pipeline when values have no changed
     if (old_state == current_state) {
         return;
     }
 
     // Workaround for AMD bug
-    constexpr GLenum all_used_stages{GL_VERTEX_SHADER_BIT | GL_GEOMETRY_SHADER_BIT |
-                                     GL_FRAGMENT_SHADER_BIT};
-    glUseProgramStages(pipeline.handle, all_used_stages, 0);
-
-    glUseProgramStages(pipeline.handle, GL_VERTEX_SHADER_BIT, current_state.vertex_shader);
-    glUseProgramStages(pipeline.handle, GL_GEOMETRY_SHADER_BIT, current_state.geometry_shader);
-    glUseProgramStages(pipeline.handle, GL_FRAGMENT_SHADER_BIT, current_state.fragment_shader);
+    static constexpr GLenum all_used_stages{GL_VERTEX_SHADER_BIT | GL_GEOMETRY_SHADER_BIT |
+                                            GL_FRAGMENT_SHADER_BIT};
+    const GLuint handle = graphics_pipeline.handle;
+    glUseProgramStages(handle, all_used_stages, 0);
+    glUseProgramStages(handle, GL_VERTEX_SHADER_BIT, current_state.vertex_shader);
+    glUseProgramStages(handle, GL_GEOMETRY_SHADER_BIT, current_state.geometry_shader);
+    glUseProgramStages(handle, GL_FRAGMENT_SHADER_BIT, current_state.fragment_shader);
 
     old_state = current_state;
 }
 
-void MaxwellUniformData::SetFromRegs(const Maxwell3D& maxwell) {
+void ProgramManager::BindComputeShader(GLuint program) {
+    is_graphics_bound = false;
+    glUseProgram(program);
+}
+
+void MaxwellUniformData::SetFromRegs(const Tegra::Engines::Maxwell3D& maxwell) {
     const auto& regs = maxwell.regs;
 
     // Y_NEGATE controls what value S2R returns for the Y_DIRECTION system value.

src/video_core/renderer_opengl/gl_shader_manager.h

@@ -9,7 +9,6 @@
 #include <glad/glad.h>
 
 #include "video_core/renderer_opengl/gl_resource_manager.h"
-#include "video_core/renderer_opengl/gl_state.h"
 #include "video_core/renderer_opengl/maxwell_to_gl.h"
 
 namespace OpenGL::GLShader {
@@ -32,49 +31,47 @@ public:
     explicit ProgramManager();
     ~ProgramManager();
 
-    void ApplyTo(OpenGLState& state);
+    void Create();
 
-    void UseProgrammableVertexShader(GLuint program) {
+    /// Updates the graphics pipeline and binds it.
+    void BindGraphicsPipeline();
+
+    /// Binds a compute shader.
+    void BindComputeShader(GLuint program);
+
+    void UseVertexShader(GLuint program) {
         current_state.vertex_shader = program;
     }
 
-    void UseProgrammableGeometryShader(GLuint program) {
+    void UseGeometryShader(GLuint program) {
         current_state.geometry_shader = program;
     }
 
-    void UseProgrammableFragmentShader(GLuint program) {
+    void UseFragmentShader(GLuint program) {
         current_state.fragment_shader = program;
     }
 
-    void UseTrivialGeometryShader() {
-        current_state.geometry_shader = 0;
-    }
-
-    void UseTrivialFragmentShader() {
-        current_state.fragment_shader = 0;
-    }
-
 private:
     struct PipelineState {
-        bool operator==(const PipelineState& rhs) const {
+        bool operator==(const PipelineState& rhs) const noexcept {
             return vertex_shader == rhs.vertex_shader && fragment_shader == rhs.fragment_shader &&
                    geometry_shader == rhs.geometry_shader;
         }
 
-        bool operator!=(const PipelineState& rhs) const {
+        bool operator!=(const PipelineState& rhs) const noexcept {
             return !operator==(rhs);
         }
 
-        GLuint vertex_shader{};
-        GLuint fragment_shader{};
-        GLuint geometry_shader{};
+        GLuint vertex_shader = 0;
+        GLuint fragment_shader = 0;
+        GLuint geometry_shader = 0;
     };
 
-    void UpdatePipeline();
-
-    OGLPipeline pipeline;
+    OGLPipeline graphics_pipeline;
+    OGLPipeline compute_pipeline;
     PipelineState current_state;
     PipelineState old_state;
+    bool is_graphics_bound = true;
 };
 
 } // namespace OpenGL::GLShader

src/video_core/renderer_opengl/gl_state.cpp

@@ -1,554 +0,0 @@
-// Copyright 2015 Citra Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#include <algorithm>
-#include <iterator>
-#include <glad/glad.h>
-#include "common/assert.h"
-#include "common/logging/log.h"
-#include "common/microprofile.h"
-#include "video_core/renderer_opengl/gl_state.h"
-
-MICROPROFILE_DEFINE(OpenGL_State, "OpenGL", "State Change", MP_RGB(192, 128, 128));
-
-namespace OpenGL {
-
-using Maxwell = Tegra::Engines::Maxwell3D::Regs;
-
-OpenGLState OpenGLState::cur_state;
-
-namespace {
-
-template <typename T>
-bool UpdateValue(T& current_value, const T new_value) {
-    const bool changed = current_value != new_value;
-    current_value = new_value;
-    return changed;
-}
-
-template <typename T1, typename T2>
-bool UpdateTie(T1 current_value, const T2 new_value) {
-    const bool changed = current_value != new_value;
-    current_value = new_value;
-    return changed;
-}
-
-template <typename T>
-std::optional<std::pair<GLuint, GLsizei>> UpdateArray(T& current_values, const T& new_values) {
-    std::optional<std::size_t> first;
-    std::size_t last;
-    for (std::size_t i = 0; i < std::size(current_values); ++i) {
-        if (!UpdateValue(current_values[i], new_values[i])) {
-            continue;
-        }
-        if (!first) {
-            first = i;
-        }
-        last = i;
-    }
-    if (!first) {
-        return std::nullopt;
-    }
-    return std::make_pair(static_cast<GLuint>(*first), static_cast<GLsizei>(last - *first + 1));
-}
-
-void Enable(GLenum cap, bool enable) {
-    if (enable) {
-        glEnable(cap);
-    } else {
-        glDisable(cap);
-    }
-}
-
-void Enable(GLenum cap, GLuint index, bool enable) {
-    if (enable) {
-        glEnablei(cap, index);
-    } else {
-        glDisablei(cap, index);
-    }
-}
-
-void Enable(GLenum cap, bool& current_value, bool new_value) {
-    if (UpdateValue(current_value, new_value)) {
-        Enable(cap, new_value);
-    }
-}
-
-void Enable(GLenum cap, GLuint index, bool& current_value, bool new_value) {
-    if (UpdateValue(current_value, new_value)) {
-        Enable(cap, index, new_value);
-    }
-}
-
-} // Anonymous namespace
-
-OpenGLState::OpenGLState() = default;
-
-void OpenGLState::SetDefaultViewports() {
-    viewports.fill(Viewport{});
-
-    depth_clamp.far_plane = false;
-    depth_clamp.near_plane = false;
-}
-
-void OpenGLState::ApplyFramebufferState() {
-    if (UpdateValue(cur_state.draw.read_framebuffer, draw.read_framebuffer)) {
-        glBindFramebuffer(GL_READ_FRAMEBUFFER, draw.read_framebuffer);
-    }
-    if (UpdateValue(cur_state.draw.draw_framebuffer, draw.draw_framebuffer)) {
-        glBindFramebuffer(GL_DRAW_FRAMEBUFFER, draw.draw_framebuffer);
-    }
-}
-
-void OpenGLState::ApplyVertexArrayState() {
-    if (UpdateValue(cur_state.draw.vertex_array, draw.vertex_array)) {
-        glBindVertexArray(draw.vertex_array);
-    }
-}
-
-void OpenGLState::ApplyShaderProgram() {
-    if (UpdateValue(cur_state.draw.shader_program, draw.shader_program)) {
-        glUseProgram(draw.shader_program);
-    }
-}
-
-void OpenGLState::ApplyProgramPipeline() {
-    if (UpdateValue(cur_state.draw.program_pipeline, draw.program_pipeline)) {
-        glBindProgramPipeline(draw.program_pipeline);
-    }
-}
-
-void OpenGLState::ApplyClipDistances() {
-    for (std::size_t i = 0; i < clip_distance.size(); ++i) {
-        Enable(GL_CLIP_DISTANCE0 + static_cast<GLenum>(i), cur_state.clip_distance[i],
-               clip_distance[i]);
-    }
-}
-
-void OpenGLState::ApplyPointSize() {
-    Enable(GL_PROGRAM_POINT_SIZE, cur_state.point.program_control, point.program_control);
-    Enable(GL_POINT_SPRITE, cur_state.point.sprite, point.sprite);
-    if (UpdateValue(cur_state.point.size, point.size)) {
-        glPointSize(point.size);
-    }
-}
-
-void OpenGLState::ApplyFragmentColorClamp() {
-    if (UpdateValue(cur_state.fragment_color_clamp.enabled, fragment_color_clamp.enabled)) {
-        glClampColor(GL_CLAMP_FRAGMENT_COLOR_ARB,
-                     fragment_color_clamp.enabled ? GL_TRUE : GL_FALSE);
-    }
-}
-
-void OpenGLState::ApplyMultisample() {
-    Enable(GL_SAMPLE_ALPHA_TO_COVERAGE, cur_state.multisample_control.alpha_to_coverage,
-           multisample_control.alpha_to_coverage);
-    Enable(GL_SAMPLE_ALPHA_TO_ONE, cur_state.multisample_control.alpha_to_one,
-           multisample_control.alpha_to_one);
-}
-
-void OpenGLState::ApplyDepthClamp() {
-    if (depth_clamp.far_plane == cur_state.depth_clamp.far_plane &&
-        depth_clamp.near_plane == cur_state.depth_clamp.near_plane) {
-        return;
-    }
-    cur_state.depth_clamp = depth_clamp;
-
-    UNIMPLEMENTED_IF_MSG(depth_clamp.far_plane != depth_clamp.near_plane,
-                         "Unimplemented Depth Clamp Separation!");
-
-    Enable(GL_DEPTH_CLAMP, depth_clamp.far_plane || depth_clamp.near_plane);
-}
-
-void OpenGLState::ApplySRgb() {
-    if (cur_state.framebuffer_srgb.enabled == framebuffer_srgb.enabled)
-        return;
-    cur_state.framebuffer_srgb.enabled = framebuffer_srgb.enabled;
-    if (framebuffer_srgb.enabled) {
-        glEnable(GL_FRAMEBUFFER_SRGB);
-    } else {
-        glDisable(GL_FRAMEBUFFER_SRGB);
-    }
-}
-
-void OpenGLState::ApplyCulling() {
-    Enable(GL_CULL_FACE, cur_state.cull.enabled, cull.enabled);
-
-    if (UpdateValue(cur_state.cull.mode, cull.mode)) {
-        glCullFace(cull.mode);
-    }
-
-    if (UpdateValue(cur_state.cull.front_face, cull.front_face)) {
-        glFrontFace(cull.front_face);
-    }
-}
-
-void OpenGLState::ApplyRasterizerDiscard() {
-    Enable(GL_RASTERIZER_DISCARD, cur_state.rasterizer_discard, rasterizer_discard);
-}
-
-void OpenGLState::ApplyColorMask() {
-    if (!dirty.color_mask) {
-        return;
-    }
-    dirty.color_mask = false;
-
-    for (std::size_t i = 0; i < Maxwell::NumRenderTargets; ++i) {
-        const auto& updated = color_mask[i];
-        auto& current = cur_state.color_mask[i];
-        if (updated.red_enabled != current.red_enabled ||
-            updated.green_enabled != current.green_enabled ||
-            updated.blue_enabled != current.blue_enabled ||
-            updated.alpha_enabled != current.alpha_enabled) {
-            current = updated;
-            glColorMaski(static_cast<GLuint>(i), updated.red_enabled, updated.green_enabled,
-                         updated.blue_enabled, updated.alpha_enabled);
-        }
-    }
-}
-
-void OpenGLState::ApplyDepth() {
-    Enable(GL_DEPTH_TEST, cur_state.depth.test_enabled, depth.test_enabled);
-
-    if (cur_state.depth.test_func != depth.test_func) {
-        cur_state.depth.test_func = depth.test_func;
-        glDepthFunc(depth.test_func);
-    }
-
-    if (cur_state.depth.write_mask != depth.write_mask) {
-        cur_state.depth.write_mask = depth.write_mask;
-        glDepthMask(depth.write_mask);
-    }
-}
-
-void OpenGLState::ApplyPrimitiveRestart() {
-    Enable(GL_PRIMITIVE_RESTART, cur_state.primitive_restart.enabled, primitive_restart.enabled);
-
-    if (cur_state.primitive_restart.index != primitive_restart.index) {
-        cur_state.primitive_restart.index = primitive_restart.index;
-        glPrimitiveRestartIndex(primitive_restart.index);
-    }
-}
-
-void OpenGLState::ApplyStencilTest() {
-    if (!dirty.stencil_state) {
-        return;
-    }
-    dirty.stencil_state = false;
-
-    Enable(GL_STENCIL_TEST, cur_state.stencil.test_enabled, stencil.test_enabled);
-
-    const auto ConfigStencil = [](GLenum face, const auto& config, auto& current) {
-        if (current.test_func != config.test_func || current.test_ref != config.test_ref ||
-            current.test_mask != config.test_mask) {
-            current.test_func = config.test_func;
-            current.test_ref = config.test_ref;
-            current.test_mask = config.test_mask;
-            glStencilFuncSeparate(face, config.test_func, config.test_ref, config.test_mask);
-        }
-        if (current.action_depth_fail != config.action_depth_fail ||
-            current.action_depth_pass != config.action_depth_pass ||
-            current.action_stencil_fail != config.action_stencil_fail) {
-            current.action_depth_fail = config.action_depth_fail;
-            current.action_depth_pass = config.action_depth_pass;
-            current.action_stencil_fail = config.action_stencil_fail;
-            glStencilOpSeparate(face, config.action_stencil_fail, config.action_depth_fail,
-                                config.action_depth_pass);
-        }
-        if (current.write_mask != config.write_mask) {
-            current.write_mask = config.write_mask;
-            glStencilMaskSeparate(face, config.write_mask);
-        }
-    };
-    ConfigStencil(GL_FRONT, stencil.front, cur_state.stencil.front);
-    ConfigStencil(GL_BACK, stencil.back, cur_state.stencil.back);
-}
-
-void OpenGLState::ApplyViewport() {
-    for (GLuint i = 0; i < static_cast<GLuint>(Maxwell::NumViewports); ++i) {
-        const auto& updated = viewports[i];
-        auto& current = cur_state.viewports[i];
-
-        if (current.x != updated.x || current.y != updated.y || current.width != updated.width ||
-            current.height != updated.height) {
-            current.x = updated.x;
-            current.y = updated.y;
-            current.width = updated.width;
-            current.height = updated.height;
-            glViewportIndexedf(i, static_cast<GLfloat>(updated.x), static_cast<GLfloat>(updated.y),
-                               static_cast<GLfloat>(updated.width),
-                               static_cast<GLfloat>(updated.height));
-        }
-        if (current.depth_range_near != updated.depth_range_near ||
-            current.depth_range_far != updated.depth_range_far) {
-            current.depth_range_near = updated.depth_range_near;
-            current.depth_range_far = updated.depth_range_far;
-            glDepthRangeIndexed(i, updated.depth_range_near, updated.depth_range_far);
-        }
-
-        Enable(GL_SCISSOR_TEST, i, current.scissor.enabled, updated.scissor.enabled);
-
-        if (current.scissor.x != updated.scissor.x || current.scissor.y != updated.scissor.y ||
-            current.scissor.width != updated.scissor.width ||
-            current.scissor.height != updated.scissor.height) {
-            current.scissor.x = updated.scissor.x;
-            current.scissor.y = updated.scissor.y;
-            current.scissor.width = updated.scissor.width;
-            current.scissor.height = updated.scissor.height;
-            glScissorIndexed(i, updated.scissor.x, updated.scissor.y, updated.scissor.width,
-                             updated.scissor.height);
-        }
-    }
-}
-
-void OpenGLState::ApplyGlobalBlending() {
-    const Blend& updated = blend[0];
-    Blend& current = cur_state.blend[0];
-
-    Enable(GL_BLEND, current.enabled, updated.enabled);
-
-    if (current.src_rgb_func != updated.src_rgb_func ||
-        current.dst_rgb_func != updated.dst_rgb_func || current.src_a_func != updated.src_a_func ||
-        current.dst_a_func != updated.dst_a_func) {
-        current.src_rgb_func = updated.src_rgb_func;
-        current.dst_rgb_func = updated.dst_rgb_func;
-        current.src_a_func = updated.src_a_func;
-        current.dst_a_func = updated.dst_a_func;
-        glBlendFuncSeparate(updated.src_rgb_func, updated.dst_rgb_func, updated.src_a_func,
-                            updated.dst_a_func);
-    }
-
-    if (current.rgb_equation != updated.rgb_equation || current.a_equation != updated.a_equation) {
-        current.rgb_equation = updated.rgb_equation;
-        current.a_equation = updated.a_equation;
-        glBlendEquationSeparate(updated.rgb_equation, updated.a_equation);
-    }
-}
-
-void OpenGLState::ApplyTargetBlending(std::size_t target, bool force) {
-    const Blend& updated = blend[target];
-    Blend& current = cur_state.blend[target];
-
-    if (current.enabled != updated.enabled || force) {
-        current.enabled = updated.enabled;
-        Enable(GL_BLEND, static_cast<GLuint>(target), updated.enabled);
-    }
-
-    if (UpdateTie(std::tie(current.src_rgb_func, current.dst_rgb_func, current.src_a_func,
-                           current.dst_a_func),
-                  std::tie(updated.src_rgb_func, updated.dst_rgb_func, updated.src_a_func,
-                           updated.dst_a_func))) {
-        glBlendFuncSeparatei(static_cast<GLuint>(target), updated.src_rgb_func,
-                             updated.dst_rgb_func, updated.src_a_func, updated.dst_a_func);
-    }
-
-    if (UpdateTie(std::tie(current.rgb_equation, current.a_equation),
-                  std::tie(updated.rgb_equation, updated.a_equation))) {
-        glBlendEquationSeparatei(static_cast<GLuint>(target), updated.rgb_equation,
-                                 updated.a_equation);
-    }
-}
-
-void OpenGLState::ApplyBlending() {
-    if (!dirty.blend_state) {
-        return;
-    }
-    dirty.blend_state = false;
-
-    if (independant_blend.enabled) {
-        const bool force = independant_blend.enabled != cur_state.independant_blend.enabled;
-        for (std::size_t target = 0; target < Maxwell::NumRenderTargets; ++target) {
-            ApplyTargetBlending(target, force);
-        }
-    } else {
-        ApplyGlobalBlending();
-    }
-    cur_state.independant_blend.enabled = independant_blend.enabled;
-
-    if (UpdateTie(
-            std::tie(cur_state.blend_color.red, cur_state.blend_color.green,
-                     cur_state.blend_color.blue, cur_state.blend_color.alpha),
-            std::tie(blend_color.red, blend_color.green, blend_color.blue, blend_color.alpha))) {
-        glBlendColor(blend_color.red, blend_color.green, blend_color.blue, blend_color.alpha);
-    }
-}
-
-void OpenGLState::ApplyLogicOp() {
-    Enable(GL_COLOR_LOGIC_OP, cur_state.logic_op.enabled, logic_op.enabled);
-
-    if (UpdateValue(cur_state.logic_op.operation, logic_op.operation)) {
-        glLogicOp(logic_op.operation);
-    }
-}
-
-void OpenGLState::ApplyPolygonOffset() {
-    if (!dirty.polygon_offset) {
-        return;
-    }
-    dirty.polygon_offset = false;
-
-    Enable(GL_POLYGON_OFFSET_FILL, cur_state.polygon_offset.fill_enable,
-           polygon_offset.fill_enable);
-    Enable(GL_POLYGON_OFFSET_LINE, cur_state.polygon_offset.line_enable,
-           polygon_offset.line_enable);
-    Enable(GL_POLYGON_OFFSET_POINT, cur_state.polygon_offset.point_enable,
-           polygon_offset.point_enable);
-
-    if (UpdateTie(std::tie(cur_state.polygon_offset.factor, cur_state.polygon_offset.units,
-                           cur_state.polygon_offset.clamp),
-                  std::tie(polygon_offset.factor, polygon_offset.units, polygon_offset.clamp))) {
-        if (GLAD_GL_EXT_polygon_offset_clamp && polygon_offset.clamp != 0) {
-            glPolygonOffsetClamp(polygon_offset.factor, polygon_offset.units, polygon_offset.clamp);
-        } else {
-            UNIMPLEMENTED_IF_MSG(polygon_offset.clamp != 0,
-                                 "Unimplemented Depth polygon offset clamp.");
-            glPolygonOffset(polygon_offset.factor, polygon_offset.units);
-        }
-    }
-}
-
-void OpenGLState::ApplyAlphaTest() {
-    Enable(GL_ALPHA_TEST, cur_state.alpha_test.enabled, alpha_test.enabled);
-    if (UpdateTie(std::tie(cur_state.alpha_test.func, cur_state.alpha_test.ref),
-                  std::tie(alpha_test.func, alpha_test.ref))) {
-        glAlphaFunc(alpha_test.func, alpha_test.ref);
-    }
-}
-
-void OpenGLState::ApplyClipControl() {
-    if (UpdateTie(std::tie(cur_state.clip_control.origin, cur_state.clip_control.depth_mode),
-                  std::tie(clip_control.origin, clip_control.depth_mode))) {
-        glClipControl(clip_control.origin, clip_control.depth_mode);
-    }
-}
-
-void OpenGLState::ApplyTextures() {
-    const std::size_t size = std::size(textures);
-    for (std::size_t i = 0; i < size; ++i) {
-        if (UpdateValue(cur_state.textures[i], textures[i])) {
-            // BindTextureUnit doesn't support binding null textures, skip those binds.
-            // TODO(Rodrigo): Stop using null textures
-            if (textures[i] != 0) {
-                glBindTextureUnit(static_cast<GLuint>(i), textures[i]);
-            }
-        }
-    }
-}
-
-void OpenGLState::ApplySamplers() {
-    const std::size_t size = std::size(samplers);
-    for (std::size_t i = 0; i < size; ++i) {
-        if (UpdateValue(cur_state.samplers[i], samplers[i])) {
-            glBindSampler(static_cast<GLuint>(i), samplers[i]);
-        }
-    }
-}
-
-void OpenGLState::ApplyImages() {
-    if (const auto update = UpdateArray(cur_state.images, images)) {
-        glBindImageTextures(update->first, update->second, images.data() + update->first);
-    }
-}
-
-void OpenGLState::Apply() {
-    MICROPROFILE_SCOPE(OpenGL_State);
-    ApplyFramebufferState();
-    ApplyVertexArrayState();
-    ApplyShaderProgram();
-    ApplyProgramPipeline();
-    ApplyClipDistances();
-    ApplyPointSize();
-    ApplyFragmentColorClamp();
-    ApplyMultisample();
-    ApplyRasterizerDiscard();
-    ApplyColorMask();
-    ApplyDepthClamp();
-    ApplyViewport();
-    ApplyStencilTest();
-    ApplySRgb();
-    ApplyCulling();
-    ApplyDepth();
-    ApplyPrimitiveRestart();
-    ApplyBlending();
-    ApplyLogicOp();
-    ApplyTextures();
-    ApplySamplers();
-    ApplyImages();
-    ApplyPolygonOffset();
-    ApplyAlphaTest();
-    ApplyClipControl();
-}
-
-void OpenGLState::EmulateViewportWithScissor() {
-    auto& current = viewports[0];
-    if (current.scissor.enabled) {
-        const GLint left = std::max(current.x, current.scissor.x);
-        const GLint right =
-            std::max(current.x + current.width, current.scissor.x + current.scissor.width);
-        const GLint bottom = std::max(current.y, current.scissor.y);
-        const GLint top =
-            std::max(current.y + current.height, current.scissor.y + current.scissor.height);
-        current.scissor.x = std::max(left, 0);
-        current.scissor.y = std::max(bottom, 0);
-        current.scissor.width = std::max(right - left, 0);
-        current.scissor.height = std::max(top - bottom, 0);
-    } else {
-        current.scissor.enabled = true;
-        current.scissor.x = current.x;
-        current.scissor.y = current.y;
-        current.scissor.width = current.width;
-        current.scissor.height = current.height;
-    }
-}
-
-OpenGLState& OpenGLState::UnbindTexture(GLuint handle) {
-    for (auto& texture : textures) {
-        if (texture == handle) {
-            texture = 0;
-        }
-    }
-    return *this;
-}
-
-OpenGLState& OpenGLState::ResetSampler(GLuint handle) {
-    for (auto& sampler : samplers) {
-        if (sampler == handle) {
-            sampler = 0;
-        }
-    }
-    return *this;
-}
-
-OpenGLState& OpenGLState::ResetProgram(GLuint handle) {
-    if (draw.shader_program == handle) {
-        draw.shader_program = 0;
-    }
-    return *this;
-}
-
-OpenGLState& OpenGLState::ResetPipeline(GLuint handle) {
-    if (draw.program_pipeline == handle) {
-        draw.program_pipeline = 0;
-    }
-    return *this;
-}
-
-OpenGLState& OpenGLState::ResetVertexArray(GLuint handle) {
-    if (draw.vertex_array == handle) {
-        draw.vertex_array = 0;
-    }
-    return *this;
-}
-
-OpenGLState& OpenGLState::ResetFramebuffer(GLuint handle) {
-    if (draw.read_framebuffer == handle) {
-        draw.read_framebuffer = 0;
-    }
-    if (draw.draw_framebuffer == handle) {
-        draw.draw_framebuffer = 0;
-    }
-    return *this;
-}
-
-} // namespace OpenGL

