service/service: Remove unncessary calls to c_str()

These can just be passed regularly, now that we use fmt instead of our
old logging system.

While we're at it, make the parameters to MakeFunctionString
std::string_views.

.travis.yml

@@ -24,7 +24,7 @@ matrix:
     - os: osx
       env: NAME="macos build"
       sudo: false
-      osx_image: xcode10
+      osx_image: xcode10.1
       install: "./.travis/macos/deps.sh"
       script: "./.travis/macos/build.sh"
       after_success: "./.travis/macos/upload.sh"

.travis/macos/build.sh

@@ -2,7 +2,7 @@
 
 set -o pipefail
 
-export MACOSX_DEPLOYMENT_TARGET=10.13
+export MACOSX_DEPLOYMENT_TARGET=10.14
 export Qt5_DIR=$(brew --prefix)/opt/qt5
 export UNICORNDIR=$(pwd)/externals/unicorn
 export PATH="/usr/local/opt/ccache/libexec:$PATH"

CMakeModules/GenerateSCMRev.cmake

@@ -73,6 +73,7 @@ set(HASH_FILES
     "${VIDEO_CORE}/shader/decode/integer_set.cpp"
     "${VIDEO_CORE}/shader/decode/integer_set_predicate.cpp"
     "${VIDEO_CORE}/shader/decode/memory.cpp"
+    "${VIDEO_CORE}/shader/decode/texture.cpp"
     "${VIDEO_CORE}/shader/decode/other.cpp"
     "${VIDEO_CORE}/shader/decode/predicate_set_predicate.cpp"
     "${VIDEO_CORE}/shader/decode/predicate_set_register.cpp"

src/audio_core/cubeb_sink.cpp

@@ -12,7 +12,7 @@
 #include "common/ring_buffer.h"
 #include "core/settings.h"
 
-#ifdef _MSC_VER
+#ifdef _WIN32
 #include <objbase.h>
 #endif
 
@@ -113,7 +113,7 @@ private:
 
 CubebSink::CubebSink(std::string_view target_device_name) {
     // Cubeb requires COM to be initialized on the thread calling cubeb_init on Windows
-#ifdef _MSC_VER
+#ifdef _WIN32
     com_init_result = CoInitializeEx(nullptr, COINIT_MULTITHREADED);
 #endif
 
@@ -152,7 +152,7 @@ CubebSink::~CubebSink() {
 
     cubeb_destroy(ctx);
 
-#ifdef _MSC_VER
+#ifdef _WIN32
     if (SUCCEEDED(com_init_result)) {
         CoUninitialize();
     }

src/audio_core/cubeb_sink.h

@@ -26,7 +26,7 @@ private:
     cubeb_devid output_device{};
     std::vector<SinkStreamPtr> sink_streams;
 
-#ifdef _MSC_VER
+#ifdef _WIN32
     u32 com_init_result = 0;
 #endif
 };

src/common/CMakeLists.txt

@@ -47,6 +47,7 @@ add_custom_command(OUTPUT scm_rev.cpp
       "${VIDEO_CORE}/shader/decode/integer_set.cpp"
       "${VIDEO_CORE}/shader/decode/integer_set_predicate.cpp"
       "${VIDEO_CORE}/shader/decode/memory.cpp"
+      "${VIDEO_CORE}/shader/decode/texture.cpp"
       "${VIDEO_CORE}/shader/decode/other.cpp"
       "${VIDEO_CORE}/shader/decode/predicate_set_predicate.cpp"
       "${VIDEO_CORE}/shader/decode/predicate_set_register.cpp"

src/common/bit_field.h

@@ -111,12 +111,6 @@
 template <std::size_t Position, std::size_t Bits, typename T>
 struct BitField {
 private:
-    // We hide the copy assigment operator here, because the default copy
-    // assignment would copy the full storage value, rather than just the bits
-    // relevant to this particular bit field.
-    // We don't delete it because we want BitField to be trivially copyable.
-    constexpr BitField& operator=(const BitField&) = default;
-
     // UnderlyingType is T for non-enum types and the underlying type of T if
     // T is an enumeration. Note that T is wrapped within an enable_if in the
     // former case to workaround compile errors which arise when using
@@ -163,9 +157,13 @@ public:
     BitField(T val) = delete;
     BitField& operator=(T val) = delete;
 
-    // Force default constructor to be created
-    // so that we can use this within unions
-    constexpr BitField() = default;
+    constexpr BitField() noexcept = default;
+
+    constexpr BitField(const BitField&) noexcept = default;
+    constexpr BitField& operator=(const BitField&) noexcept = default;
+
+    constexpr BitField(BitField&&) noexcept = default;
+    constexpr BitField& operator=(BitField&&) noexcept = default;
 
     constexpr FORCE_INLINE operator T() const {
         return Value();

src/core/CMakeLists.txt

@@ -217,6 +217,7 @@ add_library(core STATIC
     hle/service/audio/audren_u.h
     hle/service/audio/codecctl.cpp
     hle/service/audio/codecctl.h
+    hle/service/audio/errors.h
     hle/service/audio/hwopus.cpp
     hle/service/audio/hwopus.h
     hle/service/bcat/bcat.cpp

src/core/core.cpp

@@ -36,7 +36,8 @@
 #include "frontend/applets/software_keyboard.h"
 #include "frontend/applets/web_browser.h"
 #include "video_core/debug_utils/debug_utils.h"
-#include "video_core/gpu.h"
+#include "video_core/gpu_asynch.h"
+#include "video_core/gpu_synch.h"
 #include "video_core/renderer_base.h"
 #include "video_core/video_core.h"
 
@@ -78,6 +79,7 @@ FileSys::VirtualFile GetGameFileFromPath(const FileSys::VirtualFilesystem& vfs,
     return vfs->OpenFile(path, FileSys::Mode::Read);
 }
 struct System::Impl {
+    explicit Impl(System& system) : kernel{system} {}
 
     Cpu& CurrentCpuCore() {
         return cpu_core_manager.GetCurrentCore();
@@ -95,7 +97,7 @@ struct System::Impl {
         LOG_DEBUG(HW_Memory, "initialized OK");
 
         core_timing.Initialize();
-        kernel.Initialize(core_timing);
+        kernel.Initialize();
 
         const auto current_time = std::chrono::duration_cast<std::chrono::seconds>(
             std::chrono::system_clock::now().time_since_epoch());
@@ -128,10 +130,16 @@ struct System::Impl {
             return ResultStatus::ErrorVideoCore;
         }
 
-        gpu_core = std::make_unique<Tegra::GPU>(system, renderer->Rasterizer());
+        is_powered_on = true;
+
+        if (Settings::values.use_asynchronous_gpu_emulation) {
+            gpu_core = std::make_unique<VideoCommon::GPUAsynch>(system, *renderer);
+        } else {
+            gpu_core = std::make_unique<VideoCommon::GPUSynch>(system, *renderer);
+        }
 
         cpu_core_manager.Initialize(system);
-        is_powered_on = true;
+
         LOG_DEBUG(Core, "Initialized OK");
 
         // Reset counters and set time origin to current frame
@@ -182,13 +190,13 @@ struct System::Impl {
 
     void Shutdown() {
         // Log last frame performance stats
-        auto perf_results = GetAndResetPerfStats();
-        Telemetry().AddField(Telemetry::FieldType::Performance, "Shutdown_EmulationSpeed",
-                             perf_results.emulation_speed * 100.0);
-        Telemetry().AddField(Telemetry::FieldType::Performance, "Shutdown_Framerate",
-                             perf_results.game_fps);
-        Telemetry().AddField(Telemetry::FieldType::Performance, "Shutdown_Frametime",
-                             perf_results.frametime * 1000.0);
+        const auto perf_results = GetAndResetPerfStats();
+        telemetry_session->AddField(Telemetry::FieldType::Performance, "Shutdown_EmulationSpeed",
+                                    perf_results.emulation_speed * 100.0);
+        telemetry_session->AddField(Telemetry::FieldType::Performance, "Shutdown_Framerate",
+                                    perf_results.game_fps);
+        telemetry_session->AddField(Telemetry::FieldType::Performance, "Shutdown_Frametime",
+                                    perf_results.frametime * 1000.0);
 
         is_powered_on = false;
 
@@ -265,7 +273,7 @@ struct System::Impl {
     Core::FrameLimiter frame_limiter;
 };
 
-System::System() : impl{std::make_unique<Impl>()} {}
+System::System() : impl{std::make_unique<Impl>(*this)} {}
 System::~System() = default;
 
 Cpu& System::CurrentCpuCore() {

src/core/core.h

@@ -293,10 +293,6 @@ inline ARM_Interface& CurrentArmInterface() {
     return System::GetInstance().CurrentArmInterface();
 }
 
-inline TelemetrySession& Telemetry() {
-    return System::GetInstance().TelemetrySession();
-}
-
 inline Kernel::Process* CurrentProcess() {
     return System::GetInstance().CurrentProcess();
 }

src/core/core_cpu.cpp

@@ -11,6 +11,7 @@
 #endif
 #include "core/arm/exclusive_monitor.h"
 #include "core/arm/unicorn/arm_unicorn.h"
+#include "core/core.h"
 #include "core/core_cpu.h"
 #include "core/core_timing.h"
 #include "core/hle/kernel/scheduler.h"
@@ -49,9 +50,9 @@ bool CpuBarrier::Rendezvous() {
     return false;
 }
 
-Cpu::Cpu(Timing::CoreTiming& core_timing, ExclusiveMonitor& exclusive_monitor,
-         CpuBarrier& cpu_barrier, std::size_t core_index)
-    : cpu_barrier{cpu_barrier}, core_timing{core_timing}, core_index{core_index} {
+Cpu::Cpu(System& system, ExclusiveMonitor& exclusive_monitor, CpuBarrier& cpu_barrier,
+         std::size_t core_index)
+    : cpu_barrier{cpu_barrier}, core_timing{system.CoreTiming()}, core_index{core_index} {
     if (Settings::values.use_cpu_jit) {
 #ifdef ARCHITECTURE_x86_64
         arm_interface = std::make_unique<ARM_Dynarmic>(core_timing, exclusive_monitor, core_index);
@@ -63,7 +64,7 @@ Cpu::Cpu(Timing::CoreTiming& core_timing, ExclusiveMonitor& exclusive_monitor,
         arm_interface = std::make_unique<ARM_Unicorn>(core_timing);
     }
 
-    scheduler = std::make_unique<Kernel::Scheduler>(*arm_interface);
+    scheduler = std::make_unique<Kernel::Scheduler>(system, *arm_interface);
 }
 
 Cpu::~Cpu() = default;

src/core/core_cpu.h

@@ -15,6 +15,10 @@ namespace Kernel {
 class Scheduler;
 }
 
+namespace Core {
+class System;
+}
+
 namespace Core::Timing {
 class CoreTiming;
 }
@@ -45,8 +49,8 @@ private:
 
 class Cpu {
 public:
-    Cpu(Timing::CoreTiming& core_timing, ExclusiveMonitor& exclusive_monitor,
-        CpuBarrier& cpu_barrier, std::size_t core_index);
+    Cpu(System& system, ExclusiveMonitor& exclusive_monitor, CpuBarrier& cpu_barrier,
+        std::size_t core_index);
     ~Cpu();
 
     void RunLoop(bool tight_loop = true);

src/core/cpu_core_manager.cpp

@@ -27,8 +27,7 @@ void CpuCoreManager::Initialize(System& system) {
     exclusive_monitor = Cpu::MakeExclusiveMonitor(cores.size());
 
     for (std::size_t index = 0; index < cores.size(); ++index) {
-        cores[index] =
-            std::make_unique<Cpu>(system.CoreTiming(), *exclusive_monitor, *barrier, index);
+        cores[index] = std::make_unique<Cpu>(system, *exclusive_monitor, *barrier, index);
     }
 
     // Create threads for CPU cores 1-3, and build thread_to_cpu map

src/core/hle/ipc.h

@@ -4,10 +4,10 @@
 
 #pragma once
 
+#include "common/bit_field.h"
+#include "common/common_funcs.h"
 #include "common/common_types.h"
 #include "common/swap.h"
-#include "core/hle/kernel/errors.h"
-#include "core/memory.h"
 
 namespace IPC {
 

src/core/hle/ipc_helpers.h

@@ -350,7 +350,7 @@ public:
     template <class T>
     std::shared_ptr<T> PopIpcInterface() {
         ASSERT(context->Session()->IsDomain());
-        ASSERT(context->GetDomainMessageHeader()->input_object_count > 0);
+        ASSERT(context->GetDomainMessageHeader().input_object_count > 0);
         return context->GetDomainRequestHandler<T>(Pop<u32>() - 1);
     }
 };

src/core/hle/kernel/address_arbiter.cpp

@@ -9,6 +9,7 @@
 #include "common/common_types.h"
 #include "core/core.h"
 #include "core/core_cpu.h"
+#include "core/hle/kernel/address_arbiter.h"
 #include "core/hle/kernel/errors.h"
 #include "core/hle/kernel/object.h"
 #include "core/hle/kernel/process.h"
@@ -17,32 +18,144 @@
 #include "core/hle/result.h"
 #include "core/memory.h"
 
-namespace Kernel::AddressArbiter {
+namespace Kernel {
+namespace {
+// Wake up num_to_wake (or all) threads in a vector.
+void WakeThreads(const std::vector<SharedPtr<Thread>>& waiting_threads, s32 num_to_wake) {
+    // Only process up to 'target' threads, unless 'target' is <= 0, in which case process
+    // them all.
+    std::size_t last = waiting_threads.size();
+    if (num_to_wake > 0) {
+        last = num_to_wake;
+    }
 
-// Performs actual address waiting logic.
-static ResultCode WaitForAddress(VAddr address, s64 timeout) {
-    SharedPtr<Thread> current_thread = GetCurrentThread();
+    // Signal the waiting threads.
+    for (std::size_t i = 0; i < last; i++) {
+        ASSERT(waiting_threads[i]->GetStatus() == ThreadStatus::WaitArb);
+        waiting_threads[i]->SetWaitSynchronizationResult(RESULT_SUCCESS);
+        waiting_threads[i]->SetArbiterWaitAddress(0);
+        waiting_threads[i]->ResumeFromWait();
+    }
+}
+} // Anonymous namespace
+
+AddressArbiter::AddressArbiter(Core::System& system) : system{system} {}
+AddressArbiter::~AddressArbiter() = default;
+
+ResultCode AddressArbiter::SignalToAddress(VAddr address, s32 num_to_wake) {
+    const std::vector<SharedPtr<Thread>> waiting_threads = GetThreadsWaitingOnAddress(address);
+    WakeThreads(waiting_threads, num_to_wake);
+    return RESULT_SUCCESS;
+}
+
+ResultCode AddressArbiter::IncrementAndSignalToAddressIfEqual(VAddr address, s32 value,
+                                                              s32 num_to_wake) {
+    // Ensure that we can write to the address.
+    if (!Memory::IsValidVirtualAddress(address)) {
+        return ERR_INVALID_ADDRESS_STATE;
+    }
+
+    if (static_cast<s32>(Memory::Read32(address)) != value) {
+        return ERR_INVALID_STATE;
+    }
+
+    Memory::Write32(address, static_cast<u32>(value + 1));
+    return SignalToAddress(address, num_to_wake);
+}
+
+ResultCode AddressArbiter::ModifyByWaitingCountAndSignalToAddressIfEqual(VAddr address, s32 value,
+                                                                         s32 num_to_wake) {
+    // Ensure that we can write to the address.
+    if (!Memory::IsValidVirtualAddress(address)) {
+        return ERR_INVALID_ADDRESS_STATE;
+    }
+
+    // Get threads waiting on the address.
+    const std::vector<SharedPtr<Thread>> waiting_threads = GetThreadsWaitingOnAddress(address);
+
+    // Determine the modified value depending on the waiting count.
+    s32 updated_value;
+    if (waiting_threads.empty()) {
+        updated_value = value - 1;
+    } else if (num_to_wake <= 0 || waiting_threads.size() <= static_cast<u32>(num_to_wake)) {
+        updated_value = value + 1;
+    } else {
+        updated_value = value;
+    }
+
+    if (static_cast<s32>(Memory::Read32(address)) != value) {
+        return ERR_INVALID_STATE;
+    }
+
+    Memory::Write32(address, static_cast<u32>(updated_value));
+    WakeThreads(waiting_threads, num_to_wake);
+    return RESULT_SUCCESS;
+}
+
+ResultCode AddressArbiter::WaitForAddressIfLessThan(VAddr address, s32 value, s64 timeout,
+                                                    bool should_decrement) {
+    // Ensure that we can read the address.
+    if (!Memory::IsValidVirtualAddress(address)) {
+        return ERR_INVALID_ADDRESS_STATE;
+    }
+
+    const s32 cur_value = static_cast<s32>(Memory::Read32(address));
+    if (cur_value >= value) {
+        return ERR_INVALID_STATE;
+    }
+
+    if (should_decrement) {
+        Memory::Write32(address, static_cast<u32>(cur_value - 1));
+    }
+
+    // Short-circuit without rescheduling, if timeout is zero.
+    if (timeout == 0) {
+        return RESULT_TIMEOUT;
+    }
+
+    return WaitForAddress(address, timeout);
+}
+
+ResultCode AddressArbiter::WaitForAddressIfEqual(VAddr address, s32 value, s64 timeout) {
+    // Ensure that we can read the address.
+    if (!Memory::IsValidVirtualAddress(address)) {
+        return ERR_INVALID_ADDRESS_STATE;
+    }
+    // Only wait for the address if equal.
+    if (static_cast<s32>(Memory::Read32(address)) != value) {
+        return ERR_INVALID_STATE;
+    }
+    // Short-circuit without rescheduling, if timeout is zero.
+    if (timeout == 0) {
+        return RESULT_TIMEOUT;
+    }
+
+    return WaitForAddress(address, timeout);
+}
+
+ResultCode AddressArbiter::WaitForAddress(VAddr address, s64 timeout) {
+    SharedPtr<Thread> current_thread = system.CurrentScheduler().GetCurrentThread();
     current_thread->SetArbiterWaitAddress(address);
     current_thread->SetStatus(ThreadStatus::WaitArb);
     current_thread->InvalidateWakeupCallback();
 
     current_thread->WakeAfterDelay(timeout);
 
-    Core::System::GetInstance().CpuCore(current_thread->GetProcessorID()).PrepareReschedule();
+    system.CpuCore(current_thread->GetProcessorID()).PrepareReschedule();
     return RESULT_TIMEOUT;
 }
 
-// Gets the threads waiting on an address.
-static std::vector<SharedPtr<Thread>> GetThreadsWaitingOnAddress(VAddr address) {
-    const auto RetrieveWaitingThreads = [](std::size_t core_index,
-                                           std::vector<SharedPtr<Thread>>& waiting_threads,
-                                           VAddr arb_addr) {
-        const auto& scheduler = Core::System::GetInstance().Scheduler(core_index);
+std::vector<SharedPtr<Thread>> AddressArbiter::GetThreadsWaitingOnAddress(VAddr address) const {
+    const auto RetrieveWaitingThreads = [this](std::size_t core_index,
+                                               std::vector<SharedPtr<Thread>>& waiting_threads,
+                                               VAddr arb_addr) {
+        const auto& scheduler = system.Scheduler(core_index);
         const auto& thread_list = scheduler.GetThreadList();
 
         for (const auto& thread : thread_list) {
-            if (thread->GetArbiterWaitAddress() == arb_addr)
+            if (thread->GetArbiterWaitAddress() == arb_addr) {
                 waiting_threads.push_back(thread);
+            }
         }
     };
 
@@ -61,118 +174,4 @@ static std::vector<SharedPtr<Thread>> GetThreadsWaitingOnAddress(VAddr address)
 
     return threads;
 }
-
-// Wake up num_to_wake (or all) threads in a vector.
-static void WakeThreads(std::vector<SharedPtr<Thread>>& waiting_threads, s32 num_to_wake) {
-    // Only process up to 'target' threads, unless 'target' is <= 0, in which case process
-    // them all.
-    std::size_t last = waiting_threads.size();
-    if (num_to_wake > 0)
-        last = num_to_wake;
-
-    // Signal the waiting threads.
-    for (std::size_t i = 0; i < last; i++) {
-        ASSERT(waiting_threads[i]->GetStatus() == ThreadStatus::WaitArb);
-        waiting_threads[i]->SetWaitSynchronizationResult(RESULT_SUCCESS);
-        waiting_threads[i]->SetArbiterWaitAddress(0);
-        waiting_threads[i]->ResumeFromWait();
-    }
-}
-
-// Signals an address being waited on.
-ResultCode SignalToAddress(VAddr address, s32 num_to_wake) {
-    std::vector<SharedPtr<Thread>> waiting_threads = GetThreadsWaitingOnAddress(address);
-
-    WakeThreads(waiting_threads, num_to_wake);
-    return RESULT_SUCCESS;
-}
-
-// Signals an address being waited on and increments its value if equal to the value argument.
-ResultCode IncrementAndSignalToAddressIfEqual(VAddr address, s32 value, s32 num_to_wake) {
-    // Ensure that we can write to the address.
-    if (!Memory::IsValidVirtualAddress(address)) {
-        return ERR_INVALID_ADDRESS_STATE;
-    }
-
-    if (static_cast<s32>(Memory::Read32(address)) == value) {
-        Memory::Write32(address, static_cast<u32>(value + 1));
-    } else {
-        return ERR_INVALID_STATE;
-    }
-
-    return SignalToAddress(address, num_to_wake);
-}
-
-// Signals an address being waited on and modifies its value based on waiting thread count if equal
-// to the value argument.
-ResultCode ModifyByWaitingCountAndSignalToAddressIfEqual(VAddr address, s32 value,
-                                                         s32 num_to_wake) {
-    // Ensure that we can write to the address.
-    if (!Memory::IsValidVirtualAddress(address)) {
-        return ERR_INVALID_ADDRESS_STATE;
-    }
-
-    // Get threads waiting on the address.
-    std::vector<SharedPtr<Thread>> waiting_threads = GetThreadsWaitingOnAddress(address);
-
-    // Determine the modified value depending on the waiting count.
-    s32 updated_value;
-    if (waiting_threads.empty()) {
-        updated_value = value - 1;
-    } else if (num_to_wake <= 0 || waiting_threads.size() <= static_cast<u32>(num_to_wake)) {
-        updated_value = value + 1;
-    } else {
-        updated_value = value;
-    }
-
-    if (static_cast<s32>(Memory::Read32(address)) == value) {
-        Memory::Write32(address, static_cast<u32>(updated_value));
-    } else {
-        return ERR_INVALID_STATE;
-    }
-
-    WakeThreads(waiting_threads, num_to_wake);
-    return RESULT_SUCCESS;
-}
-
-// Waits on an address if the value passed is less than the argument value, optionally decrementing.
-ResultCode WaitForAddressIfLessThan(VAddr address, s32 value, s64 timeout, bool should_decrement) {
-    // Ensure that we can read the address.
-    if (!Memory::IsValidVirtualAddress(address)) {
-        return ERR_INVALID_ADDRESS_STATE;
-    }
-
-    s32 cur_value = static_cast<s32>(Memory::Read32(address));
-    if (cur_value < value) {
-        if (should_decrement) {
-            Memory::Write32(address, static_cast<u32>(cur_value - 1));
-        }
-    } else {
-        return ERR_INVALID_STATE;
-    }
-    // Short-circuit without rescheduling, if timeout is zero.
-    if (timeout == 0) {
-        return RESULT_TIMEOUT;
-    }
-
-    return WaitForAddress(address, timeout);
-}
-
-// Waits on an address if the value passed is equal to the argument value.
-ResultCode WaitForAddressIfEqual(VAddr address, s32 value, s64 timeout) {
-    // Ensure that we can read the address.
-    if (!Memory::IsValidVirtualAddress(address)) {
-        return ERR_INVALID_ADDRESS_STATE;
-    }
-    // Only wait for the address if equal.
-    if (static_cast<s32>(Memory::Read32(address)) != value) {
-        return ERR_INVALID_STATE;
-    }
-    // Short-circuit without rescheduling, if timeout is zero.
-    if (timeout == 0) {
-        return RESULT_TIMEOUT;
-    }
-
-    return WaitForAddress(address, timeout);
-}
-} // namespace Kernel::AddressArbiter
+} // namespace Kernel

src/core/hle/kernel/address_arbiter.h

@@ -5,28 +5,68 @@
 #pragma once
 
 #include "common/common_types.h"
+#include "core/hle/kernel/address_arbiter.h"
 
 union ResultCode;
 
-namespace Kernel::AddressArbiter {
+namespace Core {
+class System;
+}
 
-enum class ArbitrationType {
-    WaitIfLessThan = 0,
-    DecrementAndWaitIfLessThan = 1,
-    WaitIfEqual = 2,
+namespace Kernel {
+
+class Thread;
+
+class AddressArbiter {
+public:
+    enum class ArbitrationType {
+        WaitIfLessThan = 0,
+        DecrementAndWaitIfLessThan = 1,
+        WaitIfEqual = 2,
+    };
+
+    enum class SignalType {
+        Signal = 0,
+        IncrementAndSignalIfEqual = 1,
+        ModifyByWaitingCountAndSignalIfEqual = 2,
+    };
+
+    explicit AddressArbiter(Core::System& system);
+    ~AddressArbiter();
+
+    AddressArbiter(const AddressArbiter&) = delete;
+    AddressArbiter& operator=(const AddressArbiter&) = delete;
+
+    AddressArbiter(AddressArbiter&&) = default;
+    AddressArbiter& operator=(AddressArbiter&&) = delete;
+
+    /// Signals an address being waited on.
+    ResultCode SignalToAddress(VAddr address, s32 num_to_wake);
+
+    /// Signals an address being waited on and increments its value if equal to the value argument.
+    ResultCode IncrementAndSignalToAddressIfEqual(VAddr address, s32 value, s32 num_to_wake);
+
+    /// Signals an address being waited on and modifies its value based on waiting thread count if
+    /// equal to the value argument.
+    ResultCode ModifyByWaitingCountAndSignalToAddressIfEqual(VAddr address, s32 value,
+                                                             s32 num_to_wake);
+
+    /// Waits on an address if the value passed is less than the argument value,
+    /// optionally decrementing.
+    ResultCode WaitForAddressIfLessThan(VAddr address, s32 value, s64 timeout,
+                                        bool should_decrement);
+
+    /// Waits on an address if the value passed is equal to the argument value.
+    ResultCode WaitForAddressIfEqual(VAddr address, s32 value, s64 timeout);
+
+private:
+    // Waits on the given address with a timeout in nanoseconds
+    ResultCode WaitForAddress(VAddr address, s64 timeout);
+
+    // Gets the threads waiting on an address.
+    std::vector<SharedPtr<Thread>> GetThreadsWaitingOnAddress(VAddr address) const;
+
+    Core::System& system;
 };
 
-enum class SignalType {
-    Signal = 0,
-    IncrementAndSignalIfEqual = 1,
-    ModifyByWaitingCountAndSignalIfEqual = 2,
-};
-
-ResultCode SignalToAddress(VAddr address, s32 num_to_wake);
-ResultCode IncrementAndSignalToAddressIfEqual(VAddr address, s32 value, s32 num_to_wake);
-ResultCode ModifyByWaitingCountAndSignalToAddressIfEqual(VAddr address, s32 value, s32 num_to_wake);
-
-ResultCode WaitForAddressIfLessThan(VAddr address, s32 value, s64 timeout, bool should_decrement);
-ResultCode WaitForAddressIfEqual(VAddr address, s32 value, s64 timeout);
-
-} // namespace Kernel::AddressArbiter
+} // namespace Kernel

src/core/hle/kernel/client_session.cpp

@@ -17,21 +17,11 @@ ClientSession::~ClientSession() {
     // This destructor will be called automatically when the last ClientSession handle is closed by
     // the emulated application.
 
-    // Local references to ServerSession and SessionRequestHandler are necessary to guarantee they
+    // A local reference to the ServerSession is necessary to guarantee it
     // will be kept alive until after ClientDisconnected() returns.
     SharedPtr<ServerSession> server = parent->server;
     if (server) {
-        std::shared_ptr<SessionRequestHandler> hle_handler = server->hle_handler;
-        if (hle_handler)
-            hle_handler->ClientDisconnected(server);
-
-        // TODO(Subv): Force a wake up of all the ServerSession's waiting threads and set
-        // their WaitSynchronization result to 0xC920181A.
-
-        // Clean up the list of client threads with pending requests, they are unneeded now that the
-        // client endpoint is closed.
-        server->pending_requesting_threads.clear();
-        server->currently_handling = nullptr;
+        server->ClientDisconnected();
     }
 
     parent->client = nullptr;

src/core/hle/kernel/client_session.h

@@ -36,14 +36,15 @@ public:
 
     ResultCode SendSyncRequest(SharedPtr<Thread> thread);
 
-    std::string name; ///< Name of client port (optional)
+private:
+    explicit ClientSession(KernelCore& kernel);
+    ~ClientSession() override;
 
     /// The parent session, which links to the server endpoint.
     std::shared_ptr<Session> parent;
 
-private:
-    explicit ClientSession(KernelCore& kernel);
-    ~ClientSession() override;
+    /// Name of the client session (optional)
+    std::string name;
 };
 
 } // namespace Kernel

src/core/hle/kernel/hle_ipc.cpp

@@ -86,7 +86,7 @@ HLERequestContext::~HLERequestContext() = default;
 void HLERequestContext::ParseCommandBuffer(const HandleTable& handle_table, u32_le* src_cmdbuf,
                                            bool incoming) {
     IPC::RequestParser rp(src_cmdbuf);
-    command_header = std::make_shared<IPC::CommandHeader>(rp.PopRaw<IPC::CommandHeader>());
+    command_header = rp.PopRaw<IPC::CommandHeader>();
 
     if (command_header->type == IPC::CommandType::Close) {
         // Close does not populate the rest of the IPC header
@@ -95,8 +95,7 @@ void HLERequestContext::ParseCommandBuffer(const HandleTable& handle_table, u32_
 
     // If handle descriptor is present, add size of it
     if (command_header->enable_handle_descriptor) {
-        handle_descriptor_header =
-            std::make_shared<IPC::HandleDescriptorHeader>(rp.PopRaw<IPC::HandleDescriptorHeader>());
+        handle_descriptor_header = rp.PopRaw<IPC::HandleDescriptorHeader>();
         if (handle_descriptor_header->send_current_pid) {
             rp.Skip(2, false);
         }
@@ -140,16 +139,15 @@ void HLERequestContext::ParseCommandBuffer(const HandleTable& handle_table, u32_
         // If this is an incoming message, only CommandType "Request" has a domain header
         // All outgoing domain messages have the domain header, if only incoming has it
         if (incoming || domain_message_header) {
-            domain_message_header =
-                std::make_shared<IPC::DomainMessageHeader>(rp.PopRaw<IPC::DomainMessageHeader>());
+            domain_message_header = rp.PopRaw<IPC::DomainMessageHeader>();
         } else {
-            if (Session()->IsDomain())
+            if (Session()->IsDomain()) {
                 LOG_WARNING(IPC, "Domain request has no DomainMessageHeader!");
+            }
         }
     }
 
-    data_payload_header =
-        std::make_shared<IPC::DataPayloadHeader>(rp.PopRaw<IPC::DataPayloadHeader>());
+    data_payload_header = rp.PopRaw<IPC::DataPayloadHeader>();
 
     data_payload_offset = rp.GetCurrentOffset();
 
@@ -264,11 +262,11 @@ ResultCode HLERequestContext::WriteToOutgoingCommandBuffer(Thread& thread) {
         // Write the domain objects to the command buffer, these go after the raw untranslated data.
         // TODO(Subv): This completely ignores C buffers.
         std::size_t domain_offset = size - domain_message_header->num_objects;
-        auto& request_handlers = server_session->domain_request_handlers;
 
-        for (auto& object : domain_objects) {
-            request_handlers.emplace_back(object);
-            dst_cmdbuf[domain_offset++] = static_cast<u32_le>(request_handlers.size());
+        for (const auto& object : domain_objects) {
+            server_session->AppendDomainRequestHandler(object);
+            dst_cmdbuf[domain_offset++] =
+                static_cast<u32_le>(server_session->NumDomainRequestHandlers());
         }
     }
 

src/core/hle/kernel/hle_ipc.h

@@ -6,6 +6,7 @@
 
 #include <array>
 #include <memory>
+#include <optional>
 #include <string>
 #include <type_traits>
 #include <vector>
@@ -15,6 +16,8 @@
 #include "core/hle/ipc.h"
 #include "core/hle/kernel/object.h"
 
+union ResultCode;
+
 namespace Service {
 class ServiceFrameworkBase;
 }
@@ -166,12 +169,12 @@ public:
         return buffer_c_desciptors;
     }
 
-    const IPC::DomainMessageHeader* GetDomainMessageHeader() const {
-        return domain_message_header.get();
+    const IPC::DomainMessageHeader& GetDomainMessageHeader() const {
+        return domain_message_header.value();
     }
 
     bool HasDomainMessageHeader() const {
-        return domain_message_header != nullptr;
+        return domain_message_header.has_value();
     }
 
     /// Helper function to read a buffer using the appropriate buffer descriptor
@@ -208,14 +211,12 @@ public:
 
     template <typename T>
     SharedPtr<T> GetCopyObject(std::size_t index) {
-        ASSERT(index < copy_objects.size());
-        return DynamicObjectCast<T>(copy_objects[index]);
+        return DynamicObjectCast<T>(copy_objects.at(index));
     }
 
     template <typename T>
     SharedPtr<T> GetMoveObject(std::size_t index) {
-        ASSERT(index < move_objects.size());
-        return DynamicObjectCast<T>(move_objects[index]);
+        return DynamicObjectCast<T>(move_objects.at(index));
     }
 
     void AddMoveObject(SharedPtr<Object> object) {
@@ -232,7 +233,7 @@ public:
 
     template <typename T>
     std::shared_ptr<T> GetDomainRequestHandler(std::size_t index) const {
-        return std::static_pointer_cast<T>(domain_request_handlers[index]);
+        return std::static_pointer_cast<T>(domain_request_handlers.at(index));
     }
 
     void SetDomainRequestHandlers(
@@ -272,10 +273,10 @@ private:
     boost::container::small_vector<SharedPtr<Object>, 8> copy_objects;
     boost::container::small_vector<std::shared_ptr<SessionRequestHandler>, 8> domain_objects;
 