src/video_core/renderer_opengl/gl_state.h

@@ -1,247 +0,0 @@
-// Copyright 2015 Citra Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#pragma once
-
-#include <array>
-#include <type_traits>
-#include <glad/glad.h>
-#include "video_core/engines/maxwell_3d.h"
-
-namespace OpenGL {
-
-class OpenGLState {
-public:
-    struct {
-        bool enabled = false; // GL_FRAMEBUFFER_SRGB
-    } framebuffer_srgb;
-
-    struct {
-        bool alpha_to_coverage = false; // GL_ALPHA_TO_COVERAGE
-        bool alpha_to_one = false;      // GL_ALPHA_TO_ONE
-    } multisample_control;
-
-    struct {
-        bool enabled = false; // GL_CLAMP_FRAGMENT_COLOR_ARB
-    } fragment_color_clamp;
-
-    struct {
-        bool far_plane = false;
-        bool near_plane = false;
-    } depth_clamp; // GL_DEPTH_CLAMP
-
-    struct {
-        bool enabled = false;       // GL_CULL_FACE
-        GLenum mode = GL_BACK;      // GL_CULL_FACE_MODE
-        GLenum front_face = GL_CCW; // GL_FRONT_FACE
-    } cull;
-
-    struct {
-        bool test_enabled = false;      // GL_DEPTH_TEST
-        GLboolean write_mask = GL_TRUE; // GL_DEPTH_WRITEMASK
-        GLenum test_func = GL_LESS;     // GL_DEPTH_FUNC
-    } depth;
-
-    struct {
-        bool enabled = false;
-        GLuint index = 0;
-    } primitive_restart; // GL_PRIMITIVE_RESTART
-
-    bool rasterizer_discard = false; // GL_RASTERIZER_DISCARD
-
-    struct ColorMask {
-        GLboolean red_enabled = GL_TRUE;
-        GLboolean green_enabled = GL_TRUE;
-        GLboolean blue_enabled = GL_TRUE;
-        GLboolean alpha_enabled = GL_TRUE;
-    };
-    std::array<ColorMask, Tegra::Engines::Maxwell3D::Regs::NumRenderTargets>
-        color_mask; // GL_COLOR_WRITEMASK
-
-    struct {
-        bool test_enabled = false; // GL_STENCIL_TEST
-        struct {
-            GLenum test_func = GL_ALWAYS;         // GL_STENCIL_FUNC
-            GLint test_ref = 0;                   // GL_STENCIL_REF
-            GLuint test_mask = 0xFFFFFFFF;        // GL_STENCIL_VALUE_MASK
-            GLuint write_mask = 0xFFFFFFFF;       // GL_STENCIL_WRITEMASK
-            GLenum action_stencil_fail = GL_KEEP; // GL_STENCIL_FAIL
-            GLenum action_depth_fail = GL_KEEP;   // GL_STENCIL_PASS_DEPTH_FAIL
-            GLenum action_depth_pass = GL_KEEP;   // GL_STENCIL_PASS_DEPTH_PASS
-        } front, back;
-    } stencil;
-
-    struct Blend {
-        bool enabled = false;              // GL_BLEND
-        GLenum rgb_equation = GL_FUNC_ADD; // GL_BLEND_EQUATION_RGB
-        GLenum a_equation = GL_FUNC_ADD;   // GL_BLEND_EQUATION_ALPHA
-        GLenum src_rgb_func = GL_ONE;      // GL_BLEND_SRC_RGB
-        GLenum dst_rgb_func = GL_ZERO;     // GL_BLEND_DST_RGB
-        GLenum src_a_func = GL_ONE;        // GL_BLEND_SRC_ALPHA
-        GLenum dst_a_func = GL_ZERO;       // GL_BLEND_DST_ALPHA
-    };
-    std::array<Blend, Tegra::Engines::Maxwell3D::Regs::NumRenderTargets> blend;
-
-    struct {
-        bool enabled = false;
-    } independant_blend;
-
-    struct {
-        GLclampf red = 0.0f;
-        GLclampf green = 0.0f;
-        GLclampf blue = 0.0f;
-        GLclampf alpha = 0.0f;
-    } blend_color; // GL_BLEND_COLOR
-
-    struct {
-        bool enabled = false; // GL_LOGIC_OP_MODE
-        GLenum operation = GL_COPY;
-    } logic_op;
-
-    static constexpr std::size_t NumSamplers = 32 * 5;
-    static constexpr std::size_t NumImages = 8 * 5;
-    std::array<GLuint, NumSamplers> textures = {};
-    std::array<GLuint, NumSamplers> samplers = {};
-    std::array<GLuint, NumImages> images = {};
-
-    struct {
-        GLuint read_framebuffer = 0; // GL_READ_FRAMEBUFFER_BINDING
-        GLuint draw_framebuffer = 0; // GL_DRAW_FRAMEBUFFER_BINDING
-        GLuint vertex_array = 0;     // GL_VERTEX_ARRAY_BINDING
-        GLuint shader_program = 0;   // GL_CURRENT_PROGRAM
-        GLuint program_pipeline = 0; // GL_PROGRAM_PIPELINE_BINDING
-    } draw;
-
-    struct Viewport {
-        GLint x = 0;
-        GLint y = 0;
-        GLint width = 0;
-        GLint height = 0;
-        GLfloat depth_range_near = 0.0f; // GL_DEPTH_RANGE
-        GLfloat depth_range_far = 1.0f;  // GL_DEPTH_RANGE
-        struct {
-            bool enabled = false; // GL_SCISSOR_TEST
-            GLint x = 0;
-            GLint y = 0;
-            GLsizei width = 0;
-            GLsizei height = 0;
-        } scissor;
-    };
-    std::array<Viewport, Tegra::Engines::Maxwell3D::Regs::NumViewports> viewports;
-
-    struct {
-        bool program_control = false; // GL_PROGRAM_POINT_SIZE
-        bool sprite = false;          // GL_POINT_SPRITE
-        GLfloat size = 1.0f;          // GL_POINT_SIZE
-    } point;
-
-    struct {
-        bool point_enable = false;
-        bool line_enable = false;
-        bool fill_enable = false;
-        GLfloat units = 0.0f;
-        GLfloat factor = 0.0f;
-        GLfloat clamp = 0.0f;
-    } polygon_offset;
-
-    struct {
-        bool enabled = false;    // GL_ALPHA_TEST
-        GLenum func = GL_ALWAYS; // GL_ALPHA_TEST_FUNC
-        GLfloat ref = 0.0f;      // GL_ALPHA_TEST_REF
-    } alpha_test;
-
-    std::array<bool, 8> clip_distance = {}; // GL_CLIP_DISTANCE
-
-    struct {
-        GLenum origin = GL_LOWER_LEFT;
-        GLenum depth_mode = GL_NEGATIVE_ONE_TO_ONE;
-    } clip_control;
-
-    OpenGLState();
-
-    /// Get the currently active OpenGL state
-    static OpenGLState GetCurState() {
-        return cur_state;
-    }
-
-    void SetDefaultViewports();
-    /// Apply this state as the current OpenGL state
-    void Apply();
-
-    void ApplyFramebufferState();
-    void ApplyVertexArrayState();
-    void ApplyShaderProgram();
-    void ApplyProgramPipeline();
-    void ApplyClipDistances();
-    void ApplyPointSize();
-    void ApplyFragmentColorClamp();
-    void ApplyMultisample();
-    void ApplySRgb();
-    void ApplyCulling();
-    void ApplyRasterizerDiscard();
-    void ApplyColorMask();
-    void ApplyDepth();
-    void ApplyPrimitiveRestart();
-    void ApplyStencilTest();
-    void ApplyViewport();
-    void ApplyTargetBlending(std::size_t target, bool force);
-    void ApplyGlobalBlending();
-    void ApplyBlending();
-    void ApplyLogicOp();
-    void ApplyTextures();
-    void ApplySamplers();
-    void ApplyImages();
-    void ApplyDepthClamp();
-    void ApplyPolygonOffset();
-    void ApplyAlphaTest();
-    void ApplyClipControl();
-
-    /// Resets any references to the given resource
-    OpenGLState& UnbindTexture(GLuint handle);
-    OpenGLState& ResetSampler(GLuint handle);
-    OpenGLState& ResetProgram(GLuint handle);
-    OpenGLState& ResetPipeline(GLuint handle);
-    OpenGLState& ResetVertexArray(GLuint handle);
-    OpenGLState& ResetFramebuffer(GLuint handle);
-
-    /// Viewport does not affects glClearBuffer so emulate viewport using scissor test
-    void EmulateViewportWithScissor();
-
-    void MarkDirtyBlendState() {
-        dirty.blend_state = true;
-    }
-
-    void MarkDirtyStencilState() {
-        dirty.stencil_state = true;
-    }
-
-    void MarkDirtyPolygonOffset() {
-        dirty.polygon_offset = true;
-    }
-
-    void MarkDirtyColorMask() {
-        dirty.color_mask = true;
-    }
-
-    void AllDirty() {
-        dirty.blend_state = true;
-        dirty.stencil_state = true;
-        dirty.polygon_offset = true;
-        dirty.color_mask = true;
-    }
-
-private:
-    static OpenGLState cur_state;
-
-    struct {
-        bool blend_state;
-        bool stencil_state;
-        bool viewport_state;
-        bool polygon_offset;
-        bool color_mask;
-    } dirty{};
-};
-static_assert(std::is_trivially_copyable_v<OpenGLState>);
-
-} // namespace OpenGL