-    std::shared_ptr<IPC::CommandHeader> command_header;
-    std::shared_ptr<IPC::HandleDescriptorHeader> handle_descriptor_header;
-    std::shared_ptr<IPC::DataPayloadHeader> data_payload_header;
-    std::shared_ptr<IPC::DomainMessageHeader> domain_message_header;
+    std::optional<IPC::CommandHeader> command_header;
+    std::optional<IPC::HandleDescriptorHeader> handle_descriptor_header;
+    std::optional<IPC::DataPayloadHeader> data_payload_header;
+    std::optional<IPC::DomainMessageHeader> domain_message_header;
     std::vector<IPC::BufferDescriptorX> buffer_x_desciptors;
     std::vector<IPC::BufferDescriptorABW> buffer_a_desciptors;
     std::vector<IPC::BufferDescriptorABW> buffer_b_desciptors;

src/core/hle/kernel/kernel.cpp

@@ -12,6 +12,7 @@
 
 #include "core/core.h"
 #include "core/core_timing.h"
+#include "core/hle/kernel/address_arbiter.h"
 #include "core/hle/kernel/client_port.h"
 #include "core/hle/kernel/handle_table.h"
 #include "core/hle/kernel/kernel.h"
@@ -86,11 +87,13 @@ static void ThreadWakeupCallback(u64 thread_handle, [[maybe_unused]] int cycles_
 }
 
 struct KernelCore::Impl {
-    void Initialize(KernelCore& kernel, Core::Timing::CoreTiming& core_timing) {
+    explicit Impl(Core::System& system) : address_arbiter{system}, system{system} {}
+
+    void Initialize(KernelCore& kernel) {
         Shutdown();
 
         InitializeSystemResourceLimit(kernel);
-        InitializeThreads(core_timing);
+        InitializeThreads();
     }
 
     void Shutdown() {
@@ -122,9 +125,9 @@ struct KernelCore::Impl {
         ASSERT(system_resource_limit->SetLimitValue(ResourceType::Sessions, 900).IsSuccess());
     }
 
-    void InitializeThreads(Core::Timing::CoreTiming& core_timing) {
+    void InitializeThreads() {
         thread_wakeup_event_type =
-            core_timing.RegisterEvent("ThreadWakeupCallback", ThreadWakeupCallback);
+            system.CoreTiming().RegisterEvent("ThreadWakeupCallback", ThreadWakeupCallback);
     }
 
     std::atomic<u32> next_object_id{0};
@@ -135,6 +138,8 @@ struct KernelCore::Impl {
     std::vector<SharedPtr<Process>> process_list;
     Process* current_process = nullptr;
 
+    Kernel::AddressArbiter address_arbiter;
+
     SharedPtr<ResourceLimit> system_resource_limit;
 
     Core::Timing::EventType* thread_wakeup_event_type = nullptr;
@@ -145,15 +150,18 @@ struct KernelCore::Impl {
     /// Map of named ports managed by the kernel, which can be retrieved using
     /// the ConnectToPort SVC.
     NamedPortTable named_ports;
+
+    // System context
+    Core::System& system;
 };
 
-KernelCore::KernelCore() : impl{std::make_unique<Impl>()} {}
+KernelCore::KernelCore(Core::System& system) : impl{std::make_unique<Impl>(system)} {}
 KernelCore::~KernelCore() {
     Shutdown();
 }
 
-void KernelCore::Initialize(Core::Timing::CoreTiming& core_timing) {
-    impl->Initialize(*this, core_timing);
+void KernelCore::Initialize() {
+    impl->Initialize(*this);
 }
 
 void KernelCore::Shutdown() {
@@ -184,6 +192,14 @@ const Process* KernelCore::CurrentProcess() const {
     return impl->current_process;
 }
 
+AddressArbiter& KernelCore::AddressArbiter() {
+    return impl->address_arbiter;
+}
+
+const AddressArbiter& KernelCore::AddressArbiter() const {
+    return impl->address_arbiter;
+}
+
 void KernelCore::AddNamedPort(std::string name, SharedPtr<ClientPort> port) {
     impl->named_ports.emplace(std::move(name), std::move(port));
 }

src/core/hle/kernel/kernel.h

@@ -11,6 +11,10 @@
 template <typename T>
 class ResultVal;
 
+namespace Core {
+class System;
+}
+
 namespace Core::Timing {
 class CoreTiming;
 struct EventType;
@@ -18,6 +22,7 @@ struct EventType;
 
 namespace Kernel {
 
+class AddressArbiter;
 class ClientPort;
 class HandleTable;
 class Process;
@@ -30,7 +35,14 @@ private:
     using NamedPortTable = std::unordered_map<std::string, SharedPtr<ClientPort>>;
 
 public:
-    KernelCore();
+    /// Constructs an instance of the kernel using the given System
+    /// instance as a context for any necessary system-related state,
+    /// such as threads, CPU core state, etc.
+    ///
+    /// @post After execution of the constructor, the provided System
+    ///       object *must* outlive the kernel instance itself.
+    ///
+    explicit KernelCore(Core::System& system);
     ~KernelCore();
 
     KernelCore(const KernelCore&) = delete;
@@ -40,11 +52,7 @@ public:
     KernelCore& operator=(KernelCore&&) = delete;
 
     /// Resets the kernel to a clean slate for use.
-    ///
-    /// @param core_timing CoreTiming instance used to create any necessary
-    ///                    kernel-specific callback events.
-    ///
-    void Initialize(Core::Timing::CoreTiming& core_timing);
+    void Initialize();
 
     /// Clears all resources in use by the kernel instance.
     void Shutdown();
@@ -67,6 +75,12 @@ public:
     /// Retrieves a const pointer to the current process.
     const Process* CurrentProcess() const;
 
+    /// Provides a reference to the kernel's address arbiter.
+    Kernel::AddressArbiter& AddressArbiter();
+
+    /// Provides a const reference to the kernel's address arbiter.
+    const Kernel::AddressArbiter& AddressArbiter() const;
+
     /// Adds a port to the named port table
     void AddNamedPort(std::string name, SharedPtr<ClientPort> port);
 

src/core/hle/kernel/scheduler.cpp

@@ -19,7 +19,8 @@ namespace Kernel {
 
 std::mutex Scheduler::scheduler_mutex;
 
-Scheduler::Scheduler(Core::ARM_Interface& cpu_core) : cpu_core(cpu_core) {}
+Scheduler::Scheduler(Core::System& system, Core::ARM_Interface& cpu_core)
+    : cpu_core{cpu_core}, system{system} {}
 
 Scheduler::~Scheduler() {
     for (auto& thread : thread_list) {
@@ -61,7 +62,7 @@ Thread* Scheduler::PopNextReadyThread() {
 
 void Scheduler::SwitchContext(Thread* new_thread) {
     Thread* const previous_thread = GetCurrentThread();
-    Process* const previous_process = Core::CurrentProcess();
+    Process* const previous_process = system.Kernel().CurrentProcess();
 
     UpdateLastContextSwitchTime(previous_thread, previous_process);
 
@@ -94,8 +95,8 @@ void Scheduler::SwitchContext(Thread* new_thread) {
 
         auto* const thread_owner_process = current_thread->GetOwnerProcess();
         if (previous_process != thread_owner_process) {
-            Core::System::GetInstance().Kernel().MakeCurrentProcess(thread_owner_process);
-            SetCurrentPageTable(&Core::CurrentProcess()->VMManager().page_table);
+            system.Kernel().MakeCurrentProcess(thread_owner_process);
+            SetCurrentPageTable(&thread_owner_process->VMManager().page_table);
         }
 
         cpu_core.LoadContext(new_thread->GetContext());
@@ -111,7 +112,7 @@ void Scheduler::SwitchContext(Thread* new_thread) {
 
 void Scheduler::UpdateLastContextSwitchTime(Thread* thread, Process* process) {
     const u64 prev_switch_ticks = last_context_switch_time;
-    const u64 most_recent_switch_ticks = Core::System::GetInstance().CoreTiming().GetTicks();
+    const u64 most_recent_switch_ticks = system.CoreTiming().GetTicks();
     const u64 update_ticks = most_recent_switch_ticks - prev_switch_ticks;
 
     if (thread != nullptr) {
@@ -223,8 +224,7 @@ void Scheduler::YieldWithLoadBalancing(Thread* thread) {
     // Take the first non-nullptr one
     for (unsigned cur_core = 0; cur_core < Core::NUM_CPU_CORES; ++cur_core) {
         const auto res =
-            Core::System::GetInstance().CpuCore(cur_core).Scheduler().GetNextSuggestedThread(
-                core, priority);
+            system.CpuCore(cur_core).Scheduler().GetNextSuggestedThread(core, priority);
 
         // If scheduler provides a suggested thread
         if (res != nullptr) {

src/core/hle/kernel/scheduler.h

@@ -13,7 +13,8 @@
 
 namespace Core {
 class ARM_Interface;
-}
+class System;
+} // namespace Core
 
 namespace Kernel {
 
@@ -21,7 +22,7 @@ class Process;
 
 class Scheduler final {
 public:
-    explicit Scheduler(Core::ARM_Interface& cpu_core);
+    explicit Scheduler(Core::System& system, Core::ARM_Interface& cpu_core);
     ~Scheduler();
 
     /// Returns whether there are any threads that are ready to run.
@@ -162,6 +163,7 @@ private:
     Core::ARM_Interface& cpu_core;
     u64 last_context_switch_time = 0;
 
+    Core::System& system;
     static std::mutex scheduler_mutex;
 };
 

src/core/hle/kernel/server_session.cpp

@@ -63,42 +63,71 @@ void ServerSession::Acquire(Thread* thread) {
     pending_requesting_threads.pop_back();
 }
 
-ResultCode ServerSession::HandleDomainSyncRequest(Kernel::HLERequestContext& context) {
-    auto* const domain_message_header = context.GetDomainMessageHeader();
-    if (domain_message_header) {
-        // Set domain handlers in HLE context, used for domain objects (IPC interfaces) as inputs
-        context.SetDomainRequestHandlers(domain_request_handlers);
-
-        // If there is a DomainMessageHeader, then this is CommandType "Request"
-        const u32 object_id{context.GetDomainMessageHeader()->object_id};
-        switch (domain_message_header->command) {
-        case IPC::DomainMessageHeader::CommandType::SendMessage:
-            if (object_id > domain_request_handlers.size()) {
-                LOG_CRITICAL(IPC,
-                             "object_id {} is too big! This probably means a recent service call "
-                             "to {} needed to return a new interface!",
-                             object_id, name);
-                UNREACHABLE();
-                return RESULT_SUCCESS; // Ignore error if asserts are off
-            }
-            return domain_request_handlers[object_id - 1]->HandleSyncRequest(context);
-
-        case IPC::DomainMessageHeader::CommandType::CloseVirtualHandle: {
-            LOG_DEBUG(IPC, "CloseVirtualHandle, object_id=0x{:08X}", object_id);
-
-            domain_request_handlers[object_id - 1] = nullptr;
-
-            IPC::ResponseBuilder rb{context, 2};
-            rb.Push(RESULT_SUCCESS);
-            return RESULT_SUCCESS;
-        }
-        }
-
-        LOG_CRITICAL(IPC, "Unknown domain command={}",
-                     static_cast<int>(domain_message_header->command.Value()));
-        ASSERT(false);
+void ServerSession::ClientDisconnected() {
+    // We keep a shared pointer to the hle handler to keep it alive throughout
+    // the call to ClientDisconnected, as ClientDisconnected invalidates the
+    // hle_handler member itself during the course of the function executing.
+    std::shared_ptr<SessionRequestHandler> handler = hle_handler;
+    if (handler) {
+        // Note that after this returns, this server session's hle_handler is
+        // invalidated (set to null).
+        handler->ClientDisconnected(this);
     }
 
+    // TODO(Subv): Force a wake up of all the ServerSession's waiting threads and set
+    // their WaitSynchronization result to 0xC920181A.
+
+    // Clean up the list of client threads with pending requests, they are unneeded now that the
+    // client endpoint is closed.
+    pending_requesting_threads.clear();
+    currently_handling = nullptr;
+}
+
+void ServerSession::AppendDomainRequestHandler(std::shared_ptr<SessionRequestHandler> handler) {
+    domain_request_handlers.push_back(std::move(handler));
+}
+
+std::size_t ServerSession::NumDomainRequestHandlers() const {
+    return domain_request_handlers.size();
+}
+
+ResultCode ServerSession::HandleDomainSyncRequest(Kernel::HLERequestContext& context) {
+    if (!context.HasDomainMessageHeader()) {
+        return RESULT_SUCCESS;
+    }
+
+    // Set domain handlers in HLE context, used for domain objects (IPC interfaces) as inputs
+    context.SetDomainRequestHandlers(domain_request_handlers);
+
+    // If there is a DomainMessageHeader, then this is CommandType "Request"
+    const auto& domain_message_header = context.GetDomainMessageHeader();
+    const u32 object_id{domain_message_header.object_id};
+    switch (domain_message_header.command) {
+    case IPC::DomainMessageHeader::CommandType::SendMessage:
+        if (object_id > domain_request_handlers.size()) {
+            LOG_CRITICAL(IPC,
+                         "object_id {} is too big! This probably means a recent service call "
+                         "to {} needed to return a new interface!",
+                         object_id, name);
+            UNREACHABLE();
+            return RESULT_SUCCESS; // Ignore error if asserts are off
+        }
+        return domain_request_handlers[object_id - 1]->HandleSyncRequest(context);
+
+    case IPC::DomainMessageHeader::CommandType::CloseVirtualHandle: {
+        LOG_DEBUG(IPC, "CloseVirtualHandle, object_id=0x{:08X}", object_id);
+
+        domain_request_handlers[object_id - 1] = nullptr;
+
+        IPC::ResponseBuilder rb{context, 2};
+        rb.Push(RESULT_SUCCESS);
+        return RESULT_SUCCESS;
+    }
+    }
+
+    LOG_CRITICAL(IPC, "Unknown domain command={}",
+                 static_cast<int>(domain_message_header.command.Value()));
+    ASSERT(false);
     return RESULT_SUCCESS;
 }
 

src/core/hle/kernel/server_session.h

@@ -46,6 +46,14 @@ public:
         return HANDLE_TYPE;
     }
 
+    Session* GetParent() {
+        return parent.get();
+    }
+
+    const Session* GetParent() const {
+        return parent.get();
+    }
+
     using SessionPair = std::tuple<SharedPtr<ServerSession>, SharedPtr<ClientSession>>;
 
     /**
@@ -78,23 +86,16 @@ public:
 
     void Acquire(Thread* thread) override;
 
-    std::string name;                ///< The name of this session (optional)
-    std::shared_ptr<Session> parent; ///< The parent session, which links to the client endpoint.
-    std::shared_ptr<SessionRequestHandler>
-        hle_handler; ///< This session's HLE request handler (applicable when not a domain)
+    /// Called when a client disconnection occurs.
+    void ClientDisconnected();
 
-    /// This is the list of domain request handlers (after conversion to a domain)
-    std::vector<std::shared_ptr<SessionRequestHandler>> domain_request_handlers;
+    /// Adds a new domain request handler to the collection of request handlers within
+    /// this ServerSession instance.
+    void AppendDomainRequestHandler(std::shared_ptr<SessionRequestHandler> handler);
 
-    /// List of threads that are pending a response after a sync request. This list is processed in
-    /// a LIFO manner, thus, the last request will be dispatched first.
-    /// TODO(Subv): Verify if this is indeed processed in LIFO using a hardware test.
-    std::vector<SharedPtr<Thread>> pending_requesting_threads;
-
-    /// Thread whose request is currently being handled. A request is considered "handled" when a
-    /// response is sent via svcReplyAndReceive.
-    /// TODO(Subv): Find a better name for this.
-    SharedPtr<Thread> currently_handling;
+    /// Retrieves the total number of domain request handlers that have been
+    /// appended to this ServerSession instance.
+    std::size_t NumDomainRequestHandlers() const;
 
     /// Returns true if the session has been converted to a domain, otherwise False
     bool IsDomain() const {
@@ -129,8 +130,30 @@ private:
     /// object handle.
     ResultCode HandleDomainSyncRequest(Kernel::HLERequestContext& context);
 
+    /// The parent session, which links to the client endpoint.
+    std::shared_ptr<Session> parent;
+
+    /// This session's HLE request handler (applicable when not a domain)
+    std::shared_ptr<SessionRequestHandler> hle_handler;
+
+    /// This is the list of domain request handlers (after conversion to a domain)
+    std::vector<std::shared_ptr<SessionRequestHandler>> domain_request_handlers;
+
+    /// List of threads that are pending a response after a sync request. This list is processed in
+    /// a LIFO manner, thus, the last request will be dispatched first.
+    /// TODO(Subv): Verify if this is indeed processed in LIFO using a hardware test.
+    std::vector<SharedPtr<Thread>> pending_requesting_threads;
+
+    /// Thread whose request is currently being handled. A request is considered "handled" when a
+    /// response is sent via svcReplyAndReceive.
+    /// TODO(Subv): Find a better name for this.
+    SharedPtr<Thread> currently_handling;
+
     /// When set to True, converts the session to a domain at the end of the command
     bool convert_to_domain{};
+
+    /// The name of this session (optional)
+    std::string name;
 };
 
 } // namespace Kernel

src/core/hle/kernel/shared_memory.cpp

@@ -6,7 +6,6 @@
 
 #include "common/assert.h"
 #include "common/logging/log.h"
-#include "core/core.h"
 #include "core/hle/kernel/errors.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/shared_memory.h"
@@ -34,8 +33,8 @@ SharedPtr<SharedMemory> SharedMemory::Create(KernelCore& kernel, Process* owner_
         shared_memory->backing_block_offset = 0;
 
         // Refresh the address mappings for the current process.
-        if (Core::CurrentProcess() != nullptr) {
-            Core::CurrentProcess()->VMManager().RefreshMemoryBlockMappings(
+        if (kernel.CurrentProcess() != nullptr) {
+            kernel.CurrentProcess()->VMManager().RefreshMemoryBlockMappings(
                 shared_memory->backing_block.get());
         }
     } else {

src/core/hle/kernel/svc.cpp

@@ -20,6 +20,7 @@
 #include "core/hle/kernel/address_arbiter.h"
 #include "core/hle/kernel/client_port.h"
 #include "core/hle/kernel/client_session.h"
+#include "core/hle/kernel/errors.h"
 #include "core/hle/kernel/handle_table.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/mutex.h"
@@ -47,23 +48,6 @@ constexpr bool IsValidAddressRange(VAddr address, u64 size) {
     return address + size > address;
 }
 
-// Checks if a given address range lies within a larger address range.
-constexpr bool IsInsideAddressRange(VAddr address, u64 size, VAddr address_range_begin,
-                                    VAddr address_range_end) {
-    const VAddr end_address = address + size - 1;
-    return address_range_begin <= address && end_address <= address_range_end - 1;
-}
-
-bool IsInsideAddressSpace(const VMManager& vm, VAddr address, u64 size) {
-    return IsInsideAddressRange(address, size, vm.GetAddressSpaceBaseAddress(),
-                                vm.GetAddressSpaceEndAddress());
-}
-
-bool IsInsideNewMapRegion(const VMManager& vm, VAddr address, u64 size) {
-    return IsInsideAddressRange(address, size, vm.GetNewMapRegionBaseAddress(),
-                                vm.GetNewMapRegionEndAddress());
-}
-
 // 8 GiB
 constexpr u64 MAIN_MEMORY_SIZE = 0x200000000;
 
@@ -105,14 +89,14 @@ ResultCode MapUnmapMemorySanityChecks(const VMManager& vm_manager, VAddr dst_add
         return ERR_INVALID_ADDRESS_STATE;
     }
 
-    if (!IsInsideAddressSpace(vm_manager, src_addr, size)) {
+    if (!vm_manager.IsWithinAddressSpace(src_addr, size)) {
         LOG_ERROR(Kernel_SVC,
                   "Source is not within the address space, addr=0x{:016X}, size=0x{:016X}",
                   src_addr, size);
         return ERR_INVALID_ADDRESS_STATE;
     }
 
-    if (!IsInsideNewMapRegion(vm_manager, dst_addr, size)) {
+    if (!vm_manager.IsWithinNewMapRegion(dst_addr, size)) {
         LOG_ERROR(Kernel_SVC,
                   "Destination is not within the new map region, addr=0x{:016X}, size=0x{:016X}",
                   dst_addr, size);
@@ -238,7 +222,7 @@ static ResultCode SetMemoryPermission(VAddr addr, u64 size, u32 prot) {
     auto* const current_process = Core::CurrentProcess();
     auto& vm_manager = current_process->VMManager();
 
-    if (!IsInsideAddressSpace(vm_manager, addr, size)) {
+    if (!vm_manager.IsWithinAddressSpace(addr, size)) {
         LOG_ERROR(Kernel_SVC,
                   "Source is not within the address space, addr=0x{:016X}, size=0x{:016X}", addr,
                   size);
@@ -299,7 +283,7 @@ static ResultCode SetMemoryAttribute(VAddr address, u64 size, u32 mask, u32 attr
     }
 
     auto& vm_manager = Core::CurrentProcess()->VMManager();
-    if (!IsInsideAddressSpace(vm_manager, address, size)) {
+    if (!vm_manager.IsWithinAddressSpace(address, size)) {
         LOG_ERROR(Kernel_SVC,
                   "Given address (0x{:016X}) is outside the bounds of the address space.", address);
         return ERR_INVALID_ADDRESS_STATE;
@@ -1495,13 +1479,14 @@ static ResultCode WaitForAddress(VAddr address, u32 type, s32 value, s64 timeout
         return ERR_INVALID_ADDRESS;
     }
 
+    auto& address_arbiter = Core::System::GetInstance().Kernel().AddressArbiter();
     switch (static_cast<AddressArbiter::ArbitrationType>(type)) {
     case AddressArbiter::ArbitrationType::WaitIfLessThan:
-        return AddressArbiter::WaitForAddressIfLessThan(address, value, timeout, false);
+        return address_arbiter.WaitForAddressIfLessThan(address, value, timeout, false);
     case AddressArbiter::ArbitrationType::DecrementAndWaitIfLessThan:
-        return AddressArbiter::WaitForAddressIfLessThan(address, value, timeout, true);
+        return address_arbiter.WaitForAddressIfLessThan(address, value, timeout, true);
     case AddressArbiter::ArbitrationType::WaitIfEqual:
-        return AddressArbiter::WaitForAddressIfEqual(address, value, timeout);
+        return address_arbiter.WaitForAddressIfEqual(address, value, timeout);
     default:
         LOG_ERROR(Kernel_SVC,
                   "Invalid arbitration type, expected WaitIfLessThan, DecrementAndWaitIfLessThan "
@@ -1526,13 +1511,14 @@ static ResultCode SignalToAddress(VAddr address, u32 type, s32 value, s32 num_to
         return ERR_INVALID_ADDRESS;
     }
 
+    auto& address_arbiter = Core::System::GetInstance().Kernel().AddressArbiter();
     switch (static_cast<AddressArbiter::SignalType>(type)) {
     case AddressArbiter::SignalType::Signal:
-        return AddressArbiter::SignalToAddress(address, num_to_wake);
+        return address_arbiter.SignalToAddress(address, num_to_wake);
     case AddressArbiter::SignalType::IncrementAndSignalIfEqual:
-        return AddressArbiter::IncrementAndSignalToAddressIfEqual(address, value, num_to_wake);
+        return address_arbiter.IncrementAndSignalToAddressIfEqual(address, value, num_to_wake);
     case AddressArbiter::SignalType::ModifyByWaitingCountAndSignalIfEqual:
-        return AddressArbiter::ModifyByWaitingCountAndSignalToAddressIfEqual(address, value,
+        return address_arbiter.ModifyByWaitingCountAndSignalToAddressIfEqual(address, value,
                                                                              num_to_wake);
     default:
         LOG_ERROR(Kernel_SVC,

src/core/hle/kernel/thread.cpp

@@ -184,8 +184,6 @@ ResultVal<SharedPtr<Thread>> Thread::Create(KernelCore& kernel, std::string name
         return ERR_INVALID_PROCESSOR_ID;
     }
 
-    // TODO(yuriks): Other checks, returning 0xD9001BEA
-
     if (!Memory::IsValidVirtualAddress(owner_process, entry_point)) {
         LOG_ERROR(Kernel_SVC, "(name={}): invalid entry {:016X}", name, entry_point);
         // TODO (bunnei): Find the correct error code to use here

src/core/hle/kernel/vm_manager.cpp

@@ -17,8 +17,8 @@
 #include "core/memory_setup.h"
 
 namespace Kernel {
-
-static const char* GetMemoryStateName(MemoryState state) {
+namespace {
+const char* GetMemoryStateName(MemoryState state) {
     static constexpr const char* names[] = {
         "Unmapped",         "Io",
         "Normal",           "CodeStatic",
@@ -35,6 +35,14 @@ static const char* GetMemoryStateName(MemoryState state) {
     return names[ToSvcMemoryState(state)];
 }
 
+// Checks if a given address range lies within a larger address range.
+constexpr bool IsInsideAddressRange(VAddr address, u64 size, VAddr address_range_begin,
+                                    VAddr address_range_end) {
+    const VAddr end_address = address + size - 1;
+    return address_range_begin <= address && end_address <= address_range_end - 1;
+}
+} // Anonymous namespace
+
 bool VirtualMemoryArea::CanBeMergedWith(const VirtualMemoryArea& next) const {
     ASSERT(base + size == next.base);
     if (permissions != next.permissions || state != next.state || attribute != next.attribute ||
@@ -249,8 +257,7 @@ ResultCode VMManager::ReprotectRange(VAddr target, u64 size, VMAPermission new_p
 }
 
 ResultVal<VAddr> VMManager::HeapAllocate(VAddr target, u64 size, VMAPermission perms) {
-    if (target < GetHeapRegionBaseAddress() || target + size > GetHeapRegionEndAddress() ||
-        target + size < target) {
+    if (!IsWithinHeapRegion(target, size)) {
         return ERR_INVALID_ADDRESS;
     }
 
@@ -285,8 +292,7 @@ ResultVal<VAddr> VMManager::HeapAllocate(VAddr target, u64 size, VMAPermission p
 }
 
 ResultCode VMManager::HeapFree(VAddr target, u64 size) {
-    if (target < GetHeapRegionBaseAddress() || target + size > GetHeapRegionEndAddress() ||
-        target + size < target) {
+    if (!IsWithinHeapRegion(target, size)) {
         return ERR_INVALID_ADDRESS;
     }
 
@@ -706,6 +712,11 @@ u64 VMManager::GetAddressSpaceWidth() const {
     return address_space_width;
 }
 
+bool VMManager::IsWithinAddressSpace(VAddr address, u64 size) const {
+    return IsInsideAddressRange(address, size, GetAddressSpaceBaseAddress(),
+                                GetAddressSpaceEndAddress());
+}
+
 VAddr VMManager::GetASLRRegionBaseAddress() const {
     return aslr_region_base;
 }
@@ -750,6 +761,11 @@ u64 VMManager::GetCodeRegionSize() const {
     return code_region_end - code_region_base;
 }
 
+bool VMManager::IsWithinCodeRegion(VAddr address, u64 size) const {
+    return IsInsideAddressRange(address, size, GetCodeRegionBaseAddress(),
+                                GetCodeRegionEndAddress());
+}
+
 VAddr VMManager::GetHeapRegionBaseAddress() const {
     return heap_region_base;
 }
@@ -762,6 +778,11 @@ u64 VMManager::GetHeapRegionSize() const {
     return heap_region_end - heap_region_base;
 }
 
+bool VMManager::IsWithinHeapRegion(VAddr address, u64 size) const {
+    return IsInsideAddressRange(address, size, GetHeapRegionBaseAddress(),
+                                GetHeapRegionEndAddress());
+}
+
 VAddr VMManager::GetMapRegionBaseAddress() const {
     return map_region_base;
 }
@@ -774,6 +795,10 @@ u64 VMManager::GetMapRegionSize() const {
     return map_region_end - map_region_base;
 }
 
+bool VMManager::IsWithinMapRegion(VAddr address, u64 size) const {
+    return IsInsideAddressRange(address, size, GetMapRegionBaseAddress(), GetMapRegionEndAddress());
+}
+
 VAddr VMManager::GetNewMapRegionBaseAddress() const {
     return new_map_region_base;
 }
@@ -786,6 +811,11 @@ u64 VMManager::GetNewMapRegionSize() const {
     return new_map_region_end - new_map_region_base;
 }
 
+bool VMManager::IsWithinNewMapRegion(VAddr address, u64 size) const {
+    return IsInsideAddressRange(address, size, GetNewMapRegionBaseAddress(),
+                                GetNewMapRegionEndAddress());
+}
+
 VAddr VMManager::GetTLSIORegionBaseAddress() const {
     return tls_io_region_base;
 }
@@ -798,4 +828,9 @@ u64 VMManager::GetTLSIORegionSize() const {
     return tls_io_region_end - tls_io_region_base;
 }
 
+bool VMManager::IsWithinTLSIORegion(VAddr address, u64 size) const {
+    return IsInsideAddressRange(address, size, GetTLSIORegionBaseAddress(),
+                                GetTLSIORegionEndAddress());
+}
+
 } // namespace Kernel

src/core/hle/kernel/vm_manager.h

@@ -432,18 +432,21 @@ public:
     /// Gets the address space width in bits.
     u64 GetAddressSpaceWidth() const;
 
+    /// Determines whether or not the given address range lies within the address space.
+    bool IsWithinAddressSpace(VAddr address, u64 size) const;
+
     /// Gets the base address of the ASLR region.
     VAddr GetASLRRegionBaseAddress() const;
 
     /// Gets the end address of the ASLR region.
     VAddr GetASLRRegionEndAddress() const;
 
-    /// Determines whether or not the specified address range is within the ASLR region.
-    bool IsWithinASLRRegion(VAddr address, u64 size) const;
-
     /// Gets the size of the ASLR region
     u64 GetASLRRegionSize() const;
 
+    /// Determines whether or not the specified address range is within the ASLR region.
+    bool IsWithinASLRRegion(VAddr address, u64 size) const;
+
     /// Gets the base address of the code region.
     VAddr GetCodeRegionBaseAddress() const;
 
@@ -453,6 +456,9 @@ public:
     /// Gets the total size of the code region in bytes.
     u64 GetCodeRegionSize() const;
 
+    /// Determines whether or not the specified range is within the code region.
+    bool IsWithinCodeRegion(VAddr address, u64 size) const;
+
     /// Gets the base address of the heap region.
     VAddr GetHeapRegionBaseAddress() const;
 
@@ -462,6 +468,9 @@ public:
     /// Gets the total size of the heap region in bytes.
     u64 GetHeapRegionSize() const;
 
+    /// Determines whether or not the specified range is within the heap region.
+    bool IsWithinHeapRegion(VAddr address, u64 size) const;
+
     /// Gets the base address of the map region.
     VAddr GetMapRegionBaseAddress() const;
 
@@ -471,6 +480,9 @@ public:
     /// Gets the total size of the map region in bytes.
     u64 GetMapRegionSize() const;
 
+    /// Determines whether or not the specified range is within the map region.
+    bool IsWithinMapRegion(VAddr address, u64 size) const;
+
     /// Gets the base address of the new map region.
     VAddr GetNewMapRegionBaseAddress() const;
 
@@ -480,6 +492,9 @@ public:
     /// Gets the total size of the new map region in bytes.
     u64 GetNewMapRegionSize() const;
 
+    /// Determines whether or not the given address range is within the new map region
+    bool IsWithinNewMapRegion(VAddr address, u64 size) const;
+
     /// Gets the base address of the TLS IO region.
     VAddr GetTLSIORegionBaseAddress() const;
 
@@ -489,6 +504,9 @@ public:
     /// Gets the total size of the TLS IO region in bytes.
     u64 GetTLSIORegionSize() const;
 
+    /// Determines if the given address range is within the TLS IO region.
+    bool IsWithinTLSIORegion(VAddr address, u64 size) const;
+
     /// Each VMManager has its own page table, which is set as the main one when the owning process
     /// is scheduled.
     Memory::PageTable page_table;

src/core/hle/result.h

@@ -8,7 +8,6 @@
 #include <utility>
 #include "common/assert.h"
 #include "common/bit_field.h"
-#include "common/common_funcs.h"
 #include "common/common_types.h"
 
 // All the constants in this file come from http://switchbrew.org/index.php?title=Error_codes

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

@@ -7,6 +7,7 @@
 #include "common/string_util.h"
 #include "core/core.h"
 #include "core/frontend/applets/software_keyboard.h"
+#include "core/hle/result.h"
 #include "core/hle/service/am/am.h"
 #include "core/hle/service/am/applets/software_keyboard.h"
 

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

@@ -9,10 +9,13 @@
 #include <vector>
 
 #include "common/common_funcs.h"
+#include "common/common_types.h"
 #include "common/swap.h"
 #include "core/hle/service/am/am.h"
 #include "core/hle/service/am/applets/applets.h"
 
+union ResultCode;
+
 namespace Service::AM::Applets {
 
 enum class KeysetDisable : u32 {

src/core/hle/service/audio/audout_u.cpp

@@ -18,17 +18,11 @@
 #include "core/hle/kernel/readable_event.h"
 #include "core/hle/kernel/writable_event.h"
 #include "core/hle/service/audio/audout_u.h"
+#include "core/hle/service/audio/errors.h"
 #include "core/memory.h"
 
 namespace Service::Audio {
 
-namespace ErrCodes {
-enum {
-    ErrorUnknown = 2,
-    BufferCountExceeded = 8,
-};
-}
-
 constexpr std::array<char, 10> DefaultDevice{{"DeviceOut"}};
 constexpr int DefaultSampleRate{48000};
 
@@ -100,7 +94,7 @@ private:
 
         if (stream->IsPlaying()) {
             IPC::ResponseBuilder rb{ctx, 2};
-            rb.Push(ResultCode(ErrorModule::Audio, ErrCodes::ErrorUnknown));
+            rb.Push(ERR_OPERATION_FAILED);
             return;
         }
 
@@ -113,7 +107,9 @@ private:
     void StopAudioOut(Kernel::HLERequestContext& ctx) {
         LOG_DEBUG(Service_Audio, "called");
 
-        audio_core.StopStream(stream);
+        if (stream->IsPlaying()) {
+            audio_core.StopStream(stream);
+        }
 
         IPC::ResponseBuilder rb{ctx, 2};
         rb.Push(RESULT_SUCCESS);
@@ -143,7 +139,8 @@ private:
 
         if (!audio_core.QueueBuffer(stream, tag, std::move(samples))) {
             IPC::ResponseBuilder rb{ctx, 2};
-            rb.Push(ResultCode(ErrorModule::Audio, ErrCodes::BufferCountExceeded));
+            rb.Push(ERR_BUFFER_COUNT_EXCEEDED);
+            return;
         }
 
         IPC::ResponseBuilder rb{ctx, 2};

src/core/hle/service/audio/audren_u.cpp

@@ -17,6 +17,7 @@
 #include "core/hle/kernel/readable_event.h"
 #include "core/hle/kernel/writable_event.h"
 #include "core/hle/service/audio/audren_u.h"
+#include "core/hle/service/audio/errors.h"
 
 namespace Service::Audio {
 
@@ -146,7 +147,7 @@ private:
         // code in this case.
 