src/video_core/renderer_opengl/gl_state_tracker.cpp

@@ -0,0 +1,238 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include <array>
+#include <cstddef>
+
+#include "common/common_types.h"
+#include "core/core.h"
+#include "video_core/engines/maxwell_3d.h"
+#include "video_core/gpu.h"
+#include "video_core/renderer_opengl/gl_state_tracker.h"
+
+#define OFF(field_name) MAXWELL3D_REG_INDEX(field_name)
+#define NUM(field_name) (sizeof(Maxwell3D::Regs::field_name) / sizeof(u32))
+
+namespace OpenGL {
+
+namespace {
+
+using namespace Dirty;
+using namespace VideoCommon::Dirty;
+using Tegra::Engines::Maxwell3D;
+using Regs = Maxwell3D::Regs;
+using Tables = Maxwell3D::DirtyState::Tables;
+using Table = Maxwell3D::DirtyState::Table;
+
+void SetupDirtyColorMasks(Tables& tables) {
+    tables[0][OFF(color_mask_common)] = ColorMaskCommon;
+    for (std::size_t rt = 0; rt < Regs::NumRenderTargets; ++rt) {
+        const std::size_t offset = OFF(color_mask) + rt * NUM(color_mask[0]);
+        FillBlock(tables[0], offset, NUM(color_mask[0]), ColorMask0 + rt);
+    }
+
+    FillBlock(tables[1], OFF(color_mask), NUM(color_mask), ColorMasks);
+}
+
+void SetupDirtyVertexArrays(Tables& tables) {
+    static constexpr std::size_t num_array = 3;
+    static constexpr std::size_t instance_base_offset = 3;
+    for (std::size_t i = 0; i < 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);
+
+        const std::size_t instance_array_offset = array_offset + instance_base_offset;
+        tables[0][instance_array_offset] = static_cast<u8>(VertexInstance0 + i);
+        tables[1][instance_array_offset] = VertexInstances;
+
+        const std::size_t instance_offset = OFF(instanced_arrays) + i;
+        tables[0][instance_offset] = static_cast<u8>(VertexInstance0 + i);
+        tables[1][instance_offset] = VertexInstances;
+    }
+}
+
+void SetupDirtyVertexFormat(Tables& tables) {
+    for (std::size_t i = 0; i < Regs::NumVertexAttributes; ++i) {
+        const std::size_t offset = OFF(vertex_attrib_format) + i * NUM(vertex_attrib_format[0]);
+        FillBlock(tables[0], offset, NUM(vertex_attrib_format[0]), VertexFormat0 + i);
+    }
+
+    FillBlock(tables[1], OFF(vertex_attrib_format), Regs::NumVertexAttributes, VertexFormats);
+}
+
+void SetupDirtyViewports(Tables& tables) {
+    for (std::size_t i = 0; i < Regs::NumViewports; ++i) {
+        const std::size_t transf_offset = OFF(viewport_transform) + i * NUM(viewport_transform[0]);
+        const std::size_t viewport_offset = OFF(viewports) + i * NUM(viewports[0]);
+
+        FillBlock(tables[0], transf_offset, NUM(viewport_transform[0]), Viewport0 + i);
+        FillBlock(tables[0], viewport_offset, NUM(viewports[0]), Viewport0 + i);
+    }
+
+    FillBlock(tables[1], OFF(viewport_transform), NUM(viewport_transform), Viewports);
+    FillBlock(tables[1], OFF(viewports), NUM(viewports), Viewports);
+
+    tables[0][OFF(viewport_transform_enabled)] = ViewportTransform;
+    tables[1][OFF(viewport_transform_enabled)] = Viewports;
+}
+
+void SetupDirtyScissors(Tables& tables) {
+    for (std::size_t i = 0; i < Regs::NumViewports; ++i) {
+        const std::size_t offset = OFF(scissor_test) + i * NUM(scissor_test[0]);
+        FillBlock(tables[0], offset, NUM(scissor_test[0]), Scissor0 + i);
+    }
+    FillBlock(tables[1], OFF(scissor_test), NUM(scissor_test), Scissors);
+}
+
+void SetupDirtyShaders(Tables& tables) {
+    FillBlock(tables[0], OFF(shader_config[0]), NUM(shader_config[0]) * Regs::MaxShaderProgram,
+              Shaders);
+}
+
+void SetupDirtyDepthTest(Tables& tables) {
+    auto& table = tables[0];
+    table[OFF(depth_test_enable)] = DepthTest;
+    table[OFF(depth_write_enabled)] = DepthMask;
+    table[OFF(depth_test_func)] = DepthTest;
+}
+
+void SetupDirtyStencilTest(Tables& tables) {
+    static constexpr std::array offsets = {
+        OFF(stencil_enable),          OFF(stencil_front_func_func), OFF(stencil_front_func_ref),
+        OFF(stencil_front_func_mask), OFF(stencil_front_op_fail),   OFF(stencil_front_op_zfail),
+        OFF(stencil_front_op_zpass),  OFF(stencil_front_mask),      OFF(stencil_two_side_enable),
+        OFF(stencil_back_func_func),  OFF(stencil_back_func_ref),   OFF(stencil_back_func_mask),
+        OFF(stencil_back_op_fail),    OFF(stencil_back_op_zfail),   OFF(stencil_back_op_zpass),
+        OFF(stencil_back_mask)};
+    for (const auto offset : offsets) {
+        tables[0][offset] = StencilTest;
+    }
+}
+
+void SetupDirtyAlphaTest(Tables& tables) {
+    auto& table = tables[0];
+    table[OFF(alpha_test_ref)] = AlphaTest;
+    table[OFF(alpha_test_func)] = AlphaTest;
+    table[OFF(alpha_test_enabled)] = AlphaTest;
+}
+
+void SetupDirtyBlend(Tables& tables) {
+    FillBlock(tables[0], OFF(blend_color), NUM(blend_color), BlendColor);
+
+    tables[0][OFF(independent_blend_enable)] = BlendIndependentEnabled;
+
+    for (std::size_t i = 0; i < Regs::NumRenderTargets; ++i) {
+        const std::size_t offset = OFF(independent_blend) + i * NUM(independent_blend[0]);
+        FillBlock(tables[0], offset, NUM(independent_blend[0]), BlendState0 + i);
+
+        tables[0][OFF(blend.enable) + i] = static_cast<u8>(BlendState0 + i);
+    }
+    FillBlock(tables[1], OFF(independent_blend), NUM(independent_blend), BlendStates);
+    FillBlock(tables[1], OFF(blend), NUM(blend), BlendStates);
+}
+
+void SetupDirtyPrimitiveRestart(Tables& tables) {
+    FillBlock(tables[0], OFF(primitive_restart), NUM(primitive_restart), PrimitiveRestart);
+}
+
+void SetupDirtyPolygonOffset(Tables& tables) {
+    auto& table = tables[0];
+    table[OFF(polygon_offset_fill_enable)] = PolygonOffset;
+    table[OFF(polygon_offset_line_enable)] = PolygonOffset;
+    table[OFF(polygon_offset_point_enable)] = PolygonOffset;
+    table[OFF(polygon_offset_factor)] = PolygonOffset;
+    table[OFF(polygon_offset_units)] = PolygonOffset;
+    table[OFF(polygon_offset_clamp)] = PolygonOffset;
+}
+
+void SetupDirtyMultisampleControl(Tables& tables) {
+    FillBlock(tables[0], OFF(multisample_control), NUM(multisample_control), MultisampleControl);
+}
+
+void SetupDirtyRasterizeEnable(Tables& tables) {
+    tables[0][OFF(rasterize_enable)] = RasterizeEnable;
+}
+
+void SetupDirtyFramebufferSRGB(Tables& tables) {
+    tables[0][OFF(framebuffer_srgb)] = FramebufferSRGB;
+}
+
+void SetupDirtyLogicOp(Tables& tables) {
+    FillBlock(tables[0], OFF(logic_op), NUM(logic_op), LogicOp);
+}
+
+void SetupDirtyFragmentClampColor(Tables& tables) {
+    tables[0][OFF(frag_color_clamp)] = FragmentClampColor;
+}
+
+void SetupDirtyPointSize(Tables& tables) {
+    tables[0][OFF(vp_point_size)] = PointSize;
+    tables[0][OFF(point_size)] = PointSize;
+    tables[0][OFF(point_sprite_enable)] = PointSize;
+}
+
+void SetupDirtyClipControl(Tables& tables) {
+    auto& table = tables[0];
+    table[OFF(screen_y_control)] = ClipControl;
+    table[OFF(depth_mode)] = ClipControl;
+}
+
+void SetupDirtyDepthClampEnabled(Tables& tables) {
+    tables[0][OFF(view_volume_clip_control)] = DepthClampEnabled;
+}
+
+void SetupDirtyMisc(Tables& tables) {
+    auto& table = tables[0];
+
+    table[OFF(clip_distance_enabled)] = ClipDistances;
+
+    table[OFF(front_face)] = FrontFace;
+
+    table[OFF(cull_test_enabled)] = CullTest;
+    table[OFF(cull_face)] = CullTest;
+}
+
+} // Anonymous namespace
+
+StateTracker::StateTracker(Core::System& system) : system{system} {}
+
+void StateTracker::Initialize() {
+    auto& dirty = system.GPU().Maxwell3D().dirty;
+    auto& tables = dirty.tables;
+    SetupDirtyRenderTargets(tables);
+    SetupDirtyColorMasks(tables);
+    SetupDirtyViewports(tables);
+    SetupDirtyScissors(tables);
+    SetupDirtyVertexArrays(tables);
+    SetupDirtyVertexFormat(tables);
+    SetupDirtyShaders(tables);
+    SetupDirtyDepthTest(tables);
+    SetupDirtyStencilTest(tables);
+    SetupDirtyAlphaTest(tables);
+    SetupDirtyBlend(tables);
+    SetupDirtyPrimitiveRestart(tables);
+    SetupDirtyPolygonOffset(tables);
+    SetupDirtyMultisampleControl(tables);
+    SetupDirtyRasterizeEnable(tables);
+    SetupDirtyFramebufferSRGB(tables);
+    SetupDirtyLogicOp(tables);
+    SetupDirtyFragmentClampColor(tables);
+    SetupDirtyPointSize(tables);
+    SetupDirtyClipControl(tables);
+    SetupDirtyDepthClampEnabled(tables);
+    SetupDirtyMisc(tables);
+
+    auto& store = dirty.on_write_stores;
+    SetupCommonOnWriteStores(store);
+    store[VertexBuffers] = true;
+    for (std::size_t i = 0; i < Regs::NumVertexArrays; ++i) {
+        store[VertexBuffer0 + i] = true;
+    }
+}
+
+} // namespace OpenGL

src/video_core/renderer_opengl/gl_state_tracker.h

@@ -0,0 +1,204 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <limits>
+
+#include <glad/glad.h>
+
+#include "common/common_types.h"
+#include "core/core.h"
+#include "video_core/dirty_flags.h"
+#include "video_core/engines/maxwell_3d.h"
+
+namespace Core {
+class System;
+}
+
+namespace OpenGL {
+
+namespace Dirty {
+
+enum : u8 {
+    First = VideoCommon::Dirty::LastCommonEntry,
+
+    VertexFormats,
+    VertexFormat0,
+    VertexFormat31 = VertexFormat0 + 31,
+
+    VertexBuffers,
+    VertexBuffer0,
+    VertexBuffer31 = VertexBuffer0 + 31,
+
+    VertexInstances,
+    VertexInstance0,
+    VertexInstance31 = VertexInstance0 + 31,
+
+    ViewportTransform,
+    Viewports,
+    Viewport0,
+    Viewport15 = Viewport0 + 15,
+
+    Scissors,
+    Scissor0,
+    Scissor15 = Scissor0 + 15,
+
+    ColorMaskCommon,
+    ColorMasks,
+    ColorMask0,
+    ColorMask7 = ColorMask0 + 7,
+
+    BlendColor,
+    BlendIndependentEnabled,
+    BlendStates,
+    BlendState0,
+    BlendState7 = BlendState0 + 7,
+
+    Shaders,
+    ClipDistances,
+
+    ColorMask,
+    FrontFace,
+    CullTest,
+    DepthMask,
+    DepthTest,
+    StencilTest,
+    AlphaTest,
+    PrimitiveRestart,
+    PolygonOffset,
+    MultisampleControl,
+    RasterizeEnable,
+    FramebufferSRGB,
+    LogicOp,
+    FragmentClampColor,
+    PointSize,
+    ClipControl,
+    DepthClampEnabled,
+
+    Last
+};
+static_assert(Last <= std::numeric_limits<u8>::max());
+
+} // namespace Dirty
+
+class StateTracker {
+public:
+    explicit StateTracker(Core::System& system);
+
+    void Initialize();
+
+    void BindIndexBuffer(GLuint new_index_buffer) {
+        if (index_buffer == new_index_buffer) {
+            return;
+        }
+        index_buffer = new_index_buffer;
+        glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, new_index_buffer);
+    }
+
+    void NotifyScreenDrawVertexArray() {
+        auto& flags = system.GPU().Maxwell3D().dirty.flags;
+        flags[OpenGL::Dirty::VertexFormats] = true;
+        flags[OpenGL::Dirty::VertexFormat0 + 0] = true;
+        flags[OpenGL::Dirty::VertexFormat0 + 1] = true;
+
+        flags[OpenGL::Dirty::VertexBuffers] = true;
+        flags[OpenGL::Dirty::VertexBuffer0] = true;
+
+        flags[OpenGL::Dirty::VertexInstances] = true;
+        flags[OpenGL::Dirty::VertexInstance0 + 0] = true;
+        flags[OpenGL::Dirty::VertexInstance0 + 1] = true;
+    }
+
+    void NotifyViewport0() {
+        auto& flags = system.GPU().Maxwell3D().dirty.flags;
+        flags[OpenGL::Dirty::Viewports] = true;
+        flags[OpenGL::Dirty::Viewport0] = true;
+    }
+
+    void NotifyScissor0() {
+        auto& flags = system.GPU().Maxwell3D().dirty.flags;
+        flags[OpenGL::Dirty::Scissors] = true;
+        flags[OpenGL::Dirty::Scissor0] = true;
+    }
+
+    void NotifyColorMask0() {
+        auto& flags = system.GPU().Maxwell3D().dirty.flags;
+        flags[OpenGL::Dirty::ColorMasks] = true;
+        flags[OpenGL::Dirty::ColorMask0] = true;
+    }
+
+    void NotifyBlend0() {
+        auto& flags = system.GPU().Maxwell3D().dirty.flags;
+        flags[OpenGL::Dirty::BlendStates] = true;
+        flags[OpenGL::Dirty::BlendState0] = true;
+    }
+
+    void NotifyFramebuffer() {
+        auto& flags = system.GPU().Maxwell3D().dirty.flags;
+        flags[VideoCommon::Dirty::RenderTargets] = true;
+    }
+
+    void NotifyFrontFace() {
+        auto& flags = system.GPU().Maxwell3D().dirty.flags;
+        flags[OpenGL::Dirty::FrontFace] = true;
+    }
+
+    void NotifyCullTest() {
+        auto& flags = system.GPU().Maxwell3D().dirty.flags;
+        flags[OpenGL::Dirty::CullTest] = true;
+    }
+
+    void NotifyDepthMask() {
+        auto& flags = system.GPU().Maxwell3D().dirty.flags;
+        flags[OpenGL::Dirty::DepthMask] = true;
+    }
+
+    void NotifyDepthTest() {
+        auto& flags = system.GPU().Maxwell3D().dirty.flags;
+        flags[OpenGL::Dirty::DepthTest] = true;
+    }
+
+    void NotifyStencilTest() {
+        auto& flags = system.GPU().Maxwell3D().dirty.flags;
+        flags[OpenGL::Dirty::StencilTest] = true;
+    }
+
+    void NotifyPolygonOffset() {
+        auto& flags = system.GPU().Maxwell3D().dirty.flags;
+        flags[OpenGL::Dirty::PolygonOffset] = true;
+    }
+
+    void NotifyRasterizeEnable() {
+        auto& flags = system.GPU().Maxwell3D().dirty.flags;
+        flags[OpenGL::Dirty::RasterizeEnable] = true;
+    }
+
+    void NotifyFramebufferSRGB() {
+        auto& flags = system.GPU().Maxwell3D().dirty.flags;
+        flags[OpenGL::Dirty::FramebufferSRGB] = true;
+    }
+
+    void NotifyLogicOp() {
+        auto& flags = system.GPU().Maxwell3D().dirty.flags;
+        flags[OpenGL::Dirty::LogicOp] = true;
+    }
+
+    void NotifyClipControl() {
+        auto& flags = system.GPU().Maxwell3D().dirty.flags;
+        flags[OpenGL::Dirty::ClipControl] = true;
+    }
+
+    void NotifyAlphaTest() {
+        auto& flags = system.GPU().Maxwell3D().dirty.flags;
+        flags[OpenGL::Dirty::AlphaTest] = true;
+    }
+
+private:
+    Core::System& system;
+
+    GLuint index_buffer = 0;
+};
+
+} // namespace OpenGL

src/video_core/renderer_opengl/gl_stream_buffer.cpp

@@ -7,7 +7,6 @@
 #include "common/alignment.h"
 #include "common/assert.h"
 #include "common/microprofile.h"
-#include "video_core/renderer_opengl/gl_state.h"
 #include "video_core/renderer_opengl/gl_stream_buffer.h"
 
 MICROPROFILE_DEFINE(OpenGL_StreamBuffer, "OpenGL", "Stream Buffer Orphaning",

src/video_core/renderer_opengl/gl_texture_cache.cpp

@@ -10,7 +10,7 @@
 #include "core/core.h"
 #include "video_core/morton.h"
 #include "video_core/renderer_opengl/gl_resource_manager.h"
-#include "video_core/renderer_opengl/gl_state.h"
+#include "video_core/renderer_opengl/gl_state_tracker.h"
 #include "video_core/renderer_opengl/gl_texture_cache.h"
 #include "video_core/renderer_opengl/utils.h"
 #include "video_core/texture_cache/surface_base.h"
@@ -397,6 +397,7 @@ CachedSurfaceView::CachedSurfaceView(CachedSurface& surface, const ViewParams& p
                                      const bool is_proxy)
     : VideoCommon::ViewBase(params), surface{surface}, is_proxy{is_proxy} {
     target = GetTextureTarget(params.target);
+    format = GetFormatTuple(surface.GetSurfaceParams().pixel_format).internal_format;
     if (!is_proxy) {
         texture_view = CreateTextureView();
     }
@@ -467,25 +468,20 @@ void CachedSurfaceView::ApplySwizzle(SwizzleSource x_source, SwizzleSource y_sou
 }
 
 OGLTextureView CachedSurfaceView::CreateTextureView() const {
-    const auto& owner_params = surface.GetSurfaceParams();
     OGLTextureView texture_view;
     texture_view.Create();
 
-    const GLuint handle{texture_view.handle};
-    const FormatTuple& tuple{GetFormatTuple(owner_params.pixel_format)};
-
-    glTextureView(handle, target, surface.texture.handle, tuple.internal_format, params.base_level,
+    glTextureView(texture_view.handle, target, surface.texture.handle, format, params.base_level,
                   params.num_levels, params.base_layer, params.num_layers);
-
-    ApplyTextureDefaults(owner_params, handle);
+    ApplyTextureDefaults(surface.GetSurfaceParams(), texture_view.handle);
 
     return texture_view;
 }
 
 TextureCacheOpenGL::TextureCacheOpenGL(Core::System& system,
                                        VideoCore::RasterizerInterface& rasterizer,
-                                       const Device& device)
-    : TextureCacheBase{system, rasterizer} {
+                                       const Device& device, StateTracker& state_tracker)
+    : TextureCacheBase{system, rasterizer}, state_tracker{state_tracker} {
     src_framebuffer.Create();
     dst_framebuffer.Create();
 }
@@ -519,25 +515,26 @@ void TextureCacheOpenGL::ImageBlit(View& src_view, View& dst_view,
                                    const Tegra::Engines::Fermi2D::Config& copy_config) {
     const auto& src_params{src_view->GetSurfaceParams()};
     const auto& dst_params{dst_view->GetSurfaceParams()};
-
-    OpenGLState prev_state{OpenGLState::GetCurState()};
-    SCOPE_EXIT({
-        prev_state.AllDirty();
-        prev_state.Apply();
-    });
-
-    OpenGLState state;
-    state.draw.read_framebuffer = src_framebuffer.handle;
-    state.draw.draw_framebuffer = dst_framebuffer.handle;
-    state.framebuffer_srgb.enabled = dst_params.srgb_conversion;
-    state.AllDirty();
-    state.Apply();
-
-    u32 buffers{};
-
     UNIMPLEMENTED_IF(src_params.target == SurfaceTarget::Texture3D);
     UNIMPLEMENTED_IF(dst_params.target == SurfaceTarget::Texture3D);
 
+    state_tracker.NotifyScissor0();
+    state_tracker.NotifyFramebuffer();
+    state_tracker.NotifyRasterizeEnable();
+    state_tracker.NotifyFramebufferSRGB();
+
+    if (dst_params.srgb_conversion) {
+        glEnable(GL_FRAMEBUFFER_SRGB);
+    } else {
+        glDisable(GL_FRAMEBUFFER_SRGB);
+    }
+    glDisable(GL_RASTERIZER_DISCARD);
+    glDisablei(GL_SCISSOR_TEST, 0);
+
+    glBindFramebuffer(GL_READ_FRAMEBUFFER, src_framebuffer.handle);
+    glBindFramebuffer(GL_DRAW_FRAMEBUFFER, dst_framebuffer.handle);
+
+    GLenum buffers = 0;
     if (src_params.type == SurfaceType::ColorTexture) {
         src_view->Attach(GL_COLOR_ATTACHMENT0, GL_READ_FRAMEBUFFER);
         glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0,

src/video_core/renderer_opengl/gl_texture_cache.h

@@ -27,6 +27,7 @@ using VideoCommon::ViewParams;
 class CachedSurfaceView;
 class CachedSurface;
 class TextureCacheOpenGL;
+class StateTracker;
 
 using Surface = std::shared_ptr<CachedSurface>;
 using View = std::shared_ptr<CachedSurfaceView>;
@@ -96,6 +97,10 @@ public:
         return texture_view.handle;
     }
 
+    GLenum GetFormat() const {
+        return format;
+    }
+
     const SurfaceParams& GetSurfaceParams() const {
         return surface.GetSurfaceParams();
     }
@@ -113,6 +118,7 @@ private:
 
     CachedSurface& surface;
     GLenum target{};
+    GLenum format{};
 
     OGLTextureView texture_view;
     u32 swizzle{};
@@ -122,7 +128,7 @@ private:
 class TextureCacheOpenGL final : public TextureCacheBase {
 public:
     explicit TextureCacheOpenGL(Core::System& system, VideoCore::RasterizerInterface& rasterizer,
-                                const Device& device);
+                                const Device& device, StateTracker& state_tracker);
     ~TextureCacheOpenGL();
 
 protected:
@@ -139,6 +145,8 @@ protected:
 private:
     GLuint FetchPBO(std::size_t buffer_size);
 
+    StateTracker& state_tracker;
+
     OGLFramebuffer src_framebuffer;
     OGLFramebuffer dst_framebuffer;
     std::unordered_map<u32, OGLBuffer> copy_pbo_cache;

src/video_core/renderer_opengl/maxwell_to_gl.h

@@ -92,8 +92,32 @@ inline GLenum VertexType(Maxwell::VertexAttribute attrib) {
         }
     case Maxwell::VertexAttribute::Type::UnsignedScaled:
         switch (attrib.size) {
+        case Maxwell::VertexAttribute::Size::Size_8:
         case Maxwell::VertexAttribute::Size::Size_8_8:
+        case Maxwell::VertexAttribute::Size::Size_8_8_8:
+        case Maxwell::VertexAttribute::Size::Size_8_8_8_8:
             return GL_UNSIGNED_BYTE;
+        case Maxwell::VertexAttribute::Size::Size_16:
+        case Maxwell::VertexAttribute::Size::Size_16_16:
+        case Maxwell::VertexAttribute::Size::Size_16_16_16:
+        case Maxwell::VertexAttribute::Size::Size_16_16_16_16:
+            return GL_UNSIGNED_SHORT;
+        default:
+            LOG_ERROR(Render_OpenGL, "Unimplemented vertex size={}", attrib.SizeString());
+            return {};
+        }
+    case Maxwell::VertexAttribute::Type::SignedScaled:
+        switch (attrib.size) {
+        case Maxwell::VertexAttribute::Size::Size_8:
+        case Maxwell::VertexAttribute::Size::Size_8_8:
+        case Maxwell::VertexAttribute::Size::Size_8_8_8:
+        case Maxwell::VertexAttribute::Size::Size_8_8_8_8:
+            return GL_BYTE;
+        case Maxwell::VertexAttribute::Size::Size_16:
+        case Maxwell::VertexAttribute::Size::Size_16_16:
+        case Maxwell::VertexAttribute::Size::Size_16_16_16:
+        case Maxwell::VertexAttribute::Size::Size_16_16_16_16:
+            return GL_SHORT;
         default:
             LOG_ERROR(Render_OpenGL, "Unimplemented vertex size={}", attrib.SizeString());
             return {};
@@ -401,24 +425,24 @@ inline GLenum StencilOp(Maxwell::StencilOp stencil) {
     return GL_KEEP;
 }
 
-inline GLenum FrontFace(Maxwell::Cull::FrontFace front_face) {
+inline GLenum FrontFace(Maxwell::FrontFace front_face) {
     switch (front_face) {
-    case Maxwell::Cull::FrontFace::ClockWise:
+    case Maxwell::FrontFace::ClockWise:
         return GL_CW;
-    case Maxwell::Cull::FrontFace::CounterClockWise:
+    case Maxwell::FrontFace::CounterClockWise:
         return GL_CCW;
     }
     LOG_ERROR(Render_OpenGL, "Unimplemented front face cull={}", static_cast<u32>(front_face));
     return GL_CCW;
 }
 
-inline GLenum CullFace(Maxwell::Cull::CullFace cull_face) {
+inline GLenum CullFace(Maxwell::CullFace cull_face) {
     switch (cull_face) {
-    case Maxwell::Cull::CullFace::Front:
+    case Maxwell::CullFace::Front:
         return GL_FRONT;
-    case Maxwell::Cull::CullFace::Back:
+    case Maxwell::CullFace::Back:
         return GL_BACK;
-    case Maxwell::Cull::CullFace::FrontAndBack:
+    case Maxwell::CullFace::FrontAndBack:
         return GL_FRONT_AND_BACK;
     }
     LOG_ERROR(Render_OpenGL, "Unimplemented cull face={}", static_cast<u32>(cull_face));

src/video_core/renderer_opengl/renderer_opengl.cpp

@@ -9,26 +9,163 @@
 #include <glad/glad.h>
 #include "common/assert.h"
 #include "common/logging/log.h"
+#include "common/microprofile.h"
 #include "common/telemetry.h"
 #include "core/core.h"
 #include "core/core_timing.h"
 #include "core/frontend/emu_window.h"
-#include "core/frontend/scope_acquire_window_context.h"
 #include "core/memory.h"
 #include "core/perf_stats.h"
 #include "core/settings.h"
 #include "core/telemetry_session.h"
 #include "video_core/morton.h"
 #include "video_core/renderer_opengl/gl_rasterizer.h"
+#include "video_core/renderer_opengl/gl_shader_manager.h"
 #include "video_core/renderer_opengl/renderer_opengl.h"
 
 namespace OpenGL {
 
+// If the size of this is too small, it ends up creating a soft cap on FPS as the renderer will have
+// to wait on available presentation frames.
+constexpr std::size_t SWAP_CHAIN_SIZE = 3;
+
+struct Frame {
+    u32 width{};                      /// Width of the frame (to detect resize)
+    u32 height{};                     /// Height of the frame
+    bool color_reloaded{};            /// Texture attachment was recreated (ie: resized)
+    OpenGL::OGLRenderbuffer color{};  /// Buffer shared between the render/present FBO
+    OpenGL::OGLFramebuffer render{};  /// FBO created on the render thread
+    OpenGL::OGLFramebuffer present{}; /// FBO created on the present thread
+    GLsync render_fence{};            /// Fence created on the render thread
+    GLsync present_fence{};           /// Fence created on the presentation thread
+    bool is_srgb{};                   /// Framebuffer is sRGB or RGB
+};
+
+/**
+ * For smooth Vsync rendering, we want to always present the latest frame that the core generates,
+ * but also make sure that rendering happens at the pace that the frontend dictates. This is a
+ * helper class that the renderer uses to sync frames between the render thread and the presentation
+ * thread
+ */
+class FrameMailbox {
+public:
+    std::mutex swap_chain_lock;
+    std::condition_variable present_cv;
+    std::array<Frame, SWAP_CHAIN_SIZE> swap_chain{};
+    std::queue<Frame*> free_queue;
+    std::deque<Frame*> present_queue;
+    Frame* previous_frame{};
+
+    FrameMailbox() {
+        for (auto& frame : swap_chain) {
+            free_queue.push(&frame);
+        }
+    }
+
+    ~FrameMailbox() {
+        // lock the mutex and clear out the present and free_queues and notify any people who are
+        // blocked to prevent deadlock on shutdown
+        std::scoped_lock lock{swap_chain_lock};
+        std::queue<Frame*>().swap(free_queue);
+        present_queue.clear();
+        present_cv.notify_all();
+    }
+
+    void ReloadPresentFrame(Frame* frame, u32 height, u32 width) {
+        frame->present.Release();
+        frame->present.Create();
+        GLint previous_draw_fbo{};
+        glGetIntegerv(GL_DRAW_FRAMEBUFFER_BINDING, &previous_draw_fbo);
+        glBindFramebuffer(GL_FRAMEBUFFER, frame->present.handle);
+        glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_RENDERBUFFER,
+                                  frame->color.handle);
+        if (glCheckFramebufferStatus(GL_FRAMEBUFFER) != GL_FRAMEBUFFER_COMPLETE) {
+            LOG_CRITICAL(Render_OpenGL, "Failed to recreate present FBO!");
+        }
+        glBindFramebuffer(GL_DRAW_FRAMEBUFFER, previous_draw_fbo);
+        frame->color_reloaded = false;
+    }
+
+    void ReloadRenderFrame(Frame* frame, u32 width, u32 height) {
+        // Recreate the color texture attachment
+        frame->color.Release();
+        frame->color.Create();
+        const GLenum internal_format = frame->is_srgb ? GL_SRGB8 : GL_RGB8;
+        glNamedRenderbufferStorage(frame->color.handle, internal_format, width, height);
+
+        // Recreate the FBO for the render target
+        frame->render.Release();
+        frame->render.Create();
+        glBindFramebuffer(GL_FRAMEBUFFER, frame->render.handle);
+        glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_RENDERBUFFER,
+                                  frame->color.handle);
+        if (glCheckFramebufferStatus(GL_FRAMEBUFFER) != GL_FRAMEBUFFER_COMPLETE) {
+            LOG_CRITICAL(Render_OpenGL, "Failed to recreate render FBO!");
+        }
+
+        frame->width = width;
+        frame->height = height;
+        frame->color_reloaded = true;
+    }
+
+    Frame* GetRenderFrame() {
+        std::unique_lock lock{swap_chain_lock};
+
+        // If theres no free frames, we will reuse the oldest render frame
+        if (free_queue.empty()) {
+            auto frame = present_queue.back();
+            present_queue.pop_back();
+            return frame;
+        }
+
+        Frame* frame = free_queue.front();
+        free_queue.pop();
+        return frame;
+    }
+
+    void ReleaseRenderFrame(Frame* frame) {
+        std::unique_lock lock{swap_chain_lock};
+        present_queue.push_front(frame);
+        present_cv.notify_one();
+    }
+
+    Frame* TryGetPresentFrame(int timeout_ms) {
+        std::unique_lock lock{swap_chain_lock};
+        // wait for new entries in the present_queue
+        present_cv.wait_for(lock, std::chrono::milliseconds(timeout_ms),
+                            [&] { return !present_queue.empty(); });
+        if (present_queue.empty()) {
+            // timed out waiting for a frame to draw so return the previous frame
+            return previous_frame;
+        }
+
+        // free the previous frame and add it back to the free queue
+        if (previous_frame) {
+            free_queue.push(previous_frame);
+        }
+
+        // the newest entries are pushed to the front of the queue
+        Frame* frame = present_queue.front();
+        present_queue.pop_front();
+        // remove all old entries from the present queue and move them back to the free_queue
+        for (auto f : present_queue) {
+            free_queue.push(f);
+        }
+        present_queue.clear();
+        previous_frame = frame;
+        return frame;
+    }
+};
+
 namespace {
 
-constexpr char vertex_shader[] = R"(
+constexpr char VERTEX_SHADER[] = R"(
 #version 430 core
 
+out gl_PerVertex {
+    vec4 gl_Position;
+};
+
 layout (location = 0) in vec2 vert_position;
 layout (location = 1) in vec2 vert_tex_coord;
 layout (location = 0) out vec2 frag_tex_coord;
@@ -49,7 +186,7 @@ void main() {
 }
 )";
 
-constexpr char fragment_shader[] = R"(
+constexpr char FRAGMENT_SHADER[] = R"(
 #version 430 core
 
 layout (location = 0) in vec2 frag_tex_coord;
@@ -58,7 +195,7 @@ layout (location = 0) out vec4 color;
 layout (binding = 0) uniform sampler2D color_texture;
 
 void main() {
-    color = texture(color_texture, frag_tex_coord);
+    color = vec4(texture(color_texture, frag_tex_coord).rgb, 1.0f);
 }
 )";
 
@@ -67,8 +204,8 @@ constexpr GLint TexCoordLocation = 1;
 constexpr GLint ModelViewMatrixLocation = 0;
 
 struct ScreenRectVertex {
-    constexpr ScreenRectVertex(GLfloat x, GLfloat y, GLfloat u, GLfloat v)
-        : position{{x, y}}, tex_coord{{u, v}} {}
+    constexpr ScreenRectVertex(u32 x, u32 y, GLfloat u, GLfloat v)
+        : position{{static_cast<GLfloat>(x), static_cast<GLfloat>(y)}}, tex_coord{{u, v}} {}
 
     std::array<GLfloat, 2> position;
     std::array<GLfloat, 2> tex_coord;
@@ -158,21 +295,81 @@ void APIENTRY DebugHandler(GLenum source, GLenum type, GLuint id, GLenum severit
 } // Anonymous namespace
 