         IPC::ResponseBuilder rb{ctx, 2};
-        rb.Push(ResultCode{ErrorModule::Audio, 201});
+        rb.Push(ERR_NOT_SUPPORTED);
     }
 
     Kernel::EventPair system_event;

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

@@ -0,0 +1,15 @@
+// Copyright 2019 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "core/hle/result.h"
+
+namespace Service::Audio {
+
+constexpr ResultCode ERR_OPERATION_FAILED{ErrorModule::Audio, 2};
+constexpr ResultCode ERR_BUFFER_COUNT_EXCEEDED{ErrorModule::Audio, 8};
+constexpr ResultCode ERR_NOT_SUPPORTED{ErrorModule::Audio, 513};
+
+} // namespace Service::Audio

src/core/hle/service/audio/hwopus.cpp

@@ -9,43 +9,32 @@
 
 #include <opus.h>
 
-#include "common/common_funcs.h"
+#include "common/assert.h"
 #include "common/logging/log.h"
 #include "core/hle/ipc_helpers.h"
 #include "core/hle/kernel/hle_ipc.h"
 #include "core/hle/service/audio/hwopus.h"
 
 namespace Service::Audio {
-
+namespace {
 struct OpusDeleter {
     void operator()(void* ptr) const {
         operator delete(ptr);
     }
 };
 
-class IHardwareOpusDecoderManager final : public ServiceFramework<IHardwareOpusDecoderManager> {
+using OpusDecoderPtr = std::unique_ptr<OpusDecoder, OpusDeleter>;
+
+struct OpusPacketHeader {
+    // Packet size in bytes.
+    u32_be size;
+    // Indicates the final range of the codec's entropy coder.
+    u32_be final_range;
+};
+static_assert(sizeof(OpusPacketHeader) == 0x8, "OpusHeader is an invalid size");
+
+class OpusDecoderStateBase {
 public:
-    IHardwareOpusDecoderManager(std::unique_ptr<OpusDecoder, OpusDeleter> decoder, u32 sample_rate,
-                                u32 channel_count)
-        : ServiceFramework("IHardwareOpusDecoderManager"), decoder(std::move(decoder)),
-          sample_rate(sample_rate), channel_count(channel_count) {
-        // clang-format off
-        static const FunctionInfo functions[] = {
-            {0, &IHardwareOpusDecoderManager::DecodeInterleavedOld, "DecodeInterleavedOld"},
-            {1, nullptr, "SetContext"},
-            {2, nullptr, "DecodeInterleavedForMultiStreamOld"},
-            {3, nullptr, "SetContextForMultiStream"},
-            {4, &IHardwareOpusDecoderManager::DecodeInterleavedWithPerfOld, "DecodeInterleavedWithPerfOld"},
-            {5, nullptr, "DecodeInterleavedForMultiStreamWithPerfOld"},
-            {6, &IHardwareOpusDecoderManager::DecodeInterleaved, "DecodeInterleaved"},
-            {7, nullptr, "DecodeInterleavedForMultiStream"},
-        };
-        // clang-format on
-
-        RegisterHandlers(functions);
-    }
-
-private:
     /// Describes extra behavior that may be asked of the decoding context.
     enum class ExtraBehavior {
         /// No extra behavior.
@@ -55,30 +44,36 @@ private:
         ResetContext,
     };
 
-    void DecodeInterleavedOld(Kernel::HLERequestContext& ctx) {
-        LOG_DEBUG(Audio, "called");
+    enum class PerfTime {
+        Disabled,
+        Enabled,
+    };
 
-        DecodeInterleavedHelper(ctx, nullptr, ExtraBehavior::None);
-    }
-
-    void DecodeInterleavedWithPerfOld(Kernel::HLERequestContext& ctx) {
-        LOG_DEBUG(Audio, "called");
-
-        u64 performance = 0;
-        DecodeInterleavedHelper(ctx, &performance, ExtraBehavior::None);
-    }
-
-    void DecodeInterleaved(Kernel::HLERequestContext& ctx) {
-        LOG_DEBUG(Audio, "called");
-
-        IPC::RequestParser rp{ctx};
-        const auto extra_behavior =
-            rp.Pop<bool>() ? ExtraBehavior::ResetContext : ExtraBehavior::None;
-
-        u64 performance = 0;
-        DecodeInterleavedHelper(ctx, &performance, extra_behavior);
+    virtual ~OpusDecoderStateBase() = default;
+
+    // Decodes interleaved Opus packets. Optionally allows reporting time taken to
+    // perform the decoding, as well as any relevant extra behavior.
+    virtual void DecodeInterleaved(Kernel::HLERequestContext& ctx, PerfTime perf_time,
+                                   ExtraBehavior extra_behavior) = 0;
+};
+
+// Represents the decoder state for a non-multistream decoder.
+class OpusDecoderState final : public OpusDecoderStateBase {
+public:
+    explicit OpusDecoderState(OpusDecoderPtr decoder, u32 sample_rate, u32 channel_count)
+        : decoder{std::move(decoder)}, sample_rate{sample_rate}, channel_count{channel_count} {}
+
+    void DecodeInterleaved(Kernel::HLERequestContext& ctx, PerfTime perf_time,
+                           ExtraBehavior extra_behavior) override {
+        if (perf_time == PerfTime::Disabled) {
+            DecodeInterleavedHelper(ctx, nullptr, extra_behavior);
+        } else {
+            u64 performance = 0;
+            DecodeInterleavedHelper(ctx, &performance, extra_behavior);
+        }
     }
 
+private:
     void DecodeInterleavedHelper(Kernel::HLERequestContext& ctx, u64* performance,
                                  ExtraBehavior extra_behavior) {
         u32 consumed = 0;
@@ -89,8 +84,7 @@ private:
             ResetDecoderContext();
         }
 
-        if (!Decoder_DecodeInterleaved(consumed, sample_count, ctx.ReadBuffer(), samples,
-                                       performance)) {
+        if (!DecodeOpusData(consumed, sample_count, ctx.ReadBuffer(), samples, performance)) {
             LOG_ERROR(Audio, "Failed to decode opus data");
             IPC::ResponseBuilder rb{ctx, 2};
             // TODO(ogniK): Use correct error code
@@ -109,27 +103,27 @@ private:
         ctx.WriteBuffer(samples.data(), samples.size() * sizeof(s16));
     }
 
-    bool Decoder_DecodeInterleaved(u32& consumed, u32& sample_count, const std::vector<u8>& input,
-                                   std::vector<opus_int16>& output, u64* out_performance_time) {
+    bool DecodeOpusData(u32& consumed, u32& sample_count, const std::vector<u8>& input,
+                        std::vector<opus_int16>& output, u64* out_performance_time) const {
         const auto start_time = std::chrono::high_resolution_clock::now();
         const std::size_t raw_output_sz = output.size() * sizeof(opus_int16);
-        if (sizeof(OpusHeader) > input.size()) {
+        if (sizeof(OpusPacketHeader) > input.size()) {
             LOG_ERROR(Audio, "Input is smaller than the header size, header_sz={}, input_sz={}",
-                      sizeof(OpusHeader), input.size());
+                      sizeof(OpusPacketHeader), input.size());
             return false;
         }
 
-        OpusHeader hdr{};
-        std::memcpy(&hdr, input.data(), sizeof(OpusHeader));
-        if (sizeof(OpusHeader) + static_cast<u32>(hdr.sz) > input.size()) {
+        OpusPacketHeader hdr{};
+        std::memcpy(&hdr, input.data(), sizeof(OpusPacketHeader));
+        if (sizeof(OpusPacketHeader) + static_cast<u32>(hdr.size) > input.size()) {
             LOG_ERROR(Audio, "Input does not fit in the opus header size. data_sz={}, input_sz={}",
-                      sizeof(OpusHeader) + static_cast<u32>(hdr.sz), input.size());
+                      sizeof(OpusPacketHeader) + static_cast<u32>(hdr.size), input.size());
             return false;
         }
 
-        const auto frame = input.data() + sizeof(OpusHeader);
+        const auto frame = input.data() + sizeof(OpusPacketHeader);
         const auto decoded_sample_count = opus_packet_get_nb_samples(
-            frame, static_cast<opus_int32>(input.size() - sizeof(OpusHeader)),
+            frame, static_cast<opus_int32>(input.size() - sizeof(OpusPacketHeader)),
             static_cast<opus_int32>(sample_rate));
         if (decoded_sample_count * channel_count * sizeof(u16) > raw_output_sz) {
             LOG_ERROR(
@@ -141,18 +135,18 @@ private:
 
         const int frame_size = (static_cast<int>(raw_output_sz / sizeof(s16) / channel_count));
         const auto out_sample_count =
-            opus_decode(decoder.get(), frame, hdr.sz, output.data(), frame_size, 0);
+            opus_decode(decoder.get(), frame, hdr.size, output.data(), frame_size, 0);
         if (out_sample_count < 0) {
             LOG_ERROR(Audio,
                       "Incorrect sample count received from opus_decode, "
                       "output_sample_count={}, frame_size={}, data_sz_from_hdr={}",
-                      out_sample_count, frame_size, static_cast<u32>(hdr.sz));
+                      out_sample_count, frame_size, static_cast<u32>(hdr.size));
             return false;
         }
 
         const auto end_time = std::chrono::high_resolution_clock::now() - start_time;
         sample_count = out_sample_count;
-        consumed = static_cast<u32>(sizeof(OpusHeader) + hdr.sz);
+        consumed = static_cast<u32>(sizeof(OpusPacketHeader) + hdr.size);
         if (out_performance_time != nullptr) {
             *out_performance_time =
                 std::chrono::duration_cast<std::chrono::milliseconds>(end_time).count();
@@ -167,21 +161,66 @@ private:
         opus_decoder_ctl(decoder.get(), OPUS_RESET_STATE);
     }
 
-    struct OpusHeader {
-        u32_be sz; // Needs to be BE for some odd reason
-        INSERT_PADDING_WORDS(1);
-    };
-    static_assert(sizeof(OpusHeader) == 0x8, "OpusHeader is an invalid size");
-
-    std::unique_ptr<OpusDecoder, OpusDeleter> decoder;
+    OpusDecoderPtr decoder;
     u32 sample_rate;
     u32 channel_count;
 };
 
-static std::size_t WorkerBufferSize(u32 channel_count) {
+class IHardwareOpusDecoderManager final : public ServiceFramework<IHardwareOpusDecoderManager> {
+public:
+    explicit IHardwareOpusDecoderManager(std::unique_ptr<OpusDecoderStateBase> decoder_state)
+        : ServiceFramework("IHardwareOpusDecoderManager"), decoder_state{std::move(decoder_state)} {
+        // clang-format off
+        static const FunctionInfo functions[] = {
+            {0, &IHardwareOpusDecoderManager::DecodeInterleavedOld, "DecodeInterleavedOld"},
+            {1, nullptr, "SetContext"},
+            {2, nullptr, "DecodeInterleavedForMultiStreamOld"},
+            {3, nullptr, "SetContextForMultiStream"},
+            {4, &IHardwareOpusDecoderManager::DecodeInterleavedWithPerfOld, "DecodeInterleavedWithPerfOld"},
+            {5, nullptr, "DecodeInterleavedForMultiStreamWithPerfOld"},
+            {6, &IHardwareOpusDecoderManager::DecodeInterleaved, "DecodeInterleaved"},
+            {7, nullptr, "DecodeInterleavedForMultiStream"},
+        };
+        // clang-format on
+
+        RegisterHandlers(functions);
+    }
+
+private:
+    void DecodeInterleavedOld(Kernel::HLERequestContext& ctx) {
+        LOG_DEBUG(Audio, "called");
+
+        decoder_state->DecodeInterleaved(ctx, OpusDecoderStateBase::PerfTime::Disabled,
+                                         OpusDecoderStateBase::ExtraBehavior::None);
+    }
+
+    void DecodeInterleavedWithPerfOld(Kernel::HLERequestContext& ctx) {
+        LOG_DEBUG(Audio, "called");
+
+        decoder_state->DecodeInterleaved(ctx, OpusDecoderStateBase::PerfTime::Enabled,
+                                         OpusDecoderStateBase::ExtraBehavior::None);
+    }
+
+    void DecodeInterleaved(Kernel::HLERequestContext& ctx) {
+        LOG_DEBUG(Audio, "called");
+
+        IPC::RequestParser rp{ctx};
+        const auto extra_behavior = rp.Pop<bool>()
+                                        ? OpusDecoderStateBase::ExtraBehavior::ResetContext
+                                        : OpusDecoderStateBase::ExtraBehavior::None;
+
+        decoder_state->DecodeInterleaved(ctx, OpusDecoderStateBase::PerfTime::Enabled,
+                                         extra_behavior);
+    }
+
+    std::unique_ptr<OpusDecoderStateBase> decoder_state;
+};
+
+std::size_t WorkerBufferSize(u32 channel_count) {
     ASSERT_MSG(channel_count == 1 || channel_count == 2, "Invalid channel count");
     return opus_decoder_get_size(static_cast<int>(channel_count));
 }
+} // Anonymous namespace
 
 void HwOpus::GetWorkBufferSize(Kernel::HLERequestContext& ctx) {
     IPC::RequestParser rp{ctx};
@@ -220,8 +259,7 @@ void HwOpus::OpenOpusDecoder(Kernel::HLERequestContext& ctx) {
     const std::size_t worker_sz = WorkerBufferSize(channel_count);
     ASSERT_MSG(buffer_sz >= worker_sz, "Worker buffer too large");
 
-    std::unique_ptr<OpusDecoder, OpusDeleter> decoder{
-        static_cast<OpusDecoder*>(operator new(worker_sz))};
+    OpusDecoderPtr decoder{static_cast<OpusDecoder*>(operator new(worker_sz))};
     if (const int err = opus_decoder_init(decoder.get(), sample_rate, channel_count)) {
         LOG_ERROR(Audio, "Failed to init opus decoder with error={}", err);
         IPC::ResponseBuilder rb{ctx, 2};
@@ -232,8 +270,8 @@ void HwOpus::OpenOpusDecoder(Kernel::HLERequestContext& ctx) {
 
     IPC::ResponseBuilder rb{ctx, 2, 0, 1};
     rb.Push(RESULT_SUCCESS);
-    rb.PushIpcInterface<IHardwareOpusDecoderManager>(std::move(decoder), sample_rate,
-                                                     channel_count);
+    rb.PushIpcInterface<IHardwareOpusDecoderManager>(
+        std::make_unique<OpusDecoderState>(std::move(decoder), sample_rate, channel_count));
 }
 
 HwOpus::HwOpus() : ServiceFramework("hwopus") {

src/core/hle/service/nvdrv/devices/nvdisp_disp0.cpp

@@ -36,7 +36,7 @@ void nvdisp_disp0::flip(u32 buffer_handle, u32 offset, u32 format, u32 width, u3
 
     auto& instance = Core::System::GetInstance();
     instance.GetPerfStats().EndGameFrame();
-    instance.Renderer().SwapBuffers(framebuffer);
+    instance.GPU().SwapBuffers(framebuffer);
 }
 