 RendererOpenGL::RendererOpenGL(Core::Frontend::EmuWindow& emu_window, Core::System& system)
-    : VideoCore::RendererBase{emu_window}, emu_window{emu_window}, system{system} {}
+    : VideoCore::RendererBase{emu_window}, emu_window{emu_window}, system{system},
+      frame_mailbox{std::make_unique<FrameMailbox>()} {}
 
 RendererOpenGL::~RendererOpenGL() = default;
 
-void RendererOpenGL::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
-    // Maintain the rasterizer's state as a priority
-    OpenGLState prev_state = OpenGLState::GetCurState();
-    state.AllDirty();
-    state.Apply();
+MICROPROFILE_DEFINE(OpenGL_RenderFrame, "OpenGL", "Render Frame", MP_RGB(128, 128, 64));
+MICROPROFILE_DEFINE(OpenGL_WaitPresent, "OpenGL", "Wait For Present", MP_RGB(128, 128, 128));
 
+void RendererOpenGL::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
+    render_window.PollEvents();
+
+    if (!framebuffer) {
+        return;
+    }
+
+    PrepareRendertarget(framebuffer);
+    RenderScreenshot();
+
+    Frame* frame;
+    {
+        MICROPROFILE_SCOPE(OpenGL_WaitPresent);
+
+        frame = frame_mailbox->GetRenderFrame();
+
+        // Clean up sync objects before drawing
+
+        // INTEL driver workaround. We can't delete the previous render sync object until we are
+        // sure that the presentation is done
+        if (frame->present_fence) {
+            glClientWaitSync(frame->present_fence, 0, GL_TIMEOUT_IGNORED);
+        }
+
+        // delete the draw fence if the frame wasn't presented
+        if (frame->render_fence) {
+            glDeleteSync(frame->render_fence);
+            frame->render_fence = 0;
+        }
+
+        // wait for the presentation to be done
+        if (frame->present_fence) {
+            glWaitSync(frame->present_fence, 0, GL_TIMEOUT_IGNORED);
+            glDeleteSync(frame->present_fence);
+            frame->present_fence = 0;
+        }
+    }
+
+    {
+        MICROPROFILE_SCOPE(OpenGL_RenderFrame);
+        const auto& layout = render_window.GetFramebufferLayout();
+
+        // Recreate the frame if the size of the window has changed
+        if (layout.width != frame->width || layout.height != frame->height ||
+            screen_info.display_srgb != frame->is_srgb) {
+            LOG_DEBUG(Render_OpenGL, "Reloading render frame");
+            frame->is_srgb = screen_info.display_srgb;
+            frame_mailbox->ReloadRenderFrame(frame, layout.width, layout.height);
+        }
+        glBindFramebuffer(GL_DRAW_FRAMEBUFFER, frame->render.handle);
+        DrawScreen(layout);
+        // Create a fence for the frontend to wait on and swap this frame to OffTex
+        frame->render_fence = glFenceSync(GL_SYNC_GPU_COMMANDS_COMPLETE, 0);
+        glFlush();
+        frame_mailbox->ReleaseRenderFrame(frame);
+        m_current_frame++;
+        rasterizer->TickFrame();
+    }
+}
+
+void RendererOpenGL::PrepareRendertarget(const Tegra::FramebufferConfig* framebuffer) {
     if (framebuffer) {
         // If framebuffer is provided, reload it from memory to a texture
         if (screen_info.texture.width != static_cast<GLsizei>(framebuffer->width) ||
             screen_info.texture.height != static_cast<GLsizei>(framebuffer->height) ||
-            screen_info.texture.pixel_format != framebuffer->pixel_format) {
+            screen_info.texture.pixel_format != framebuffer->pixel_format ||
+            gl_framebuffer_data.empty()) {
             // Reallocate texture if the framebuffer size has changed.
             // This is expected to not happen very often and hence should not be a
             // performance problem.
@@ -181,22 +378,7 @@ void RendererOpenGL::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
 
         // Load the framebuffer from memory, draw it to the screen, and swap buffers
         LoadFBToScreenInfo(*framebuffer);
-
-        if (renderer_settings.screenshot_requested)
-            CaptureScreenshot();
-
-        DrawScreen(render_window.GetFramebufferLayout());
-
-        rasterizer->TickFrame();
-
-        render_window.SwapBuffers();
     }
-
-    render_window.PollEvents();
-
-    // Restore the rasterizer state
-    prev_state.AllDirty();
-    prev_state.Apply();
 }
 
 void RendererOpenGL::LoadFBToScreenInfo(const Tegra::FramebufferConfig& framebuffer) {
@@ -249,31 +431,24 @@ void RendererOpenGL::InitOpenGLObjects() {
     glClearColor(Settings::values.bg_red, Settings::values.bg_green, Settings::values.bg_blue,
                  0.0f);
 
-    // Link shaders and get variable locations
-    shader.CreateFromSource(vertex_shader, nullptr, fragment_shader);
-    state.draw.shader_program = shader.handle;
-    state.AllDirty();
-    state.Apply();
+    // Create shader programs
+    OGLShader vertex_shader;
+    vertex_shader.Create(VERTEX_SHADER, GL_VERTEX_SHADER);
+
+    OGLShader fragment_shader;
+    fragment_shader.Create(FRAGMENT_SHADER, GL_FRAGMENT_SHADER);
+
+    vertex_program.Create(true, false, vertex_shader.handle);
+    fragment_program.Create(true, false, fragment_shader.handle);
+
+    // Create program pipeline
+    program_manager.Create();
 
     // Generate VBO handle for drawing
     vertex_buffer.Create();
 
-    // Generate VAO
-    vertex_array.Create();
-    state.draw.vertex_array = vertex_array.handle;
-
     // Attach vertex data to VAO
     glNamedBufferData(vertex_buffer.handle, sizeof(ScreenRectVertex) * 4, nullptr, GL_STREAM_DRAW);
-    glVertexArrayAttribFormat(vertex_array.handle, PositionLocation, 2, GL_FLOAT, GL_FALSE,
-                              offsetof(ScreenRectVertex, position));
-    glVertexArrayAttribFormat(vertex_array.handle, TexCoordLocation, 2, GL_FLOAT, GL_FALSE,
-                              offsetof(ScreenRectVertex, tex_coord));
-    glVertexArrayAttribBinding(vertex_array.handle, PositionLocation, 0);
-    glVertexArrayAttribBinding(vertex_array.handle, TexCoordLocation, 0);
-    glEnableVertexArrayAttrib(vertex_array.handle, PositionLocation);
-    glEnableVertexArrayAttrib(vertex_array.handle, TexCoordLocation);
-    glVertexArrayVertexBuffer(vertex_array.handle, 0, vertex_buffer.handle, 0,
-                              sizeof(ScreenRectVertex));
 
     // Allocate textures for the screen
     screen_info.texture.resource.Create(GL_TEXTURE_2D);
@@ -306,7 +481,8 @@ void RendererOpenGL::CreateRasterizer() {
     if (rasterizer) {
         return;
     }
-    rasterizer = std::make_unique<RasterizerOpenGL>(system, emu_window, screen_info);
+    rasterizer = std::make_unique<RasterizerOpenGL>(system, emu_window, screen_info,
+                                                    program_manager, state_tracker);
 }
 
 void RendererOpenGL::ConfigureFramebufferTexture(TextureInfo& texture,
@@ -345,8 +521,19 @@ void RendererOpenGL::ConfigureFramebufferTexture(TextureInfo& texture,
     glTextureStorage2D(texture.resource.handle, 1, internal_format, texture.width, texture.height);
 }
 
-void RendererOpenGL::DrawScreenTriangles(const ScreenInfo& screen_info, float x, float y, float w,
-                                         float h) {
+void RendererOpenGL::DrawScreen(const Layout::FramebufferLayout& layout) {
+    if (renderer_settings.set_background_color) {
+        // Update background color before drawing
+        glClearColor(Settings::values.bg_red, Settings::values.bg_green, Settings::values.bg_blue,
+                     0.0f);
+    }
+
+    // Set projection matrix
+    const std::array ortho_matrix =
+        MakeOrthographicMatrix(static_cast<float>(layout.width), static_cast<float>(layout.height));
+    glProgramUniformMatrix3x2fv(vertex_program.handle, ModelViewMatrixLocation, 1, GL_FALSE,
+                                std::data(ortho_matrix));
+
     const auto& texcoords = screen_info.display_texcoords;
     auto left = texcoords.left;
     auto right = texcoords.right;
@@ -378,60 +565,127 @@ void RendererOpenGL::DrawScreenTriangles(const ScreenInfo& screen_info, float x,
                   static_cast<f32>(screen_info.texture.height);
     }
 
+    const auto& screen = layout.screen;
     const std::array vertices = {
-        ScreenRectVertex(x, y, texcoords.top * scale_u, left * scale_v),
-        ScreenRectVertex(x + w, y, texcoords.bottom * scale_u, left * scale_v),
-        ScreenRectVertex(x, y + h, texcoords.top * scale_u, right * scale_v),
-        ScreenRectVertex(x + w, y + h, texcoords.bottom * scale_u, right * scale_v),
+        ScreenRectVertex(screen.left, screen.top, texcoords.top * scale_u, left * scale_v),
+        ScreenRectVertex(screen.right, screen.top, texcoords.bottom * scale_u, left * scale_v),
+        ScreenRectVertex(screen.left, screen.bottom, texcoords.top * scale_u, right * scale_v),
+        ScreenRectVertex(screen.right, screen.bottom, texcoords.bottom * scale_u, right * scale_v),
     };
-
-    state.textures[0] = screen_info.display_texture;
-    state.framebuffer_srgb.enabled = screen_info.display_srgb;
-    state.AllDirty();
-    state.Apply();
     glNamedBufferSubData(vertex_buffer.handle, 0, sizeof(vertices), std::data(vertices));
+
+    // TODO: Signal state tracker about these changes
+    state_tracker.NotifyScreenDrawVertexArray();
+    state_tracker.NotifyViewport0();
+    state_tracker.NotifyScissor0();
+    state_tracker.NotifyColorMask0();
+    state_tracker.NotifyBlend0();
+    state_tracker.NotifyFramebuffer();
+    state_tracker.NotifyFrontFace();
+    state_tracker.NotifyCullTest();
+    state_tracker.NotifyDepthTest();
+    state_tracker.NotifyStencilTest();
+    state_tracker.NotifyPolygonOffset();
+    state_tracker.NotifyRasterizeEnable();
+    state_tracker.NotifyFramebufferSRGB();
+    state_tracker.NotifyLogicOp();
+    state_tracker.NotifyClipControl();
+    state_tracker.NotifyAlphaTest();
+
+    program_manager.UseVertexShader(vertex_program.handle);
+    program_manager.UseGeometryShader(0);
+    program_manager.UseFragmentShader(fragment_program.handle);
+    program_manager.BindGraphicsPipeline();
+
+    glEnable(GL_CULL_FACE);
+    if (screen_info.display_srgb) {
+        glEnable(GL_FRAMEBUFFER_SRGB);
+    } else {
+        glDisable(GL_FRAMEBUFFER_SRGB);
+    }
+    glDisable(GL_COLOR_LOGIC_OP);
+    glDisable(GL_DEPTH_TEST);
+    glDisable(GL_STENCIL_TEST);
+    glDisable(GL_POLYGON_OFFSET_FILL);
+    glDisable(GL_RASTERIZER_DISCARD);
+    glDisable(GL_ALPHA_TEST);
+    glDisablei(GL_BLEND, 0);
+    glDisablei(GL_SCISSOR_TEST, 0);
+    glCullFace(GL_BACK);
+    glFrontFace(GL_CW);
+    glColorMaski(0, GL_TRUE, GL_TRUE, GL_TRUE, GL_TRUE);
+    glClipControl(GL_LOWER_LEFT, GL_ZERO_TO_ONE);
+    glViewportIndexedf(0, 0.0f, 0.0f, static_cast<GLfloat>(layout.width),
+                       static_cast<GLfloat>(layout.height));
+    glDepthRangeIndexed(0, 0.0, 0.0);
+
+    glEnableVertexAttribArray(PositionLocation);
+    glEnableVertexAttribArray(TexCoordLocation);
+    glVertexAttribDivisor(PositionLocation, 0);
+    glVertexAttribDivisor(TexCoordLocation, 0);
+    glVertexAttribFormat(PositionLocation, 2, GL_FLOAT, GL_FALSE,
+                         offsetof(ScreenRectVertex, position));
+    glVertexAttribFormat(TexCoordLocation, 2, GL_FLOAT, GL_FALSE,
+                         offsetof(ScreenRectVertex, tex_coord));
+    glVertexAttribBinding(PositionLocation, 0);
+    glVertexAttribBinding(TexCoordLocation, 0);
+    glBindVertexBuffer(0, vertex_buffer.handle, 0, sizeof(ScreenRectVertex));
+
+    glBindTextureUnit(0, screen_info.display_texture);
+    glBindSampler(0, 0);
+
+    glClear(GL_COLOR_BUFFER_BIT);
     glDrawArrays(GL_TRIANGLE_STRIP, 0, 4);
-    // Restore default state
-    state.framebuffer_srgb.enabled = false;
-    state.textures[0] = 0;
-    state.AllDirty();
-    state.Apply();
 }
 
-void RendererOpenGL::DrawScreen(const Layout::FramebufferLayout& layout) {
-    if (renderer_settings.set_background_color) {
-        // Update background color before drawing
-        glClearColor(Settings::values.bg_red, Settings::values.bg_green, Settings::values.bg_blue,
-                     0.0f);
+void RendererOpenGL::TryPresent(int timeout_ms) {
+    const auto& layout = render_window.GetFramebufferLayout();
+    auto frame = frame_mailbox->TryGetPresentFrame(timeout_ms);
+    if (!frame) {
+        LOG_DEBUG(Render_OpenGL, "TryGetPresentFrame returned no frame to present");
+        return;
     }
 
-    const auto& screen = layout.screen;
-
-    glViewport(0, 0, layout.width, layout.height);
+    // Clearing before a full overwrite of a fbo can signal to drivers that they can avoid a
+    // readback since we won't be doing any blending
     glClear(GL_COLOR_BUFFER_BIT);
 
-    // Set projection matrix
-    const std::array ortho_matrix =
-        MakeOrthographicMatrix(static_cast<float>(layout.width), static_cast<float>(layout.height));
-    glUniformMatrix3x2fv(ModelViewMatrixLocation, 1, GL_FALSE, ortho_matrix.data());
+    // Recreate the presentation FBO if the color attachment was changed
+    if (frame->color_reloaded) {
+        LOG_DEBUG(Render_OpenGL, "Reloading present frame");
+        frame_mailbox->ReloadPresentFrame(frame, layout.width, layout.height);
+    }
+    glWaitSync(frame->render_fence, 0, GL_TIMEOUT_IGNORED);
+    // INTEL workaround.
+    // Normally we could just delete the draw fence here, but due to driver bugs, we can just delete
+    // it on the emulation thread without too much penalty
+    // glDeleteSync(frame.render_sync);
+    // frame.render_sync = 0;
 
-    DrawScreenTriangles(screen_info, static_cast<float>(screen.left),
-                        static_cast<float>(screen.top), static_cast<float>(screen.GetWidth()),
-                        static_cast<float>(screen.GetHeight()));
+    glBindFramebuffer(GL_READ_FRAMEBUFFER, frame->present.handle);
+    glBlitFramebuffer(0, 0, frame->width, frame->height, 0, 0, layout.width, layout.height,
+                      GL_COLOR_BUFFER_BIT, GL_LINEAR);
 
-    m_current_frame++;
+    // Insert fence for the main thread to block on
+    frame->present_fence = glFenceSync(GL_SYNC_GPU_COMMANDS_COMPLETE, 0);
+    glFlush();
+
+    glBindFramebuffer(GL_READ_FRAMEBUFFER, 0);
 }
 
-void RendererOpenGL::UpdateFramerate() {}
+void RendererOpenGL::RenderScreenshot() {
+    if (!renderer_settings.screenshot_requested) {
+        return;
+    }
+
+    GLint old_read_fb;
+    GLint old_draw_fb;
+    glGetIntegerv(GL_READ_FRAMEBUFFER_BINDING, &old_read_fb);
+    glGetIntegerv(GL_DRAW_FRAMEBUFFER_BINDING, &old_draw_fb);
 
-void RendererOpenGL::CaptureScreenshot() {
     // Draw the current frame to the screenshot framebuffer
     screenshot_framebuffer.Create();
-    GLuint old_read_fb = state.draw.read_framebuffer;
-    GLuint old_draw_fb = state.draw.draw_framebuffer;
-    state.draw.read_framebuffer = state.draw.draw_framebuffer = screenshot_framebuffer.handle;
-    state.AllDirty();
-    state.Apply();
+    glBindFramebuffer(GL_FRAMEBUFFER, screenshot_framebuffer.handle);
 
     Layout::FramebufferLayout layout{renderer_settings.screenshot_framebuffer_layout};
 
@@ -448,19 +702,16 @@ void RendererOpenGL::CaptureScreenshot() {
                  renderer_settings.screenshot_bits);
 
     screenshot_framebuffer.Release();
-    state.draw.read_framebuffer = old_read_fb;
-    state.draw.draw_framebuffer = old_draw_fb;
-    state.AllDirty();
-    state.Apply();
     glDeleteRenderbuffers(1, &renderbuffer);
 