 } // namespace Service::Nvidia::Devices

src/core/hle/service/nvdrv/devices/nvhost_as_gpu.cpp

@@ -178,7 +178,7 @@ u32 nvhost_as_gpu::UnmapBuffer(const std::vector<u8>& input, std::vector<u8>& ou
     auto& gpu = system_instance.GPU();
     auto cpu_addr = gpu.MemoryManager().GpuToCpuAddress(params.offset);
     ASSERT(cpu_addr);
-    system_instance.Renderer().Rasterizer().FlushAndInvalidateRegion(*cpu_addr, itr->second.size);
+    gpu.FlushAndInvalidateRegion(*cpu_addr, itr->second.size);
 
     params.offset = gpu.MemoryManager().UnmapBuffer(params.offset, itr->second.size);
 

src/core/hle/service/nvdrv/devices/nvhost_gpu.cpp

@@ -136,16 +136,6 @@ u32 nvhost_gpu::AllocateObjectContext(const std::vector<u8>& input, std::vector<
     return 0;
 }
 
-static void PushGPUEntries(Tegra::CommandList&& entries) {
-    if (entries.empty()) {
-        return;
-    }
-
-    auto& dma_pusher{Core::System::GetInstance().GPU().DmaPusher()};
-    dma_pusher.Push(std::move(entries));
-    dma_pusher.DispatchCalls();
-}
-
 u32 nvhost_gpu::SubmitGPFIFO(const std::vector<u8>& input, std::vector<u8>& output) {
     if (input.size() < sizeof(IoctlSubmitGpfifo)) {
         UNIMPLEMENTED();
@@ -163,7 +153,7 @@ u32 nvhost_gpu::SubmitGPFIFO(const std::vector<u8>& input, std::vector<u8>& outp
     std::memcpy(entries.data(), &input[sizeof(IoctlSubmitGpfifo)],
                 params.num_entries * sizeof(Tegra::CommandListHeader));
 
-    PushGPUEntries(std::move(entries));
+    Core::System::GetInstance().GPU().PushGPUEntries(std::move(entries));
 
     params.fence_out.id = 0;
     params.fence_out.value = 0;
@@ -184,7 +174,7 @@ u32 nvhost_gpu::KickoffPB(const std::vector<u8>& input, std::vector<u8>& output)
     Memory::ReadBlock(params.address, entries.data(),
                       params.num_entries * sizeof(Tegra::CommandListHeader));
 
-    PushGPUEntries(std::move(entries));
+    Core::System::GetInstance().GPU().PushGPUEntries(std::move(entries));
 
     params.fence_out.id = 0;
     params.fence_out.value = 0;

src/core/hle/service/nvflinger/nvflinger.cpp

@@ -186,7 +186,7 @@ void NVFlinger::Compose() {
 
             // There was no queued buffer to draw, render previous frame
             system_instance.GetPerfStats().EndGameFrame();
-            system_instance.Renderer().SwapBuffers({});
+            system_instance.GPU().SwapBuffers({});
             continue;
         }
 

src/core/hle/service/service.cpp

@@ -76,7 +76,8 @@ namespace Service {
  * Creates a function string for logging, complete with the name (or header code, depending
  * on what's passed in) the port name, and all the cmd_buff arguments.
  */
-[[maybe_unused]] static std::string MakeFunctionString(const char* name, const char* port_name,
+[[maybe_unused]] static std::string MakeFunctionString(std::string_view name,
+                                                       std::string_view port_name,
                                                        const u32* cmd_buff) {
     // Number of params == bits 0-5 + bits 6-11
     int num_params = (cmd_buff[0] & 0x3F) + ((cmd_buff[0] >> 6) & 0x3F);
@@ -158,9 +159,7 @@ void ServiceFrameworkBase::InvokeRequest(Kernel::HLERequestContext& ctx) {
         return ReportUnimplementedFunction(ctx, info);
     }
 
-    LOG_TRACE(
-        Service, "{}",
-        MakeFunctionString(info->name, GetServiceName().c_str(), ctx.CommandBuffer()).c_str());
+    LOG_TRACE(Service, "{}", MakeFunctionString(info->name, GetServiceName(), ctx.CommandBuffer()));
     handler_invoker(this, info->handler_callback, ctx);
 }
 

src/core/hle/service/sm/controller.cpp

@@ -30,7 +30,7 @@ void Controller::DuplicateSession(Kernel::HLERequestContext& ctx) {
 
     IPC::ResponseBuilder rb{ctx, 2, 0, 1, IPC::ResponseBuilder::Flags::AlwaysMoveHandles};
     rb.Push(RESULT_SUCCESS);
-    Kernel::SharedPtr<Kernel::ClientSession> session{ctx.Session()->parent->client};
+    Kernel::SharedPtr<Kernel::ClientSession> session{ctx.Session()->GetParent()->client};
     rb.PushMoveObjects(session);
 
     LOG_DEBUG(Service, "session={}", session->GetObjectId());

src/core/memory.cpp

@@ -171,9 +171,6 @@ T Read(const VAddr vaddr) {
         return value;
     }
 
-    // The memory access might do an MMIO or cached access, so we have to lock the HLE kernel state
-    std::lock_guard<std::recursive_mutex> lock(HLE::g_hle_lock);
-
     PageType type = current_page_table->attributes[vaddr >> PAGE_BITS];
     switch (type) {
     case PageType::Unmapped:
@@ -204,9 +201,6 @@ void Write(const VAddr vaddr, const T data) {
         return;
     }
 
-    // The memory access might do an MMIO or cached access, so we have to lock the HLE kernel state
-    std::lock_guard<std::recursive_mutex> lock(HLE::g_hle_lock);
-
     PageType type = current_page_table->attributes[vaddr >> PAGE_BITS];
     switch (type) {
     case PageType::Unmapped:
@@ -362,16 +356,16 @@ void RasterizerFlushVirtualRegion(VAddr start, u64 size, FlushMode mode) {
         const VAddr overlap_end = std::min(end, region_end);
         const VAddr overlap_size = overlap_end - overlap_start;
 
-        auto& rasterizer = system_instance.Renderer().Rasterizer();
+        auto& gpu = system_instance.GPU();
         switch (mode) {
         case FlushMode::Flush:
-            rasterizer.FlushRegion(overlap_start, overlap_size);
+            gpu.FlushRegion(overlap_start, overlap_size);
             break;
         case FlushMode::Invalidate:
-            rasterizer.InvalidateRegion(overlap_start, overlap_size);
+            gpu.InvalidateRegion(overlap_start, overlap_size);
             break;
         case FlushMode::FlushAndInvalidate:
-            rasterizer.FlushAndInvalidateRegion(overlap_start, overlap_size);
+            gpu.FlushAndInvalidateRegion(overlap_start, overlap_size);
             break;
         }
     };

src/core/settings.cpp

@@ -91,7 +91,10 @@ void LogSettings() {
     LogSetting("Renderer_UseResolutionFactor", Settings::values.resolution_factor);
     LogSetting("Renderer_UseFrameLimit", Settings::values.use_frame_limit);
     LogSetting("Renderer_FrameLimit", Settings::values.frame_limit);
+    LogSetting("Renderer_UseDiskShaderCache", Settings::values.use_disk_shader_cache);
     LogSetting("Renderer_UseAccurateGpuEmulation", Settings::values.use_accurate_gpu_emulation);
+    LogSetting("Renderer_UseAsynchronousGpuEmulation",
+               Settings::values.use_asynchronous_gpu_emulation);
     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

@@ -393,6 +393,7 @@ struct Values {
     u16 frame_limit;
     bool use_disk_shader_cache;
     bool use_accurate_gpu_emulation;
+    bool use_asynchronous_gpu_emulation;
 
     float bg_red;
     float bg_green;

src/core/telemetry_session.cpp

@@ -162,6 +162,8 @@ TelemetrySession::TelemetrySession() {
              Settings::values.use_disk_shader_cache);
     AddField(Telemetry::FieldType::UserConfig, "Renderer_UseAccurateGpuEmulation",
              Settings::values.use_accurate_gpu_emulation);
+    AddField(Telemetry::FieldType::UserConfig, "Renderer_UseAsynchronousGpuEmulation",
+             Settings::values.use_asynchronous_gpu_emulation);
     AddField(Telemetry::FieldType::UserConfig, "System_UseDockedMode",
              Settings::values.use_docked_mode);
 }

src/tests/core/arm/arm_test_common.cpp

@@ -13,11 +13,11 @@
 namespace ArmTests {
 
 TestEnvironment::TestEnvironment(bool mutable_memory_)
-    : mutable_memory(mutable_memory_), test_memory(std::make_shared<TestMemory>(this)) {
-
+    : mutable_memory(mutable_memory_),
+      test_memory(std::make_shared<TestMemory>(this)), kernel{Core::System::GetInstance()} {
     auto process = Kernel::Process::Create(kernel, "");
     kernel.MakeCurrentProcess(process.get());
-    page_table = &Core::CurrentProcess()->VMManager().page_table;
+    page_table = &process->VMManager().page_table;
 
     std::fill(page_table->pointers.begin(), page_table->pointers.end(), nullptr);
     page_table->special_regions.clear();

src/video_core/CMakeLists.txt

@@ -17,6 +17,12 @@ add_library(video_core STATIC
     engines/shader_header.h
     gpu.cpp
     gpu.h
+    gpu_asynch.cpp
+    gpu_asynch.h
+    gpu_synch.cpp
+    gpu_synch.h
+    gpu_thread.cpp
+    gpu_thread.h
     macro_interpreter.cpp
     macro_interpreter.h
     memory_manager.cpp
@@ -74,6 +80,7 @@ add_library(video_core STATIC
     shader/decode/hfma2.cpp
     shader/decode/conversion.cpp
     shader/decode/memory.cpp
+    shader/decode/texture.cpp
     shader/decode/float_set_predicate.cpp
     shader/decode/integer_set_predicate.cpp
     shader/decode/half_set_predicate.cpp
@@ -94,6 +101,8 @@ add_library(video_core STATIC
     surface.h
     textures/astc.cpp
     textures/astc.h
+    textures/convert.cpp
+    textures/convert.h
     textures/decoders.cpp
     textures/decoders.h
     textures/texture.h
@@ -104,6 +113,8 @@ add_library(video_core STATIC
 if (ENABLE_VULKAN)
     target_sources(video_core PRIVATE
         renderer_vulkan/declarations.h
+        renderer_vulkan/maxwell_to_vk.cpp
+        renderer_vulkan/maxwell_to_vk.h
         renderer_vulkan/vk_buffer_cache.cpp
         renderer_vulkan/vk_buffer_cache.h
         renderer_vulkan/vk_device.cpp

src/video_core/dma_pusher.cpp

@@ -39,7 +39,7 @@ bool DmaPusher::Step() {
     }
 
     const CommandList& command_list{dma_pushbuffer.front()};
-    const CommandListHeader& command_list_header{command_list[dma_pushbuffer_subindex++]};
+    const CommandListHeader command_list_header{command_list[dma_pushbuffer_subindex++]};
     GPUVAddr dma_get = command_list_header.addr;
     GPUVAddr dma_put = dma_get + command_list_header.size * sizeof(u32);
     bool non_main = command_list_header.is_non_main;

src/video_core/engines/fermi_2d.cpp

@@ -2,12 +2,11 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
-#include "core/core.h"
-#include "core/memory.h"
+#include "common/assert.h"
+#include "common/logging/log.h"
+#include "common/math_util.h"
 #include "video_core/engines/fermi_2d.h"
-#include "video_core/engines/maxwell_3d.h"
 #include "video_core/rasterizer_interface.h"
-#include "video_core/textures/decoders.h"
 
 namespace Tegra::Engines {
 

src/video_core/engines/fermi_2d.h

@@ -5,7 +5,7 @@
 #pragma once
 
 #include <array>
-#include "common/assert.h"
+#include <cstddef>
 #include "common/bit_field.h"
 #include "common/common_funcs.h"
 #include "common/common_types.h"

src/video_core/engines/kepler_compute.cpp

@@ -2,9 +2,8 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include "common/assert.h"
 #include "common/logging/log.h"
-#include "core/core.h"
-#include "core/memory.h"
 #include "video_core/engines/kepler_compute.h"
 #include "video_core/memory_manager.h"
 

src/video_core/engines/kepler_compute.h

@@ -5,8 +5,7 @@
 #pragma once
 
 #include <array>
-#include "common/assert.h"
-#include "common/bit_field.h"
+#include <cstddef>
 #include "common/common_funcs.h"
 #include "common/common_types.h"
 #include "video_core/gpu.h"

src/video_core/engines/kepler_memory.cpp

@@ -48,7 +48,7 @@ void KeplerMemory::ProcessData(u32 data) {
     // We have to invalidate the destination region to evict any outdated surfaces from the cache.
     // We do this before actually writing the new data because the destination address might contain
     // a dirty surface that will have to be written back to memory.
-    rasterizer.InvalidateRegion(*dest_address, sizeof(u32));
+    Core::System::GetInstance().GPU().InvalidateRegion(*dest_address, sizeof(u32));
 
     Memory::Write32(*dest_address, data);
     system.GPU().Maxwell3D().dirty_flags.OnMemoryWrite();

src/video_core/engines/kepler_memory.h

@@ -5,6 +5,7 @@
 #pragma once
 
 #include <array>
+#include <cstddef>
 #include "common/bit_field.h"
 #include "common/common_funcs.h"
 #include "common/common_types.h"

src/video_core/engines/maxwell_dma.cpp

@@ -3,6 +3,7 @@
 // Refer to the license.txt file included.
 
 #include "common/assert.h"
+#include "common/logging/log.h"
 #include "core/core.h"
 #include "core/memory.h"
 #include "video_core/engines/maxwell_3d.h"
@@ -91,12 +92,12 @@ void MaxwellDMA::HandleCopy() {
     const auto FlushAndInvalidate = [&](u32 src_size, u64 dst_size) {
         // TODO(Subv): For now, manually flush the regions until we implement GPU-accelerated
         // copying.
-        rasterizer.FlushRegion(*source_cpu, src_size);
+        Core::System::GetInstance().GPU().FlushRegion(*source_cpu, src_size);
 
         // We have to invalidate the destination region to evict any outdated surfaces from the
         // cache. We do this before actually writing the new data because the destination address
         // might contain a dirty surface that will have to be written back to memory.
-        rasterizer.InvalidateRegion(*dest_cpu, dst_size);
+        Core::System::GetInstance().GPU().InvalidateRegion(*dest_cpu, dst_size);
     };
 
     if (regs.exec.is_dst_linear && !regs.exec.is_src_linear) {

src/video_core/engines/maxwell_dma.h

@@ -5,6 +5,7 @@
 #pragma once
 
 #include <array>
+#include <cstddef>
 #include "common/bit_field.h"
 #include "common/common_funcs.h"
 #include "common/common_types.h"

src/video_core/engines/shader_bytecode.h

@@ -6,7 +6,6 @@
 
 #include <bitset>
 #include <optional>
-#include <string>
 #include <tuple>
 #include <vector>
 
@@ -325,11 +324,11 @@ enum class TextureQueryType : u64 {
 
 enum class TextureProcessMode : u64 {
     None = 0,
-    LZ = 1,  // Unknown, appears to be the same as none.
+    LZ = 1,  // Load LOD of zero.
     LB = 2,  // Load Bias.
-    LL = 3,  // Load LOD (LevelOfDetail)
-    LBA = 6, // Load Bias. The A is unknown, does not appear to differ with LB
-    LLA = 7  // Load LOD. The A is unknown, does not appear to differ with LL
+    LL = 3,  // Load LOD.
+    LBA = 6, // Load Bias. The A is unknown, does not appear to differ with LB.
+    LLA = 7  // Load LOD. The A is unknown, does not appear to differ with LL.
 };
 
 enum class TextureMiscMode : u64 {
@@ -1446,6 +1445,7 @@ public:
         Flow,
         Synch,
         Memory,
+        Texture,
         FloatSet,
         FloatSetPredicate,
         IntegerSet,
@@ -1576,14 +1576,14 @@ private:
             INST("1110111101010---", Id::ST_L, Type::Memory, "ST_L"),
             INST("1110111011010---", Id::LDG, Type::Memory, "LDG"),
             INST("1110111011011---", Id::STG, Type::Memory, "STG"),
-            INST("110000----111---", Id::TEX, Type::Memory, "TEX"),
-            INST("1101111101001---", Id::TXQ, Type::Memory, "TXQ"),
-            INST("1101-00---------", Id::TEXS, Type::Memory, "TEXS"),
-            INST("1101101---------", Id::TLDS, Type::Memory, "TLDS"),
-            INST("110010----111---", Id::TLD4, Type::Memory, "TLD4"),
-            INST("1101111100------", Id::TLD4S, Type::Memory, "TLD4S"),
-            INST("110111110110----", Id::TMML_B, Type::Memory, "TMML_B"),
-            INST("1101111101011---", Id::TMML, Type::Memory, "TMML"),
+            INST("110000----111---", Id::TEX, Type::Texture, "TEX"),
+            INST("1101111101001---", Id::TXQ, Type::Texture, "TXQ"),
+            INST("1101-00---------", Id::TEXS, Type::Texture, "TEXS"),
+            INST("1101101---------", Id::TLDS, Type::Texture, "TLDS"),
+            INST("110010----111---", Id::TLD4, Type::Texture, "TLD4"),
+            INST("1101111100------", Id::TLD4S, Type::Texture, "TLD4S"),
+            INST("110111110110----", Id::TMML_B, Type::Texture, "TMML_B"),
+            INST("1101111101011---", Id::TMML, Type::Texture, "TMML"),
             INST("111000110000----", Id::EXIT, Type::Trivial, "EXIT"),
             INST("11100000--------", Id::IPA, Type::Trivial, "IPA"),
             INST("1111101111100---", Id::OUT_R, Type::Trivial, "OUT_R"),

src/video_core/gpu.cpp

@@ -12,7 +12,7 @@
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/engines/maxwell_dma.h"
 #include "video_core/gpu.h"
-#include "video_core/rasterizer_interface.h"
+#include "video_core/renderer_base.h"
 
 namespace Tegra {
 
@@ -28,7 +28,8 @@ u32 FramebufferConfig::BytesPerPixel(PixelFormat format) {
     UNREACHABLE();
 }
 