+    glBindFramebuffer(GL_READ_FRAMEBUFFER, old_read_fb);
+    glBindFramebuffer(GL_DRAW_FRAMEBUFFER, old_draw_fb);
+
     renderer_settings.screenshot_complete_callback();
     renderer_settings.screenshot_requested = false;
 }
 
 bool RendererOpenGL::Init() {
-    Core::Frontend::ScopeAcquireWindowContext acquire_context{render_window};
-
     if (GLAD_GL_KHR_debug) {
         glEnable(GL_DEBUG_OUTPUT);
         glDebugMessageCallback(DebugHandler, nullptr);

src/video_core/renderer_opengl/renderer_opengl.h

@@ -10,7 +10,8 @@
 #include "common/math_util.h"
 #include "video_core/renderer_base.h"
 #include "video_core/renderer_opengl/gl_resource_manager.h"
-#include "video_core/renderer_opengl/gl_state.h"
+#include "video_core/renderer_opengl/gl_shader_manager.h"
+#include "video_core/renderer_opengl/gl_state_tracker.h"
 
 namespace Core {
 class System;
@@ -44,19 +45,23 @@ struct ScreenInfo {
     TextureInfo texture;
 };
 
+struct PresentationTexture {
+    u32 width = 0;
+    u32 height = 0;
+    OGLTexture texture;
+};
+
+class FrameMailbox;
+
 class RendererOpenGL final : public VideoCore::RendererBase {
 public:
     explicit RendererOpenGL(Core::Frontend::EmuWindow& emu_window, Core::System& system);
     ~RendererOpenGL() override;
 
-    /// Swap buffers (render frame)
-    void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) override;
-
-    /// Initialize the renderer
     bool Init() override;
-
-    /// Shutdown the renderer
     void ShutDown() override;
+    void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) override;
+    void TryPresent(int timeout_ms) override;
 
 private:
     /// Initializes the OpenGL state and creates persistent objects.
@@ -72,12 +77,7 @@ private:
     /// Draws the emulated screens to the emulator window.
     void DrawScreen(const Layout::FramebufferLayout& layout);
 
-    void DrawScreenTriangles(const ScreenInfo& screen_info, float x, float y, float w, float h);
-
-    /// Updates the framerate.
-    void UpdateFramerate();
-
-    void CaptureScreenshot();
+    void RenderScreenshot();
 
     /// Loads framebuffer from emulated memory into the active OpenGL texture.
     void LoadFBToScreenInfo(const Tegra::FramebufferConfig& framebuffer);
@@ -87,26 +87,34 @@ private:
     void LoadColorToActiveGLTexture(u8 color_r, u8 color_g, u8 color_b, u8 color_a,
                                     const TextureInfo& texture);
 
+    void PrepareRendertarget(const Tegra::FramebufferConfig* framebuffer);
+
     Core::Frontend::EmuWindow& emu_window;
     Core::System& system;
 
-    OpenGLState state;
+    StateTracker state_tracker{system};
 
     // OpenGL object IDs
-    OGLVertexArray vertex_array;
     OGLBuffer vertex_buffer;
-    OGLProgram shader;
+    OGLProgram vertex_program;
+    OGLProgram fragment_program;
     OGLFramebuffer screenshot_framebuffer;
 
     /// Display information for Switch screen
     ScreenInfo screen_info;
 
+    /// Global dummy shader pipeline
+    GLShader::ProgramManager program_manager;
+
     /// OpenGL framebuffer data
     std::vector<u8> gl_framebuffer_data;
 
     /// Used for transforming the framebuffer orientation
     Tegra::FramebufferConfig::TransformFlags framebuffer_transform_flags;
     Common::Rectangle<int> framebuffer_crop_rect;
+
+    /// Frame presentation mailbox
+    std::unique_ptr<FrameMailbox> frame_mailbox;
 };
 
 } // namespace OpenGL

src/video_core/renderer_opengl/utils.cpp

@@ -9,6 +9,7 @@
 #include <glad/glad.h>
 
 #include "common/common_types.h"
+#include "video_core/renderer_opengl/gl_state_tracker.h"
 #include "video_core/renderer_opengl/utils.h"
 
 namespace OpenGL {
@@ -20,12 +21,12 @@ struct VertexArrayPushBuffer::Entry {
     GLsizei stride{};
 };
 
-VertexArrayPushBuffer::VertexArrayPushBuffer() = default;
+VertexArrayPushBuffer::VertexArrayPushBuffer(StateTracker& state_tracker)
+    : state_tracker{state_tracker} {}
 
 VertexArrayPushBuffer::~VertexArrayPushBuffer() = default;
 
-void VertexArrayPushBuffer::Setup(GLuint vao_) {
-    vao = vao_;
+void VertexArrayPushBuffer::Setup() {
     index_buffer = nullptr;
     vertex_buffers.clear();
 }
@@ -41,13 +42,11 @@ void VertexArrayPushBuffer::SetVertexBuffer(GLuint binding_index, const GLuint*
 
 void VertexArrayPushBuffer::Bind() {
     if (index_buffer) {
-        glVertexArrayElementBuffer(vao, *index_buffer);
+        state_tracker.BindIndexBuffer(*index_buffer);
     }
 
-    // TODO(Rodrigo): Find a way to ARB_multi_bind this
     for (const auto& entry : vertex_buffers) {
-        glVertexArrayVertexBuffer(vao, entry.binding_index, *entry.buffer, entry.offset,
-                                  entry.stride);
+        glBindVertexBuffer(entry.binding_index, *entry.buffer, entry.offset, entry.stride);
     }
 }
 

src/video_core/renderer_opengl/utils.h

@@ -11,12 +11,14 @@
 
 namespace OpenGL {
 
+class StateTracker;
+
 class VertexArrayPushBuffer final {
 public:
-    explicit VertexArrayPushBuffer();
+    explicit VertexArrayPushBuffer(StateTracker& state_tracker);
     ~VertexArrayPushBuffer();
 
-    void Setup(GLuint vao_);
+    void Setup();
 
     void SetIndexBuffer(const GLuint* buffer);
 
@@ -28,7 +30,8 @@ public:
 private:
     struct Entry;
 
-    GLuint vao{};
+    StateTracker& state_tracker;
+
     const GLuint* index_buffer{};
     std::vector<Entry> vertex_buffers;
 };

src/video_core/renderer_vulkan/fixed_pipeline_state.cpp

@@ -112,19 +112,18 @@ constexpr FixedPipelineState::Rasterizer GetRasterizerState(const Maxwell& regs)
     const auto& clip = regs.view_volume_clip_control;
     const bool depth_clamp_enabled = clip.depth_clamp_near == 1 || clip.depth_clamp_far == 1;
 
-    Maxwell::Cull::FrontFace front_face = regs.cull.front_face;
+    Maxwell::FrontFace front_face = regs.front_face;
     if (regs.screen_y_control.triangle_rast_flip != 0 &&
         regs.viewport_transform[0].scale_y > 0.0f) {
-        if (front_face == Maxwell::Cull::FrontFace::CounterClockWise)
-            front_face = Maxwell::Cull::FrontFace::ClockWise;
-        else if (front_face == Maxwell::Cull::FrontFace::ClockWise)
-            front_face = Maxwell::Cull::FrontFace::CounterClockWise;
+        if (front_face == Maxwell::FrontFace::CounterClockWise)
+            front_face = Maxwell::FrontFace::ClockWise;
+        else if (front_face == Maxwell::FrontFace::ClockWise)
+            front_face = Maxwell::FrontFace::CounterClockWise;
     }
 
     const bool gl_ndc = regs.depth_mode == Maxwell::DepthMode::MinusOneToOne;
-    return FixedPipelineState::Rasterizer(regs.cull.enabled, depth_bias_enabled,
-                                          depth_clamp_enabled, gl_ndc, regs.cull.cull_face,
-                                          front_face);
+    return FixedPipelineState::Rasterizer(regs.cull_test_enabled, depth_bias_enabled,
+                                          depth_clamp_enabled, gl_ndc, regs.cull_face, front_face);
 }
 
 } // Anonymous namespace

src/video_core/renderer_vulkan/fixed_pipeline_state.h

@@ -171,8 +171,8 @@ struct FixedPipelineState {
 
     struct Rasterizer {
         constexpr Rasterizer(bool cull_enable, bool depth_bias_enable, bool depth_clamp_enable,
-                             bool ndc_minus_one_to_one, Maxwell::Cull::CullFace cull_face,
-                             Maxwell::Cull::FrontFace front_face)
+                             bool ndc_minus_one_to_one, Maxwell::CullFace cull_face,
+                             Maxwell::FrontFace front_face)
             : cull_enable{cull_enable}, depth_bias_enable{depth_bias_enable},
               depth_clamp_enable{depth_clamp_enable}, ndc_minus_one_to_one{ndc_minus_one_to_one},
               cull_face{cull_face}, front_face{front_face} {}
@@ -182,8 +182,8 @@ struct FixedPipelineState {
         bool depth_bias_enable;
         bool depth_clamp_enable;
         bool ndc_minus_one_to_one;
-        Maxwell::Cull::CullFace cull_face;
-        Maxwell::Cull::FrontFace front_face;
+        Maxwell::CullFace cull_face;
+        Maxwell::FrontFace front_face;
 
         std::size_t Hash() const noexcept;
 

src/video_core/renderer_vulkan/maxwell_to_vk.cpp

@@ -120,7 +120,7 @@ struct FormatTuple {
     {vk::Format::eA8B8G8R8UintPack32, Attachable | Storage},     // ABGR8UI
     {vk::Format::eB5G6R5UnormPack16, {}},                        // B5G6R5U
     {vk::Format::eA2B10G10R10UnormPack32, Attachable | Storage}, // A2B10G10R10U
-    {vk::Format::eA1R5G5B5UnormPack16, Attachable | Storage},    // A1B5G5R5U (flipped with swizzle)
+    {vk::Format::eA1R5G5B5UnormPack16, Attachable},              // A1B5G5R5U (flipped with swizzle)
     {vk::Format::eR8Unorm, Attachable | Storage},                // R8U
     {vk::Format::eR8Uint, Attachable | Storage},                 // R8UI
     {vk::Format::eR16G16B16A16Sfloat, Attachable | Storage},     // RGBA16F
@@ -371,8 +371,22 @@ vk::Format VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttr
         }
     case Maxwell::VertexAttribute::Type::UnsignedScaled:
         switch (size) {
+        case Maxwell::VertexAttribute::Size::Size_8:
+            return vk::Format::eR8Uscaled;
         case Maxwell::VertexAttribute::Size::Size_8_8:
             return vk::Format::eR8G8Uscaled;
+        case Maxwell::VertexAttribute::Size::Size_8_8_8:
+            return vk::Format::eR8G8B8Uscaled;
+        case Maxwell::VertexAttribute::Size::Size_8_8_8_8:
+            return vk::Format::eR8G8B8A8Uscaled;
+        case Maxwell::VertexAttribute::Size::Size_16:
+            return vk::Format::eR16Uscaled;
+        case Maxwell::VertexAttribute::Size::Size_16_16:
+            return vk::Format::eR16G16Uscaled;
+        case Maxwell::VertexAttribute::Size::Size_16_16_16:
+            return vk::Format::eR16G16B16Uscaled;
+        case Maxwell::VertexAttribute::Size::Size_16_16_16_16:
+            return vk::Format::eR16G16B16A16Uscaled;
         default:
             break;
         }
@@ -572,24 +586,24 @@ vk::BlendFactor BlendFactor(Maxwell::Blend::Factor factor) {
     return {};
 }
 
-vk::FrontFace FrontFace(Maxwell::Cull::FrontFace front_face) {
+vk::FrontFace FrontFace(Maxwell::FrontFace front_face) {
     switch (front_face) {
-    case Maxwell::Cull::FrontFace::ClockWise:
+    case Maxwell::FrontFace::ClockWise:
         return vk::FrontFace::eClockwise;
-    case Maxwell::Cull::FrontFace::CounterClockWise:
+    case Maxwell::FrontFace::CounterClockWise:
         return vk::FrontFace::eCounterClockwise;
     }
     UNIMPLEMENTED_MSG("Unimplemented front face={}", static_cast<u32>(front_face));
     return {};
 }
 
-vk::CullModeFlags CullFace(Maxwell::Cull::CullFace cull_face) {
+vk::CullModeFlags CullFace(Maxwell::CullFace cull_face) {
     switch (cull_face) {
-    case Maxwell::Cull::CullFace::Front:
+    case Maxwell::CullFace::Front:
         return vk::CullModeFlagBits::eFront;
-    case Maxwell::Cull::CullFace::Back:
+    case Maxwell::CullFace::Back:
         return vk::CullModeFlagBits::eBack;
-    case Maxwell::Cull::CullFace::FrontAndBack:
+    case Maxwell::CullFace::FrontAndBack:
         return vk::CullModeFlagBits::eFrontAndBack;
     }
     UNIMPLEMENTED_MSG("Unimplemented cull face={}", static_cast<u32>(cull_face));

src/video_core/renderer_vulkan/maxwell_to_vk.h

@@ -54,9 +54,9 @@ vk::BlendOp BlendEquation(Maxwell::Blend::Equation equation);
 
 vk::BlendFactor BlendFactor(Maxwell::Blend::Factor factor);
 
-vk::FrontFace FrontFace(Maxwell::Cull::FrontFace front_face);
+vk::FrontFace FrontFace(Maxwell::FrontFace front_face);
 
-vk::CullModeFlags CullFace(Maxwell::Cull::CullFace cull_face);
+vk::CullModeFlags CullFace(Maxwell::CullFace cull_face);
 
 vk::ComponentSwizzle SwizzleSource(Tegra::Texture::SwizzleSource swizzle);
 

src/video_core/renderer_vulkan/renderer_vulkan.cpp

@@ -27,6 +27,7 @@
 #include "video_core/renderer_vulkan/vk_rasterizer.h"
 #include "video_core/renderer_vulkan/vk_resource_manager.h"
 #include "video_core/renderer_vulkan/vk_scheduler.h"
+#include "video_core/renderer_vulkan/vk_state_tracker.h"
 #include "video_core/renderer_vulkan/vk_swapchain.h"
 
 namespace Vulkan {
@@ -106,8 +107,14 @@ RendererVulkan::~RendererVulkan() {
 }
 
 void RendererVulkan::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
+    render_window.PollEvents();
+
+    if (!framebuffer) {
+        return;
+    }
+
     const auto& layout = render_window.GetFramebufferLayout();
-    if (framebuffer && layout.width > 0 && layout.height > 0 && render_window.IsShown()) {
+    if (layout.width > 0 && layout.height > 0 && render_window.IsShown()) {
         const VAddr framebuffer_addr = framebuffer->address + framebuffer->offset;
         const bool use_accelerated =
             rasterizer->AccelerateDisplay(*framebuffer, framebuffer_addr, framebuffer->stride);
@@ -128,13 +135,16 @@ void RendererVulkan::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
             blit_screen->Recreate();
         }
 
-        render_window.SwapBuffers();
         rasterizer->TickFrame();
     }
 
     render_window.PollEvents();
 }
 
+void RendererVulkan::TryPresent(int /*timeout_ms*/) {
+    // TODO (bunnei): ImplementMe
+}
+
 bool RendererVulkan::Init() {
     PFN_vkGetInstanceProcAddr vkGetInstanceProcAddr{};
     render_window.RetrieveVulkanHandlers(&vkGetInstanceProcAddr, &instance, &surface);
@@ -168,10 +178,13 @@ bool RendererVulkan::Init() {
     swapchain = std::make_unique<VKSwapchain>(surface, *device);
     swapchain->Create(framebuffer.width, framebuffer.height, false);
 
-    scheduler = std::make_unique<VKScheduler>(*device, *resource_manager);
+    state_tracker = std::make_unique<StateTracker>(system);
+
+    scheduler = std::make_unique<VKScheduler>(*device, *resource_manager, *state_tracker);
 
     rasterizer = std::make_unique<RasterizerVulkan>(system, render_window, screen_info, *device,
-                                                    *resource_manager, *memory_manager, *scheduler);
+                                                    *resource_manager, *memory_manager,
+                                                    *state_tracker, *scheduler);
 
     blit_screen = std::make_unique<VKBlitScreen>(system, render_window, *rasterizer, *device,
                                                  *resource_manager, *memory_manager, *swapchain,
@@ -262,4 +275,4 @@ void RendererVulkan::Report() const {
     telemetry_session.AddField(field, "GPU_Vulkan_Extensions", extensions);
 }
 
-} // namespace Vulkan
+} // namespace Vulkan

src/video_core/renderer_vulkan/renderer_vulkan.h

@@ -4,8 +4,10 @@
 
 #pragma once
 
+#include <memory>
 #include <optional>
 #include <vector>
+
 #include "video_core/renderer_base.h"
 #include "video_core/renderer_vulkan/declarations.h"
 
@@ -15,6 +17,7 @@ class System;
 
 namespace Vulkan {
 
+class StateTracker;
 class VKBlitScreen;
 class VKDevice;
 class VKFence;
@@ -36,14 +39,10 @@ public:
     explicit RendererVulkan(Core::Frontend::EmuWindow& window, Core::System& system);
     ~RendererVulkan() override;
 