-GPU::GPU(Core::System& system, VideoCore::RasterizerInterface& rasterizer) {
+GPU::GPU(Core::System& system, VideoCore::RendererBase& renderer) : renderer{renderer} {
+    auto& rasterizer{renderer.Rasterizer()};
     memory_manager = std::make_unique<Tegra::MemoryManager>();
     dma_pusher = std::make_unique<Tegra::DmaPusher>(*this);
     maxwell_3d = std::make_unique<Engines::Maxwell3D>(system, rasterizer, *memory_manager);

src/video_core/gpu.h

@@ -16,8 +16,8 @@ class System;
 }
 
 namespace VideoCore {
-class RasterizerInterface;
-}
+class RendererBase;
+} // namespace VideoCore
 
 namespace Tegra {
 
@@ -119,10 +119,11 @@ enum class EngineID {
     MAXWELL_DMA_COPY_A = 0xB0B5,
 };
 
-class GPU final {
+class GPU {
 public:
-    explicit GPU(Core::System& system, VideoCore::RasterizerInterface& rasterizer);
-    ~GPU();
+    explicit GPU(Core::System& system, VideoCore::RendererBase& renderer);
+
+    virtual ~GPU();
 
     struct MethodCall {
         u32 method{};
@@ -200,8 +201,42 @@ public:
         };
     } regs{};
 
+    /// Push GPU command entries to be processed
+    virtual void PushGPUEntries(Tegra::CommandList&& entries) = 0;
+
+    /// Swap buffers (render frame)
+    virtual void SwapBuffers(
+        std::optional<std::reference_wrapper<const Tegra::FramebufferConfig>> framebuffer) = 0;
+
+    /// Notify rasterizer that any caches of the specified region should be flushed to Switch memory
+    virtual void FlushRegion(VAddr addr, u64 size) = 0;
+
+    /// Notify rasterizer that any caches of the specified region should be invalidated
+    virtual void InvalidateRegion(VAddr addr, u64 size) = 0;
+
+    /// Notify rasterizer that any caches of the specified region should be flushed and invalidated
+    virtual void FlushAndInvalidateRegion(VAddr addr, u64 size) = 0;
+
 private:
+    void ProcessBindMethod(const MethodCall& method_call);
+    void ProcessSemaphoreTriggerMethod();
+    void ProcessSemaphoreRelease();
+    void ProcessSemaphoreAcquire();
+
+    /// Calls a GPU puller method.
+    void CallPullerMethod(const MethodCall& method_call);
+
+    /// Calls a GPU engine method.
+    void CallEngineMethod(const MethodCall& method_call);
+
+    /// Determines where the method should be executed.
+    bool ExecuteMethodOnEngine(const MethodCall& method_call);
+
+protected:
     std::unique_ptr<Tegra::DmaPusher> dma_pusher;
+    VideoCore::RendererBase& renderer;
+
+private:
     std::unique_ptr<Tegra::MemoryManager> memory_manager;
 
     /// Mapping of command subchannels to their bound engine ids.
@@ -217,18 +252,6 @@ private:
     std::unique_ptr<Engines::MaxwellDMA> maxwell_dma;
     /// Inline memory engine
     std::unique_ptr<Engines::KeplerMemory> kepler_memory;
-
-    void ProcessBindMethod(const MethodCall& method_call);
-    void ProcessSemaphoreTriggerMethod();
-    void ProcessSemaphoreRelease();
-    void ProcessSemaphoreAcquire();
-
-    // Calls a GPU puller method.
-    void CallPullerMethod(const MethodCall& method_call);
-    // Calls a GPU engine method.
-    void CallEngineMethod(const MethodCall& method_call);
-    // Determines where the method should be executed.
-    bool ExecuteMethodOnEngine(const MethodCall& method_call);
 };
 
 #define ASSERT_REG_POSITION(field_name, position)                                                  \

src/video_core/gpu_asynch.cpp

@@ -0,0 +1,37 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "video_core/gpu_asynch.h"
+#include "video_core/gpu_thread.h"
+#include "video_core/renderer_base.h"
+
+namespace VideoCommon {
+
+GPUAsynch::GPUAsynch(Core::System& system, VideoCore::RendererBase& renderer)
+    : Tegra::GPU(system, renderer), gpu_thread{renderer, *dma_pusher} {}
+
+GPUAsynch::~GPUAsynch() = default;
+
+void GPUAsynch::PushGPUEntries(Tegra::CommandList&& entries) {
+    gpu_thread.SubmitList(std::move(entries));
+}
+
+void GPUAsynch::SwapBuffers(
+    std::optional<std::reference_wrapper<const Tegra::FramebufferConfig>> framebuffer) {
+    gpu_thread.SwapBuffers(std::move(framebuffer));
+}
+
+void GPUAsynch::FlushRegion(VAddr addr, u64 size) {
+    gpu_thread.FlushRegion(addr, size);
+}
+
+void GPUAsynch::InvalidateRegion(VAddr addr, u64 size) {
+    gpu_thread.InvalidateRegion(addr, size);
+}
+
+void GPUAsynch::FlushAndInvalidateRegion(VAddr addr, u64 size) {
+    gpu_thread.FlushAndInvalidateRegion(addr, size);
+}
+
+} // namespace VideoCommon

src/video_core/gpu_asynch.h

@@ -0,0 +1,37 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "video_core/gpu.h"
+#include "video_core/gpu_thread.h"
+
+namespace VideoCore {
+class RendererBase;
+} // namespace VideoCore
+
+namespace VideoCommon {
+
+namespace GPUThread {
+class ThreadManager;
+} // namespace GPUThread
+
+/// Implementation of GPU interface that runs the GPU asynchronously
+class GPUAsynch : public Tegra::GPU {
+public:
+    explicit GPUAsynch(Core::System& system, VideoCore::RendererBase& renderer);
+    ~GPUAsynch() override;
+
+    void PushGPUEntries(Tegra::CommandList&& entries) override;
+    void SwapBuffers(
+        std::optional<std::reference_wrapper<const Tegra::FramebufferConfig>> framebuffer) override;
+    void FlushRegion(VAddr addr, u64 size) override;
+    void InvalidateRegion(VAddr addr, u64 size) override;
+    void FlushAndInvalidateRegion(VAddr addr, u64 size) override;
+
+private:
+    GPUThread::ThreadManager gpu_thread;
+};
+
+} // namespace VideoCommon

src/video_core/gpu_synch.cpp

@@ -0,0 +1,37 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "video_core/gpu_synch.h"
+#include "video_core/renderer_base.h"
+
+namespace VideoCommon {
+
+GPUSynch::GPUSynch(Core::System& system, VideoCore::RendererBase& renderer)
+    : Tegra::GPU(system, renderer) {}
+
+GPUSynch::~GPUSynch() = default;
+
+void GPUSynch::PushGPUEntries(Tegra::CommandList&& entries) {
+    dma_pusher->Push(std::move(entries));
+    dma_pusher->DispatchCalls();
+}
+
+void GPUSynch::SwapBuffers(
+    std::optional<std::reference_wrapper<const Tegra::FramebufferConfig>> framebuffer) {
+    renderer.SwapBuffers(std::move(framebuffer));
+}
+
+void GPUSynch::FlushRegion(VAddr addr, u64 size) {
+    renderer.Rasterizer().FlushRegion(addr, size);
+}
+
+void GPUSynch::InvalidateRegion(VAddr addr, u64 size) {
+    renderer.Rasterizer().InvalidateRegion(addr, size);
+}
+
+void GPUSynch::FlushAndInvalidateRegion(VAddr addr, u64 size) {
+    renderer.Rasterizer().FlushAndInvalidateRegion(addr, size);
+}
+
+} // namespace VideoCommon

src/video_core/gpu_synch.h

@@ -0,0 +1,29 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "video_core/gpu.h"
+
+namespace VideoCore {
+class RendererBase;
+} // namespace VideoCore
+
+namespace VideoCommon {
+
+/// Implementation of GPU interface that runs the GPU synchronously
+class GPUSynch : public Tegra::GPU {
+public:
+    explicit GPUSynch(Core::System& system, VideoCore::RendererBase& renderer);
+    ~GPUSynch() override;
+
+    void PushGPUEntries(Tegra::CommandList&& entries) override;
+    void SwapBuffers(
+        std::optional<std::reference_wrapper<const Tegra::FramebufferConfig>> framebuffer) override;
+    void FlushRegion(VAddr addr, u64 size) override;
+    void InvalidateRegion(VAddr addr, u64 size) override;
+    void FlushAndInvalidateRegion(VAddr addr, u64 size) override;
+};
+
+} // namespace VideoCommon

src/video_core/gpu_thread.cpp

@@ -0,0 +1,152 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "common/microprofile.h"
+#include "core/frontend/scope_acquire_window_context.h"
+#include "core/settings.h"
+#include "video_core/dma_pusher.h"
+#include "video_core/gpu.h"
+#include "video_core/gpu_thread.h"
+#include "video_core/renderer_base.h"
+
+namespace VideoCommon::GPUThread {
+
+/// Executes a single GPU thread command
+static void ExecuteCommand(CommandData* command, VideoCore::RendererBase& renderer,
+                           Tegra::DmaPusher& dma_pusher) {
+    if (const auto submit_list = std::get_if<SubmitListCommand>(command)) {
+        dma_pusher.Push(std::move(submit_list->entries));
+        dma_pusher.DispatchCalls();
+    } else if (const auto data = std::get_if<SwapBuffersCommand>(command)) {
+        renderer.SwapBuffers(data->framebuffer);
+    } else if (const auto data = std::get_if<FlushRegionCommand>(command)) {
+        renderer.Rasterizer().FlushRegion(data->addr, data->size);
+    } else if (const auto data = std::get_if<InvalidateRegionCommand>(command)) {
+        renderer.Rasterizer().InvalidateRegion(data->addr, data->size);
+    } else if (const auto data = std::get_if<FlushAndInvalidateRegionCommand>(command)) {
+        renderer.Rasterizer().FlushAndInvalidateRegion(data->addr, data->size);
+    } else {
+        UNREACHABLE();
+    }
+}
+
+/// Runs the GPU thread
+static void RunThread(VideoCore::RendererBase& renderer, Tegra::DmaPusher& dma_pusher,
+                      SynchState& state) {
+
+    MicroProfileOnThreadCreate("GpuThread");
+
+    auto WaitForWakeup = [&]() {
+        std::unique_lock<std::mutex> lock{state.signal_mutex};
+        state.signal_condition.wait(lock, [&] { return !state.is_idle || !state.is_running; });
+    };
+
+    // Wait for first GPU command before acquiring the window context
+    WaitForWakeup();
+
+    // If emulation was stopped during disk shader loading, abort before trying to acquire context
+    if (!state.is_running) {
+        return;
+    }
+
+    Core::Frontend::ScopeAcquireWindowContext acquire_context{renderer.GetRenderWindow()};
+
+    while (state.is_running) {
+        if (!state.is_running) {
+            return;
+        }
+
+        {
+            // Thread has been woken up, so make the previous write queue the next read queue
+            std::lock_guard<std::mutex> lock{state.signal_mutex};
+            std::swap(state.push_queue, state.pop_queue);
+        }
+
+        // Execute all of the GPU commands
+        while (!state.pop_queue->empty()) {
+            ExecuteCommand(&state.pop_queue->front(), renderer, dma_pusher);
+            state.pop_queue->pop();
+        }
+
+        state.UpdateIdleState();
+
+        // Signal that the GPU thread has finished processing commands
+        if (state.is_idle) {
+            state.idle_condition.notify_one();
+        }
+
+        // Wait for CPU thread to send more GPU commands
+        WaitForWakeup();
+    }
+}
+
+ThreadManager::ThreadManager(VideoCore::RendererBase& renderer, Tegra::DmaPusher& dma_pusher)
+    : renderer{renderer}, dma_pusher{dma_pusher}, thread{RunThread, std::ref(renderer),
+                                                         std::ref(dma_pusher), std::ref(state)},
+      thread_id{thread.get_id()} {}
+
+ThreadManager::~ThreadManager() {
+    {
+        // Notify GPU thread that a shutdown is pending
+        std::lock_guard<std::mutex> lock{state.signal_mutex};
+        state.is_running = false;
+    }
+
+    state.signal_condition.notify_one();
+    thread.join();
+}
+
+void ThreadManager::SubmitList(Tegra::CommandList&& entries) {
+    if (entries.empty()) {
+        return;
+    }
+
+    PushCommand(SubmitListCommand(std::move(entries)), false, false);
+}
+
+void ThreadManager::SwapBuffers(
+    std::optional<std::reference_wrapper<const Tegra::FramebufferConfig>> framebuffer) {
+    PushCommand(SwapBuffersCommand(std::move(framebuffer)), true, false);
+}
+
+void ThreadManager::FlushRegion(VAddr addr, u64 size) {
+    // Block the CPU when using accurate emulation
+    PushCommand(FlushRegionCommand(addr, size), Settings::values.use_accurate_gpu_emulation, false);
+}
+
+void ThreadManager::InvalidateRegion(VAddr addr, u64 size) {
+    PushCommand(InvalidateRegionCommand(addr, size), true, true);
+}
+
+void ThreadManager::FlushAndInvalidateRegion(VAddr addr, u64 size) {
+    InvalidateRegion(addr, size);
+}
+
+void ThreadManager::PushCommand(CommandData&& command_data, bool wait_for_idle, bool allow_on_cpu) {
+    {
+        std::lock_guard<std::mutex> lock{state.signal_mutex};
+
+        if ((allow_on_cpu && state.is_idle) || IsGpuThread()) {
+            // Execute the command synchronously on the current thread
+            ExecuteCommand(&command_data, renderer, dma_pusher);
+            return;
+        }
+
+        // Push the command to the GPU thread
+        state.UpdateIdleState();
+        state.push_queue->emplace(command_data);
+    }
+
+    // Signal the GPU thread that commands are pending
+    state.signal_condition.notify_one();
+
+    if (wait_for_idle) {
+        // Wait for the GPU to be idle (all commands to be executed)
+        std::unique_lock<std::mutex> lock{state.idle_mutex};
+        state.idle_condition.wait(lock, [this] { return static_cast<bool>(state.is_idle); });
+    }
+}
+
+} // namespace VideoCommon::GPUThread