-    /// Swap buffers (render frame)
-    void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) override;
-
-    /// Initialize the renderer
     bool Init() override;
-
-    /// Shutdown the renderer
     void ShutDown() override;
+    void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) override;
+    void TryPresent(int timeout_ms) override;
 
 private:
     std::optional<vk::DebugUtilsMessengerEXT> CreateDebugCallback(
@@ -65,6 +64,7 @@ private:
     std::unique_ptr<VKSwapchain> swapchain;
     std::unique_ptr<VKMemoryManager> memory_manager;
     std::unique_ptr<VKResourceManager> resource_manager;
+    std::unique_ptr<StateTracker> state_tracker;
     std::unique_ptr<VKScheduler> scheduler;
     std::unique_ptr<VKBlitScreen> blit_screen;
 };

src/video_core/renderer_vulkan/vk_compute_pipeline.cpp

@@ -73,7 +73,7 @@ UniqueDescriptorUpdateTemplate VKComputePipeline::CreateDescriptorUpdateTemplate
     std::vector<vk::DescriptorUpdateTemplateEntry> template_entries;
     u32 binding = 0;
     u32 offset = 0;
-    FillDescriptorUpdateTemplateEntries(device, entries, binding, offset, template_entries);
+    FillDescriptorUpdateTemplateEntries(entries, binding, offset, template_entries);
     if (template_entries.empty()) {
         // If the shader doesn't use descriptor sets, skip template creation.
         return UniqueDescriptorUpdateTemplate{};

src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp

@@ -97,8 +97,7 @@ UniqueDescriptorUpdateTemplate VKGraphicsPipeline::CreateDescriptorUpdateTemplat
     u32 offset = 0;
     for (const auto& stage : program) {
         if (stage) {
-            FillDescriptorUpdateTemplateEntries(device, stage->entries, binding, offset,
-                                                template_entries);
+            FillDescriptorUpdateTemplateEntries(stage->entries, binding, offset, template_entries);
         }
     }
     if (template_entries.empty()) {

src/video_core/renderer_vulkan/vk_pipeline_cache.cpp

@@ -36,6 +36,13 @@ using Tegra::Engines::ShaderType;
 
 namespace {
 
+// C++20's using enum
+constexpr auto eUniformBuffer = vk::DescriptorType::eUniformBuffer;
+constexpr auto eStorageBuffer = vk::DescriptorType::eStorageBuffer;
+constexpr auto eUniformTexelBuffer = vk::DescriptorType::eUniformTexelBuffer;
+constexpr auto eCombinedImageSampler = vk::DescriptorType::eCombinedImageSampler;
+constexpr auto eStorageImage = vk::DescriptorType::eStorageImage;
+
 constexpr VideoCommon::Shader::CompilerSettings compiler_settings{
     VideoCommon::Shader::CompileDepth::FullDecompile};
 
@@ -119,23 +126,32 @@ ShaderType GetShaderType(Maxwell::ShaderProgram program) {
     }
 }
 
+template <vk::DescriptorType descriptor_type, class Container>
+void AddBindings(std::vector<vk::DescriptorSetLayoutBinding>& bindings, u32& binding,
+                 vk::ShaderStageFlags stage_flags, const Container& container) {
+    const u32 num_entries = static_cast<u32>(std::size(container));
+    for (std::size_t i = 0; i < num_entries; ++i) {
+        u32 count = 1;
+        if constexpr (descriptor_type == eCombinedImageSampler) {
+            // Combined image samplers can be arrayed.
+            count = container[i].Size();
+        }
+        bindings.emplace_back(binding++, descriptor_type, count, stage_flags, nullptr);
+    }
+}
+
 u32 FillDescriptorLayout(const ShaderEntries& entries,
                          std::vector<vk::DescriptorSetLayoutBinding>& bindings,
                          Maxwell::ShaderProgram program_type, u32 base_binding) {
     const ShaderType stage = GetStageFromProgram(program_type);
-    const vk::ShaderStageFlags stage_flags = MaxwellToVK::ShaderStage(stage);
+    const vk::ShaderStageFlags flags = MaxwellToVK::ShaderStage(stage);
 
     u32 binding = base_binding;
-    const auto AddBindings = [&](vk::DescriptorType descriptor_type, std::size_t num_entries) {
-        for (std::size_t i = 0; i < num_entries; ++i) {
-            bindings.emplace_back(binding++, descriptor_type, 1, stage_flags, nullptr);
-        }
-    };
-    AddBindings(vk::DescriptorType::eUniformBuffer, entries.const_buffers.size());
-    AddBindings(vk::DescriptorType::eStorageBuffer, entries.global_buffers.size());
-    AddBindings(vk::DescriptorType::eUniformTexelBuffer, entries.texel_buffers.size());
-    AddBindings(vk::DescriptorType::eCombinedImageSampler, entries.samplers.size());
-    AddBindings(vk::DescriptorType::eStorageImage, entries.images.size());
+    AddBindings<eUniformBuffer>(bindings, binding, flags, entries.const_buffers);
+    AddBindings<eStorageBuffer>(bindings, binding, flags, entries.global_buffers);
+    AddBindings<eUniformTexelBuffer>(bindings, binding, flags, entries.texel_buffers);
+    AddBindings<eCombinedImageSampler>(bindings, binding, flags, entries.samplers);
+    AddBindings<eStorageImage>(bindings, binding, flags, entries.images);
     return binding;
 }
 
@@ -172,11 +188,6 @@ VKPipelineCache::~VKPipelineCache() = default;
 
 std::array<Shader, Maxwell::MaxShaderProgram> VKPipelineCache::GetShaders() {
     const auto& gpu = system.GPU().Maxwell3D();
-    auto& dirty = system.GPU().Maxwell3D().dirty.shaders;
-    if (!dirty) {
-        return last_shaders;
-    }
-    dirty = false;
 
     std::array<Shader, Maxwell::MaxShaderProgram> shaders;
     for (std::size_t index = 0; index < Maxwell::MaxShaderProgram; ++index) {
@@ -361,32 +372,45 @@ VKPipelineCache::DecompileShaders(const GraphicsPipelineCacheKey& key) {
     return {std::move(program), std::move(bindings)};
 }
 
-void FillDescriptorUpdateTemplateEntries(
-    const VKDevice& device, const ShaderEntries& entries, u32& binding, u32& offset,
-    std::vector<vk::DescriptorUpdateTemplateEntry>& template_entries) {
-    static constexpr auto entry_size = static_cast<u32>(sizeof(DescriptorUpdateEntry));
-    const auto AddEntry = [&](vk::DescriptorType descriptor_type, std::size_t count_) {
-        const u32 count = static_cast<u32>(count_);
-        if (descriptor_type == vk::DescriptorType::eUniformTexelBuffer &&
-            device.GetDriverID() == vk::DriverIdKHR::eNvidiaProprietary) {
-            // Nvidia has a bug where updating multiple uniform texels at once causes the driver to
-            // crash.
-            for (u32 i = 0; i < count; ++i) {
-                template_entries.emplace_back(binding + i, 0, 1, descriptor_type,
-                                              offset + i * entry_size, entry_size);
-            }
-        } else if (count != 0) {
-            template_entries.emplace_back(binding, 0, count, descriptor_type, offset, entry_size);
-        }
-        offset += count * entry_size;
-        binding += count;
-    };
+template <vk::DescriptorType descriptor_type, class Container>
+void AddEntry(std::vector<vk::DescriptorUpdateTemplateEntry>& template_entries, u32& binding,
+              u32& offset, const Container& container) {
+    static constexpr u32 entry_size = static_cast<u32>(sizeof(DescriptorUpdateEntry));
+    const u32 count = static_cast<u32>(std::size(container));
 
-    AddEntry(vk::DescriptorType::eUniformBuffer, entries.const_buffers.size());
-    AddEntry(vk::DescriptorType::eStorageBuffer, entries.global_buffers.size());
-    AddEntry(vk::DescriptorType::eUniformTexelBuffer, entries.texel_buffers.size());
-    AddEntry(vk::DescriptorType::eCombinedImageSampler, entries.samplers.size());
-    AddEntry(vk::DescriptorType::eStorageImage, entries.images.size());
+    if constexpr (descriptor_type == eCombinedImageSampler) {
+        for (u32 i = 0; i < count; ++i) {
+            const u32 num_samplers = container[i].Size();
+            template_entries.emplace_back(binding, 0, num_samplers, descriptor_type, offset,
+                                          entry_size);
+            ++binding;
+            offset += num_samplers * entry_size;
+        }
+        return;
+    }
+
+    if constexpr (descriptor_type == eUniformTexelBuffer) {
+        // Nvidia has a bug where updating multiple uniform texels at once causes the driver to
+        // crash.
+        for (u32 i = 0; i < count; ++i) {
+            template_entries.emplace_back(binding + i, 0, 1, descriptor_type,
+                                          offset + i * entry_size, entry_size);
+        }
+    } else if (count > 0) {
+        template_entries.emplace_back(binding, 0, count, descriptor_type, offset, entry_size);
+    }
+    offset += count * entry_size;
+    binding += count;
+}
+
+void FillDescriptorUpdateTemplateEntries(
+    const ShaderEntries& entries, u32& binding, u32& offset,
+    std::vector<vk::DescriptorUpdateTemplateEntry>& template_entries) {
+    AddEntry<eUniformBuffer>(template_entries, offset, binding, entries.const_buffers);
+    AddEntry<eStorageBuffer>(template_entries, offset, binding, entries.global_buffers);
+    AddEntry<eUniformTexelBuffer>(template_entries, offset, binding, entries.texel_buffers);
+    AddEntry<eCombinedImageSampler>(template_entries, offset, binding, entries.samplers);
+    AddEntry<eStorageImage>(template_entries, offset, binding, entries.images);
 }
 
 } // namespace Vulkan

src/video_core/renderer_vulkan/vk_pipeline_cache.h

@@ -194,7 +194,7 @@ private:
 };
 
 void FillDescriptorUpdateTemplateEntries(
-    const VKDevice& device, const ShaderEntries& entries, u32& binding, u32& offset,
+    const ShaderEntries& entries, u32& binding, u32& offset,
     std::vector<vk::DescriptorUpdateTemplateEntry>& template_entries);
 
 } // namespace Vulkan

src/video_core/renderer_vulkan/vk_rasterizer.cpp

@@ -36,6 +36,7 @@
 #include "video_core/renderer_vulkan/vk_sampler_cache.h"
 #include "video_core/renderer_vulkan/vk_scheduler.h"
 #include "video_core/renderer_vulkan/vk_staging_buffer_pool.h"
+#include "video_core/renderer_vulkan/vk_state_tracker.h"
 #include "video_core/renderer_vulkan/vk_texture_cache.h"
 #include "video_core/renderer_vulkan/vk_update_descriptor.h"
 
@@ -105,17 +106,20 @@ void TransitionImages(const std::vector<ImageView>& views, vk::PipelineStageFlag
 
 template <typename Engine, typename Entry>
 Tegra::Texture::FullTextureInfo GetTextureInfo(const Engine& engine, const Entry& entry,
-                                               std::size_t stage) {
+                                               std::size_t stage, std::size_t index = 0) {
     const auto stage_type = static_cast<Tegra::Engines::ShaderType>(stage);
     if (entry.IsBindless()) {
         const Tegra::Texture::TextureHandle tex_handle =
             engine.AccessConstBuffer32(stage_type, entry.GetBuffer(), entry.GetOffset());
         return engine.GetTextureInfo(tex_handle);
     }
+    const auto& gpu_profile = engine.AccessGuestDriverProfile();
+    const u32 entry_offset = static_cast<u32>(index * gpu_profile.GetTextureHandlerSize());
+    const u32 offset = entry.GetOffset() + entry_offset;
     if constexpr (std::is_same_v<Engine, Tegra::Engines::Maxwell3D>) {
-        return engine.GetStageTexture(stage_type, entry.GetOffset());
+        return engine.GetStageTexture(stage_type, offset);
     } else {
-        return engine.GetTexture(entry.GetOffset());
+        return engine.GetTexture(offset);
     }
 }
 
@@ -277,10 +281,11 @@ void RasterizerVulkan::DrawParameters::Draw(vk::CommandBuffer cmdbuf,
 RasterizerVulkan::RasterizerVulkan(Core::System& system, Core::Frontend::EmuWindow& renderer,
                                    VKScreenInfo& screen_info, const VKDevice& device,
                                    VKResourceManager& resource_manager,
-                                   VKMemoryManager& memory_manager, VKScheduler& scheduler)
+                                   VKMemoryManager& memory_manager, StateTracker& state_tracker,
+                                   VKScheduler& scheduler)
     : RasterizerAccelerated{system.Memory()}, system{system}, render_window{renderer},
       screen_info{screen_info}, device{device}, resource_manager{resource_manager},
-      memory_manager{memory_manager}, scheduler{scheduler},
+      memory_manager{memory_manager}, state_tracker{state_tracker}, scheduler{scheduler},
       staging_pool(device, memory_manager, scheduler), descriptor_pool(device),
       update_descriptor_queue(device, scheduler),
       quad_array_pass(device, scheduler, descriptor_pool, staging_pool, update_descriptor_queue),
@@ -545,6 +550,10 @@ bool RasterizerVulkan::AccelerateDisplay(const Tegra::FramebufferConfig& config,
     return true;
 }
 
+void RasterizerVulkan::SetupDirtyFlags() {
+    state_tracker.Initialize();
+}
+
 void RasterizerVulkan::FlushWork() {
     static constexpr u32 DRAWS_TO_DISPATCH = 4096;
 
@@ -568,9 +577,9 @@ void RasterizerVulkan::FlushWork() {
 
 RasterizerVulkan::Texceptions RasterizerVulkan::UpdateAttachments() {
     MICROPROFILE_SCOPE(Vulkan_RenderTargets);
-    auto& dirty = system.GPU().Maxwell3D().dirty;
-    const bool update_rendertargets = dirty.render_settings;
-    dirty.render_settings = false;
+    auto& dirty = system.GPU().Maxwell3D().dirty.flags;
+    const bool update_rendertargets = dirty[VideoCommon::Dirty::RenderTargets];
+    dirty[VideoCommon::Dirty::RenderTargets] = false;
 
     texture_cache.GuardRenderTargets(true);
 
@@ -720,13 +729,13 @@ void RasterizerVulkan::SetupImageTransitions(
 }
 
 void RasterizerVulkan::UpdateDynamicStates() {
-    auto& gpu = system.GPU().Maxwell3D();
-    UpdateViewportsState(gpu);
-    UpdateScissorsState(gpu);
-    UpdateDepthBias(gpu);
-    UpdateBlendConstants(gpu);
-    UpdateDepthBounds(gpu);
-    UpdateStencilFaces(gpu);
+    auto& regs = system.GPU().Maxwell3D().regs;
+    UpdateViewportsState(regs);
+    UpdateScissorsState(regs);
+    UpdateDepthBias(regs);
+    UpdateBlendConstants(regs);
+    UpdateDepthBounds(regs);
+    UpdateStencilFaces(regs);
 }
 
 void RasterizerVulkan::SetupVertexArrays(FixedPipelineState::VertexInput& vertex_input,
@@ -836,8 +845,10 @@ void RasterizerVulkan::SetupGraphicsTextures(const ShaderEntries& entries, std::
     MICROPROFILE_SCOPE(Vulkan_Textures);
     const auto& gpu = system.GPU().Maxwell3D();
     for (const auto& entry : entries.samplers) {
-        const auto texture = GetTextureInfo(gpu, entry, stage);
-        SetupTexture(texture, entry);
+        for (std::size_t i = 0; i < entry.Size(); ++i) {
+            const auto texture = GetTextureInfo(gpu, entry, stage, i);
+            SetupTexture(texture, entry);
+        }
     }
 }
 
@@ -886,8 +897,10 @@ void RasterizerVulkan::SetupComputeTextures(const ShaderEntries& entries) {
     MICROPROFILE_SCOPE(Vulkan_Textures);
     const auto& gpu = system.GPU().KeplerCompute();
     for (const auto& entry : entries.samplers) {
-        const auto texture = GetTextureInfo(gpu, entry, ComputeShaderIndex);
-        SetupTexture(texture, entry);
+        for (std::size_t i = 0; i < entry.Size(); ++i) {
+            const auto texture = GetTextureInfo(gpu, entry, ComputeShaderIndex, i);
+            SetupTexture(texture, entry);
+        }
     }
 }
 
@@ -972,12 +985,10 @@ void RasterizerVulkan::SetupImage(const Tegra::Texture::TICEntry& tic, const Ima
     image_views.push_back(ImageView{std::move(view), image_layout});
 }
 
-void RasterizerVulkan::UpdateViewportsState(Tegra::Engines::Maxwell3D& gpu) {
-    if (!gpu.dirty.viewport_transform && scheduler.TouchViewports()) {
+void RasterizerVulkan::UpdateViewportsState(Tegra::Engines::Maxwell3D::Regs& regs) {
+    if (!state_tracker.TouchViewports()) {
         return;
     }
-    gpu.dirty.viewport_transform = false;
-    const auto& regs = gpu.regs;
     const std::array viewports{
         GetViewportState(device, regs, 0),  GetViewportState(device, regs, 1),
         GetViewportState(device, regs, 2),  GetViewportState(device, regs, 3),
@@ -992,12 +1003,10 @@ void RasterizerVulkan::UpdateViewportsState(Tegra::Engines::Maxwell3D& gpu) {
     });
 }
 
-void RasterizerVulkan::UpdateScissorsState(Tegra::Engines::Maxwell3D& gpu) {
-    if (!gpu.dirty.scissor_test && scheduler.TouchScissors()) {
+void RasterizerVulkan::UpdateScissorsState(Tegra::Engines::Maxwell3D::Regs& regs) {
+    if (!state_tracker.TouchScissors()) {
         return;
     }
-    gpu.dirty.scissor_test = false;
-    const auto& regs = gpu.regs;
     const std::array scissors = {
         GetScissorState(regs, 0),  GetScissorState(regs, 1),  GetScissorState(regs, 2),
         GetScissorState(regs, 3),  GetScissorState(regs, 4),  GetScissorState(regs, 5),
@@ -1010,46 +1019,39 @@ void RasterizerVulkan::UpdateScissorsState(Tegra::Engines::Maxwell3D& gpu) {
     });
 }
 
-void RasterizerVulkan::UpdateDepthBias(Tegra::Engines::Maxwell3D& gpu) {
-    if (!gpu.dirty.polygon_offset && scheduler.TouchDepthBias()) {
+void RasterizerVulkan::UpdateDepthBias(Tegra::Engines::Maxwell3D::Regs& regs) {
+    if (!state_tracker.TouchDepthBias()) {
         return;
     }
-    gpu.dirty.polygon_offset = false;
-    const auto& regs = gpu.regs;
     scheduler.Record([constant = regs.polygon_offset_units, clamp = regs.polygon_offset_clamp,
                       factor = regs.polygon_offset_factor](auto cmdbuf, auto& dld) {
         cmdbuf.setDepthBias(constant, clamp, factor / 2.0f, dld);
     });
 }
 
-void RasterizerVulkan::UpdateBlendConstants(Tegra::Engines::Maxwell3D& gpu) {
-    if (!gpu.dirty.blend_state && scheduler.TouchBlendConstants()) {
+void RasterizerVulkan::UpdateBlendConstants(Tegra::Engines::Maxwell3D::Regs& regs) {
+    if (!state_tracker.TouchBlendConstants()) {
         return;
     }
-    gpu.dirty.blend_state = false;
-    const std::array blend_color = {gpu.regs.blend_color.r, gpu.regs.blend_color.g,
-                                    gpu.regs.blend_color.b, gpu.regs.blend_color.a};
+    const std::array blend_color = {regs.blend_color.r, regs.blend_color.g, regs.blend_color.b,
+                                    regs.blend_color.a};
     scheduler.Record([blend_color](auto cmdbuf, auto& dld) {
         cmdbuf.setBlendConstants(blend_color.data(), dld);
     });
 }
 
-void RasterizerVulkan::UpdateDepthBounds(Tegra::Engines::Maxwell3D& gpu) {
-    if (!gpu.dirty.depth_bounds_values && scheduler.TouchDepthBounds()) {
+void RasterizerVulkan::UpdateDepthBounds(Tegra::Engines::Maxwell3D::Regs& regs) {
+    if (!state_tracker.TouchDepthBounds()) {
         return;
     }
-    gpu.dirty.depth_bounds_values = false;
-    const auto& regs = gpu.regs;
     scheduler.Record([min = regs.depth_bounds[0], max = regs.depth_bounds[1]](
                          auto cmdbuf, auto& dld) { cmdbuf.setDepthBounds(min, max, dld); });
 }
 
-void RasterizerVulkan::UpdateStencilFaces(Tegra::Engines::Maxwell3D& gpu) {
-    if (!gpu.dirty.stencil_test && scheduler.TouchStencilValues()) {
+void RasterizerVulkan::UpdateStencilFaces(Tegra::Engines::Maxwell3D::Regs& regs) {
+    if (!state_tracker.TouchStencilProperties()) {
         return;
     }
-    gpu.dirty.stencil_test = false;
-    const auto& regs = gpu.regs;
     if (regs.stencil_two_side_enable) {
         // Separate values per face
         scheduler.Record(

src/video_core/renderer_vulkan/vk_rasterizer.h

@@ -96,6 +96,7 @@ struct hash<Vulkan::FramebufferCacheKey> {
 
 namespace Vulkan {
 
+class StateTracker;
 class BufferBindings;
 
 struct ImageView {
@@ -108,7 +109,7 @@ public:
     explicit RasterizerVulkan(Core::System& system, Core::Frontend::EmuWindow& render_window,
                               VKScreenInfo& screen_info, const VKDevice& device,
                               VKResourceManager& resource_manager, VKMemoryManager& memory_manager,
-                              VKScheduler& scheduler);
+                              StateTracker& state_tracker, VKScheduler& scheduler);
     ~RasterizerVulkan() override;
 
     void Draw(bool is_indexed, bool is_instanced) override;
@@ -127,6 +128,7 @@ public:
                                const Tegra::Engines::Fermi2D::Config& copy_config) override;
     bool AccelerateDisplay(const Tegra::FramebufferConfig& config, VAddr framebuffer_addr,
                            u32 pixel_stride) override;
+    void SetupDirtyFlags() override;
 
     /// Maximum supported size that a constbuffer can have in bytes.
     static constexpr std::size_t MaxConstbufferSize = 0x10000;
@@ -215,12 +217,12 @@ private:
 
     void SetupImage(const Tegra::Texture::TICEntry& tic, const ImageEntry& entry);
 
-    void UpdateViewportsState(Tegra::Engines::Maxwell3D& gpu);
-    void UpdateScissorsState(Tegra::Engines::Maxwell3D& gpu);
-    void UpdateDepthBias(Tegra::Engines::Maxwell3D& gpu);
-    void UpdateBlendConstants(Tegra::Engines::Maxwell3D& gpu);
-    void UpdateDepthBounds(Tegra::Engines::Maxwell3D& gpu);
-    void UpdateStencilFaces(Tegra::Engines::Maxwell3D& gpu);
+    void UpdateViewportsState(Tegra::Engines::Maxwell3D::Regs& regs);
+    void UpdateScissorsState(Tegra::Engines::Maxwell3D::Regs& regs);
+    void UpdateDepthBias(Tegra::Engines::Maxwell3D::Regs& regs);
+    void UpdateBlendConstants(Tegra::Engines::Maxwell3D::Regs& regs);
+    void UpdateDepthBounds(Tegra::Engines::Maxwell3D::Regs& regs);
+    void UpdateStencilFaces(Tegra::Engines::Maxwell3D::Regs& regs);
 
     std::size_t CalculateGraphicsStreamBufferSize(bool is_indexed) const;
 
@@ -241,6 +243,7 @@ private:
     const VKDevice& device;
     VKResourceManager& resource_manager;
     VKMemoryManager& memory_manager;
+    StateTracker& state_tracker;
     VKScheduler& scheduler;
 
     VKStagingBufferPool staging_pool;

src/video_core/renderer_vulkan/vk_scheduler.cpp

@@ -2,6 +2,12 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <memory>
+#include <mutex>
+#include <optional>
+#include <thread>
+#include <utility>
+
 #include "common/assert.h"
 #include "common/microprofile.h"
 #include "video_core/renderer_vulkan/declarations.h"
@@ -9,6 +15,7 @@
 #include "video_core/renderer_vulkan/vk_query_cache.h"
 #include "video_core/renderer_vulkan/vk_resource_manager.h"
 #include "video_core/renderer_vulkan/vk_scheduler.h"
+#include "video_core/renderer_vulkan/vk_state_tracker.h"
 
 namespace Vulkan {
 
@@ -29,9 +36,10 @@ void VKScheduler::CommandChunk::ExecuteAll(vk::CommandBuffer cmdbuf,
     last = nullptr;
 }
 
-VKScheduler::VKScheduler(const VKDevice& device, VKResourceManager& resource_manager)
-    : device{device}, resource_manager{resource_manager}, next_fence{
-                                                              &resource_manager.CommitFence()} {
+VKScheduler::VKScheduler(const VKDevice& device, VKResourceManager& resource_manager,
+                         StateTracker& state_tracker)
+    : device{device}, resource_manager{resource_manager}, state_tracker{state_tracker},
+      next_fence{&resource_manager.CommitFence()} {
     AcquireNewChunk();
     AllocateNewContext();
     worker_thread = std::thread(&VKScheduler::WorkerThread, this);
@@ -157,12 +165,7 @@ void VKScheduler::AllocateNewContext() {
 
 void VKScheduler::InvalidateState() {
     state.graphics_pipeline = nullptr;
-    state.viewports = false;
-    state.scissors = false;
-    state.depth_bias = false;
-    state.blend_constants = false;
-    state.depth_bounds = false;
-    state.stencil_values = false;
+    state_tracker.InvalidateCommandBufferState();
 }
 
 void VKScheduler::EndPendingOperations() {

src/video_core/renderer_vulkan/vk_scheduler.h

@@ -17,6 +17,7 @@
 
 namespace Vulkan {
 
+class StateTracker;
 class VKDevice;
 class VKFence;
 class VKQueryCache;
@@ -43,7 +44,8 @@ private:
 /// OpenGL-like operations on Vulkan command buffers.
 class VKScheduler {
 public:
-    explicit VKScheduler(const VKDevice& device, VKResourceManager& resource_manager);
+    explicit VKScheduler(const VKDevice& device, VKResourceManager& resource_manager,
+                         StateTracker& state_tracker);
     ~VKScheduler();
 
     /// Sends the current execution context to the GPU.
@@ -74,36 +76,6 @@ public:
         query_cache = &query_cache_;
     }
 
-    /// Returns true when viewports have been set in the current command buffer.
-    bool TouchViewports() {
-        return std::exchange(state.viewports, true);
-    }
-
-    /// Returns true when scissors have been set in the current command buffer.
-    bool TouchScissors() {
-        return std::exchange(state.scissors, true);
-    }
-
-    /// Returns true when depth bias have been set in the current command buffer.
-    bool TouchDepthBias() {
-        return std::exchange(state.depth_bias, true);
-    }
-
-    /// Returns true when blend constants have been set in the current command buffer.
-    bool TouchBlendConstants() {
-        return std::exchange(state.blend_constants, true);
-    }
-
-    /// Returns true when depth bounds have been set in the current command buffer.
-    bool TouchDepthBounds() {
-        return std::exchange(state.depth_bounds, true);
-    }
-
-    /// Returns true when stencil values have been set in the current command buffer.
-    bool TouchStencilValues() {
-        return std::exchange(state.stencil_values, true);
-    }
-
     /// Send work to a separate thread.
     template <typename T>
     void Record(T&& command) {
@@ -217,6 +189,8 @@ private:
 
     const VKDevice& device;
     VKResourceManager& resource_manager;
+    StateTracker& state_tracker;
+
     VKQueryCache* query_cache = nullptr;
 
     vk::CommandBuffer current_cmdbuf;
@@ -226,12 +200,6 @@ private:
     struct State {
         std::optional<vk::RenderPassBeginInfo> renderpass;
         vk::Pipeline graphics_pipeline;
-        bool viewports = false;
-        bool scissors = false;
-        bool depth_bias = false;
-        bool blend_constants = false;
-        bool depth_bounds = false;
-        bool stencil_values = false;
     } state;
 
     std::unique_ptr<CommandChunk> chunk;

src/video_core/renderer_vulkan/vk_shader_decompiler.cpp

@@ -69,8 +69,9 @@ struct TexelBuffer {
 
 struct SampledImage {
     Id image_type{};
-    Id sampled_image_type{};
-    Id sampler{};
+    Id sampler_type{};
+    Id sampler_pointer_type{};
+    Id variable{};
 };
 
 struct StorageImage {
@@ -833,16 +834,20 @@ private:
             constexpr int sampled = 1;
             constexpr auto format = spv::ImageFormat::Unknown;
             const Id image_type = TypeImage(t_float, dim, depth, arrayed, ms, sampled, format);
-            const Id sampled_image_type = TypeSampledImage(image_type);
-            const Id pointer_type =
-                TypePointer(spv::StorageClass::UniformConstant, sampled_image_type);
+            const Id sampler_type = TypeSampledImage(image_type);
+            const Id sampler_pointer_type =
+                TypePointer(spv::StorageClass::UniformConstant, sampler_type);
+            const Id type = sampler.IsIndexed()
+                                ? TypeArray(sampler_type, Constant(t_uint, sampler.Size()))
+                                : sampler_type;
+            const Id pointer_type = TypePointer(spv::StorageClass::UniformConstant, type);
             const Id id = OpVariable(pointer_type, spv::StorageClass::UniformConstant);
             AddGlobalVariable(Name(id, fmt::format("sampler_{}", sampler.GetIndex())));
             Decorate(id, spv::Decoration::Binding, binding++);
             Decorate(id, spv::Decoration::DescriptorSet, DESCRIPTOR_SET);
 
-            sampled_images.emplace(sampler.GetIndex(),
-                                   SampledImage{image_type, sampled_image_type, id});
+            sampled_images.emplace(sampler.GetIndex(), SampledImage{image_type, sampler_type,
+                                                                    sampler_pointer_type, id});
         }
         return binding;
     }
@@ -1525,7 +1530,12 @@ private:
         ASSERT(!meta.sampler.IsBuffer());
 
         const auto& entry = sampled_images.at(meta.sampler.GetIndex());
-        return OpLoad(entry.sampled_image_type, entry.sampler);
+        Id sampler = entry.variable;
+        if (meta.sampler.IsIndexed()) {
+            const Id index = AsInt(Visit(meta.index));
+            sampler = OpAccessChain(entry.sampler_pointer_type, sampler, index);
+        }
+        return OpLoad(entry.sampler_type, sampler);
     }
 
     Id GetTextureImage(Operation operation) {
@@ -2211,16 +2221,14 @@ private:
         switch (specialization.attribute_types.at(location)) {
         case Maxwell::VertexAttribute::Type::SignedNorm:
         case Maxwell::VertexAttribute::Type::UnsignedNorm:
+        case Maxwell::VertexAttribute::Type::UnsignedScaled:
+        case Maxwell::VertexAttribute::Type::SignedScaled:
         case Maxwell::VertexAttribute::Type::Float:
             return {Type::Float, t_in_float, t_in_float4};
         case Maxwell::VertexAttribute::Type::SignedInt:
             return {Type::Int, t_in_int, t_in_int4};
         case Maxwell::VertexAttribute::Type::UnsignedInt:
             return {Type::Uint, t_in_uint, t_in_uint4};
-        case Maxwell::VertexAttribute::Type::UnsignedScaled:
-        case Maxwell::VertexAttribute::Type::SignedScaled:
-            UNIMPLEMENTED();
-            return {Type::Float, t_in_float, t_in_float4};
         default:
             UNREACHABLE();
             return {Type::Float, t_in_float, t_in_float4};

src/video_core/renderer_vulkan/vk_state_tracker.cpp

@@ -0,0 +1,101 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include <cstddef>
+#include <iterator>
+
+#include "common/common_types.h"
+#include "core/core.h"
+#include "video_core/dirty_flags.h"
+#include "video_core/engines/maxwell_3d.h"
+#include "video_core/gpu.h"
+#include "video_core/renderer_vulkan/vk_state_tracker.h"
+
+#define OFF(field_name) MAXWELL3D_REG_INDEX(field_name)
+#define NUM(field_name) (sizeof(Maxwell3D::Regs::field_name) / sizeof(u32))
+
+namespace Vulkan {
+
+namespace {
+
+using namespace Dirty;
+using namespace VideoCommon::Dirty;
+using Tegra::Engines::Maxwell3D;
+using Regs = Maxwell3D::Regs;
+using Tables = Maxwell3D::DirtyState::Tables;
+using Table = Maxwell3D::DirtyState::Table;
+using Flags = Maxwell3D::DirtyState::Flags;
+
+Flags MakeInvalidationFlags() {
+    Flags flags{};
+    flags[Viewports] = true;
+    flags[Scissors] = true;
+    flags[DepthBias] = true;
+    flags[BlendConstants] = true;
+    flags[DepthBounds] = true;
+    flags[StencilProperties] = true;
+    return flags;
+}
+
+void SetupDirtyViewports(Tables& tables) {
+    FillBlock(tables[0], OFF(viewport_transform), NUM(viewport_transform), Viewports);
+    FillBlock(tables[0], OFF(viewports), NUM(viewports), Viewports);
+    tables[0][OFF(viewport_transform_enabled)] = Viewports;
+}
+
+void SetupDirtyScissors(Tables& tables) {
+    FillBlock(tables[0], OFF(scissor_test), NUM(scissor_test), Scissors);
+}
+
+void SetupDirtyDepthBias(Tables& tables) {
+    auto& table = tables[0];
+    table[OFF(polygon_offset_units)] = DepthBias;
+    table[OFF(polygon_offset_clamp)] = DepthBias;
+    table[OFF(polygon_offset_factor)] = DepthBias;
+}
+
+void SetupDirtyBlendConstants(Tables& tables) {
+    FillBlock(tables[0], OFF(blend_color), NUM(blend_color), BlendConstants);
+}
+
+void SetupDirtyDepthBounds(Tables& tables) {
+    FillBlock(tables[0], OFF(depth_bounds), NUM(depth_bounds), DepthBounds);
+}
+
+void SetupDirtyStencilProperties(Tables& tables) {
+    auto& table = tables[0];
+    table[OFF(stencil_two_side_enable)] = StencilProperties;
+    table[OFF(stencil_front_func_ref)] = StencilProperties;
+    table[OFF(stencil_front_mask)] = StencilProperties;
+    table[OFF(stencil_front_func_mask)] = StencilProperties;
+    table[OFF(stencil_back_func_ref)] = StencilProperties;
+    table[OFF(stencil_back_mask)] = StencilProperties;
+    table[OFF(stencil_back_func_mask)] = StencilProperties;
+}
+
+} // Anonymous namespace
+
+StateTracker::StateTracker(Core::System& system)
+    : system{system}, invalidation_flags{MakeInvalidationFlags()} {}
+
+void StateTracker::Initialize() {
+    auto& dirty = system.GPU().Maxwell3D().dirty;
+    auto& tables = dirty.tables;
+    SetupDirtyRenderTargets(tables);
+    SetupDirtyViewports(tables);
+    SetupDirtyScissors(tables);
+    SetupDirtyDepthBias(tables);
+    SetupDirtyBlendConstants(tables);
+    SetupDirtyDepthBounds(tables);
+    SetupDirtyStencilProperties(tables);
+
+    SetupCommonOnWriteStores(dirty.on_write_stores);
+}
+
+void StateTracker::InvalidateCommandBufferState() {
+    system.GPU().Maxwell3D().dirty.flags |= invalidation_flags;
+}
+
+} // namespace Vulkan