src/video_core/gpu_thread.h

@@ -0,0 +1,133 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+#include <atomic>
+#include <condition_variable>
+#include <memory>
+#include <mutex>
+#include <optional>
+#include <thread>
+#include <variant>
+
+namespace Tegra {
+struct FramebufferConfig;
+class DmaPusher;
+} // namespace Tegra
+
+namespace VideoCore {
+class RendererBase;
+} // namespace VideoCore
+
+namespace VideoCommon::GPUThread {
+
+/// Command to signal to the GPU thread that a command list is ready for processing
+struct SubmitListCommand final {
+    explicit SubmitListCommand(Tegra::CommandList&& entries) : entries{std::move(entries)} {}
+
+    Tegra::CommandList entries;
+};
+
+/// Command to signal to the GPU thread that a swap buffers is pending
+struct SwapBuffersCommand final {
+    explicit SwapBuffersCommand(std::optional<const Tegra::FramebufferConfig> framebuffer)
+        : framebuffer{std::move(framebuffer)} {}
+
+    std::optional<const Tegra::FramebufferConfig> framebuffer;
+};
+
+/// Command to signal to the GPU thread to flush a region
+struct FlushRegionCommand final {
+    explicit constexpr FlushRegionCommand(VAddr addr, u64 size) : addr{addr}, size{size} {}
+
+    const VAddr addr;
+    const u64 size;
+};
+
+/// Command to signal to the GPU thread to invalidate a region
+struct InvalidateRegionCommand final {
+    explicit constexpr InvalidateRegionCommand(VAddr addr, u64 size) : addr{addr}, size{size} {}
+
+    const VAddr addr;
+    const u64 size;
+};
+
+/// Command to signal to the GPU thread to flush and invalidate a region
+struct FlushAndInvalidateRegionCommand final {
+    explicit constexpr FlushAndInvalidateRegionCommand(VAddr addr, u64 size)
+        : addr{addr}, size{size} {}
+
+    const VAddr addr;
+    const u64 size;
+};
+
+using CommandData = std::variant<SubmitListCommand, SwapBuffersCommand, FlushRegionCommand,
+                                 InvalidateRegionCommand, FlushAndInvalidateRegionCommand>;
+
+/// Struct used to synchronize the GPU thread
+struct SynchState final {
+    std::atomic<bool> is_running{true};
+    std::atomic<bool> is_idle{true};
+    std::condition_variable signal_condition;
+    std::mutex signal_mutex;
+    std::condition_variable idle_condition;
+    std::mutex idle_mutex;
+
+    // We use two queues for sending commands to the GPU thread, one for writing (push_queue) to and
+    // one for reading from (pop_queue). These are swapped whenever the current pop_queue becomes
+    // empty. This allows for efficient thread-safe access, as it does not require any copies.
+
+    using CommandQueue = std::queue<CommandData>;
+    std::array<CommandQueue, 2> command_queues;
+    CommandQueue* push_queue{&command_queues[0]};
+    CommandQueue* pop_queue{&command_queues[1]};
+
+    void UpdateIdleState() {
+        std::lock_guard<std::mutex> lock{idle_mutex};
+        is_idle = command_queues[0].empty() && command_queues[1].empty();
+    }
+};
+
+/// Class used to manage the GPU thread
+class ThreadManager final {
+public:
+    explicit ThreadManager(VideoCore::RendererBase& renderer, Tegra::DmaPusher& dma_pusher);
+    ~ThreadManager();
+
+    /// Push GPU command entries to be processed
+    void SubmitList(Tegra::CommandList&& entries);
+
+    /// Swap buffers (render frame)
+    void SwapBuffers(
+        std::optional<std::reference_wrapper<const Tegra::FramebufferConfig>> framebuffer);
+
+    /// Notify rasterizer that any caches of the specified region should be flushed to Switch memory
+    void FlushRegion(VAddr addr, u64 size);
+
+    /// Notify rasterizer that any caches of the specified region should be invalidated
+    void InvalidateRegion(VAddr addr, u64 size);
+
+    /// Notify rasterizer that any caches of the specified region should be flushed and invalidated
+    void FlushAndInvalidateRegion(VAddr addr, u64 size);
+
+private:
+    /// Pushes a command to be executed by the GPU thread
+    void PushCommand(CommandData&& command_data, bool wait_for_idle, bool allow_on_cpu);
+
+    /// Returns true if this is called by the GPU thread
+    bool IsGpuThread() const {
+        return std::this_thread::get_id() == thread_id;
+    }
+
+private:
+    SynchState state;
+    VideoCore::RendererBase& renderer;
+    Tegra::DmaPusher& dma_pusher;
+    std::thread thread;
+    std::thread::id thread_id;
+};
+
+} // namespace VideoCommon::GPUThread

src/video_core/renderer_base.cpp

@@ -2,6 +2,7 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include "common/logging/log.h"
 #include "core/frontend/emu_window.h"
 #include "core/settings.h"
 #include "video_core/renderer_base.h"

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -118,7 +118,7 @@ RasterizerOpenGL::RasterizerOpenGL(Core::Frontend::EmuWindow& window, Core::Syst
 
     glGetIntegerv(GL_UNIFORM_BUFFER_OFFSET_ALIGNMENT, &uniform_buffer_alignment);
 
-    LOG_CRITICAL(Render_OpenGL, "Sync fixed function OpenGL state here!");
+    LOG_DEBUG(Render_OpenGL, "Sync fixed function OpenGL state here");
     CheckExtensions();
 }
 
@@ -177,7 +177,7 @@ GLuint RasterizerOpenGL::SetupVertexFormat() {
                 continue;
 
             const auto& buffer = regs.vertex_array[attrib.buffer];
-            LOG_TRACE(HW_GPU,
+            LOG_TRACE(Render_OpenGL,
                       "vertex attrib {}, count={}, size={}, type={}, offset={}, normalize={}",
                       index, attrib.ComponentCount(), attrib.SizeString(), attrib.TypeString(),
                       attrib.offset.Value(), attrib.IsNormalized());
@@ -343,9 +343,8 @@ void RasterizerOpenGL::SetupShaders(GLenum primitive_mode) {
             shader_program_manager->UseProgrammableFragmentShader(program_handle);
             break;
         default:
-            LOG_CRITICAL(HW_GPU, "Unimplemented shader index={}, enable={}, offset=0x{:08X}", index,
-                         shader_config.enable.Value(), shader_config.offset);
-            UNREACHABLE();
+            UNIMPLEMENTED_MSG("Unimplemented shader index={}, enable={}, offset=0x{:08X}", index,
+                              shader_config.enable.Value(), shader_config.offset);
         }
 
         const auto stage_enum = static_cast<Maxwell::ShaderStage>(stage);
@@ -739,33 +738,17 @@ void RasterizerOpenGL::DrawArrays() {
     state.Apply();
 
     res_cache.SignalPreDrawCall();
-
-    // Execute draw call
     params.DispatchDraw();
-
     res_cache.SignalPostDrawCall();
 
-    // Disable scissor test
-    state.viewports[0].scissor.enabled = false;
-
     accelerate_draw = AccelDraw::Disabled;
-
-    // Unbind textures for potential future use as framebuffer attachments
-    for (auto& texture_unit : state.texture_units) {
-        texture_unit.Unbind();
-    }
-    state.Apply();
 }
 
 void RasterizerOpenGL::FlushAll() {}
 
 void RasterizerOpenGL::FlushRegion(VAddr addr, u64 size) {
     MICROPROFILE_SCOPE(OpenGL_CacheManagement);
-
-    if (Settings::values.use_accurate_gpu_emulation) {
-        // Only flush if use_accurate_gpu_emulation is enabled, as it incurs a performance hit
-        res_cache.FlushRegion(addr, size);
-    }
+    res_cache.FlushRegion(addr, size);
 }
 
 void RasterizerOpenGL::InvalidateRegion(VAddr addr, u64 size) {
@@ -809,7 +792,10 @@ bool RasterizerOpenGL::AccelerateDisplay(const Tegra::FramebufferConfig& config,
         VideoCore::Surface::PixelFormatFromGPUPixelFormat(config.pixel_format)};
     ASSERT_MSG(params.width == config.width, "Framebuffer width is different");
     ASSERT_MSG(params.height == config.height, "Framebuffer height is different");
-    ASSERT_MSG(params.pixel_format == pixel_format, "Framebuffer pixel_format is different");
+
+    if (params.pixel_format != pixel_format) {
+        LOG_WARNING(Render_OpenGL, "Framebuffer pixel_format is different");
+    }
 
     screen_info.display_texture = surface->Texture().handle;
 
@@ -955,8 +941,8 @@ void RasterizerOpenGL::SetupConstBuffers(Tegra::Engines::Maxwell3D::Regs::Shader
             size = buffer.size;
 
             if (size > MaxConstbufferSize) {
-                LOG_CRITICAL(HW_GPU, "indirect constbuffer size {} exceeds maximum {}", size,
-                             MaxConstbufferSize);
+                LOG_WARNING(Render_OpenGL, "Indirect constbuffer size {} exceeds maximum {}", size,
+                            MaxConstbufferSize);
                 size = MaxConstbufferSize;
             }
         } else {
@@ -1016,10 +1002,9 @@ void RasterizerOpenGL::SetupTextures(Maxwell::ShaderStage stage, const Shader& s
 
         texture_samplers[current_bindpoint].SyncWithConfig(texture.tsc);
 
-        Surface surface = res_cache.GetTextureSurface(texture, entry);
-        if (surface != nullptr) {
+        if (Surface surface = res_cache.GetTextureSurface(texture, entry); surface) {
             state.texture_units[current_bindpoint].texture =
-                entry.IsArray() ? surface->TextureLayer().handle : surface->Texture().handle;
+                surface->Texture(entry.IsArray()).handle;
             surface->UpdateSwizzle(texture.tic.x_source, texture.tic.y_source, texture.tic.z_source,
                                    texture.tic.w_source);
         } else {
@@ -1251,11 +1236,7 @@ void RasterizerOpenGL::SyncScissorTest(OpenGLState& current_state) {
 
 void RasterizerOpenGL::SyncTransformFeedback() {
     const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
-
-    if (regs.tfb_enabled != 0) {
-        LOG_CRITICAL(Render_OpenGL, "Transform feedbacks are not implemented");
-        UNREACHABLE();
-    }
+    UNIMPLEMENTED_IF_MSG(regs.tfb_enabled != 0, "Transform feedbacks are not implemented");
 }
 
 void RasterizerOpenGL::SyncPointState() {
@@ -1275,12 +1256,8 @@ void RasterizerOpenGL::SyncPolygonOffset() {
 
 void RasterizerOpenGL::CheckAlphaTests() {
     const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
-
-    if (regs.alpha_test_enabled != 0 && regs.rt_control.count > 1) {
-        LOG_CRITICAL(Render_OpenGL, "Alpha Testing is enabled with Multiple Render Targets, "
-                                    "this behavior is undefined.");
-        UNREACHABLE();
-    }
+    UNIMPLEMENTED_IF_MSG(regs.alpha_test_enabled != 0 && regs.rt_control.count > 1,
+                         "Alpha Testing is enabled with more than one rendertarget");
 }
 
 } // namespace OpenGL

src/video_core/renderer_opengl/gl_rasterizer_cache.cpp

@@ -21,7 +21,7 @@
 #include "video_core/renderer_opengl/gl_rasterizer_cache.h"
 #include "video_core/renderer_opengl/utils.h"
 #include "video_core/surface.h"
-#include "video_core/textures/astc.h"
+#include "video_core/textures/convert.h"
 #include "video_core/textures/decoders.h"
 
 namespace OpenGL {
@@ -400,6 +400,27 @@ static const FormatTuple& GetFormatTuple(PixelFormat pixel_format, ComponentType
     return format;
 }
 
+/// Returns the discrepant array target
+constexpr GLenum GetArrayDiscrepantTarget(SurfaceTarget target) {
+    switch (target) {
+    case SurfaceTarget::Texture1D:
+        return GL_TEXTURE_1D_ARRAY;
+    case SurfaceTarget::Texture2D:
+        return GL_TEXTURE_2D_ARRAY;
+    case SurfaceTarget::Texture3D:
+        return GL_NONE;
+    case SurfaceTarget::Texture1DArray:
+        return GL_TEXTURE_1D;
+    case SurfaceTarget::Texture2DArray:
+        return GL_TEXTURE_2D;
+    case SurfaceTarget::TextureCubemap:
+        return GL_TEXTURE_CUBE_MAP_ARRAY;
+    case SurfaceTarget::TextureCubeArray:
+        return GL_TEXTURE_CUBE_MAP;
+    }
+    return GL_NONE;
+}
+
 Common::Rectangle<u32> SurfaceParams::GetRect(u32 mip_level) const {
     u32 actual_height{std::max(1U, unaligned_height >> mip_level)};
     if (IsPixelFormatASTC(pixel_format)) {
@@ -597,103 +618,6 @@ CachedSurface::CachedSurface(const SurfaceParams& params)
     }
 }
 
-static void ConvertS8Z24ToZ24S8(std::vector<u8>& data, u32 width, u32 height, bool reverse) {
-    union S8Z24 {
-        BitField<0, 24, u32> z24;
-        BitField<24, 8, u32> s8;
-    };
-    static_assert(sizeof(S8Z24) == 4, "S8Z24 is incorrect size");
-
-    union Z24S8 {
-        BitField<0, 8, u32> s8;
-        BitField<8, 24, u32> z24;
-    };
-    static_assert(sizeof(Z24S8) == 4, "Z24S8 is incorrect size");
-
-    S8Z24 s8z24_pixel{};
-    Z24S8 z24s8_pixel{};
-    constexpr auto bpp{GetBytesPerPixel(PixelFormat::S8Z24)};
-    for (std::size_t y = 0; y < height; ++y) {
-        for (std::size_t x = 0; x < width; ++x) {
-            const std::size_t offset{bpp * (y * width + x)};
-            if (reverse) {
-                std::memcpy(&z24s8_pixel, &data[offset], sizeof(Z24S8));
-                s8z24_pixel.s8.Assign(z24s8_pixel.s8);
-                s8z24_pixel.z24.Assign(z24s8_pixel.z24);
-                std::memcpy(&data[offset], &s8z24_pixel, sizeof(S8Z24));
-            } else {
-                std::memcpy(&s8z24_pixel, &data[offset], sizeof(S8Z24));
-                z24s8_pixel.s8.Assign(s8z24_pixel.s8);
-                z24s8_pixel.z24.Assign(s8z24_pixel.z24);
-                std::memcpy(&data[offset], &z24s8_pixel, sizeof(Z24S8));
-            }
-        }
-    }
-}
-
-/**
- * Helper function to perform software conversion (as needed) when loading a buffer from Switch
- * memory. This is for Maxwell pixel formats that cannot be represented as-is in OpenGL or with
- * typical desktop GPUs.
- */
-static void ConvertFormatAsNeeded_LoadGLBuffer(std::vector<u8>& data, PixelFormat pixel_format,
-                                               u32 width, u32 height, u32 depth) {
-    switch (pixel_format) {
-    case PixelFormat::ASTC_2D_4X4:
-    case PixelFormat::ASTC_2D_8X8:
-    case PixelFormat::ASTC_2D_8X5:
-    case PixelFormat::ASTC_2D_5X4:
-    case PixelFormat::ASTC_2D_5X5:
-    case PixelFormat::ASTC_2D_4X4_SRGB:
-    case PixelFormat::ASTC_2D_8X8_SRGB:
-    case PixelFormat::ASTC_2D_8X5_SRGB:
-    case PixelFormat::ASTC_2D_5X4_SRGB:
-    case PixelFormat::ASTC_2D_5X5_SRGB:
-    case PixelFormat::ASTC_2D_10X8:
-    case PixelFormat::ASTC_2D_10X8_SRGB: {
-        // Convert ASTC pixel formats to RGBA8, as most desktop GPUs do not support ASTC.
-        u32 block_width{};
-        u32 block_height{};
-        std::tie(block_width, block_height) = GetASTCBlockSize(pixel_format);
-        data =
-            Tegra::Texture::ASTC::Decompress(data, width, height, depth, block_width, block_height);
-        break;
-    }
-    case PixelFormat::S8Z24:
-        // Convert the S8Z24 depth format to Z24S8, as OpenGL does not support S8Z24.
-        ConvertS8Z24ToZ24S8(data, width, height, false);
-        break;
-    }
-}
-
-/**
- * Helper function to perform software conversion (as needed) when flushing a buffer from OpenGL to
- * Switch memory. This is for Maxwell pixel formats that cannot be represented as-is in OpenGL or
- * with typical desktop GPUs.
- */
-static void ConvertFormatAsNeeded_FlushGLBuffer(std::vector<u8>& data, PixelFormat pixel_format,
-                                                u32 width, u32 height) {
-    switch (pixel_format) {
-    case PixelFormat::ASTC_2D_4X4:
-    case PixelFormat::ASTC_2D_8X8:
-    case PixelFormat::ASTC_2D_4X4_SRGB:
-    case PixelFormat::ASTC_2D_8X8_SRGB:
-    case PixelFormat::ASTC_2D_5X5:
-    case PixelFormat::ASTC_2D_5X5_SRGB:
-    case PixelFormat::ASTC_2D_10X8:
-    case PixelFormat::ASTC_2D_10X8_SRGB: {
-        LOG_CRITICAL(HW_GPU, "Conversion of format {} after texture flushing is not implemented",
-                     static_cast<u32>(pixel_format));
-        UNREACHABLE();
-        break;
-    }
-    case PixelFormat::S8Z24:
-        // Convert the Z24S8 depth format to S8Z24, as OpenGL does not support S8Z24.
-        ConvertS8Z24ToZ24S8(data, width, height, true);
-        break;
-    }
-}
-
 MICROPROFILE_DEFINE(OpenGL_SurfaceLoad, "OpenGL", "Surface Load", MP_RGB(128, 192, 64));
 void CachedSurface::LoadGLBuffer() {
     MICROPROFILE_SCOPE(OpenGL_SurfaceLoad);
@@ -722,8 +646,16 @@ void CachedSurface::LoadGLBuffer() {
         }
     }
     for (u32 i = 0; i < params.max_mip_level; i++) {
-        ConvertFormatAsNeeded_LoadGLBuffer(gl_buffer[i], params.pixel_format, params.MipWidth(i),
-                                           params.MipHeight(i), params.MipDepth(i));
+        const u32 width = params.MipWidth(i);
+        const u32 height = params.MipHeight(i);
+        const u32 depth = params.MipDepth(i);
+        if (VideoCore::Surface::IsPixelFormatASTC(params.pixel_format)) {
+            // Reserve size for RGBA8 conversion
+            constexpr std::size_t rgba_bpp = 4;
+            gl_buffer[i].resize(std::max(gl_buffer[i].size(), width * height * depth * rgba_bpp));
+        }
+        Tegra::Texture::ConvertFromGuestToHost(gl_buffer[i].data(), params.pixel_format, width,
+                                               height, depth, true, true);
     }
 }
 
@@ -746,8 +678,8 @@ void CachedSurface::FlushGLBuffer() {
     glGetTextureImage(texture.handle, 0, tuple.format, tuple.type,
                       static_cast<GLsizei>(gl_buffer[0].size()), gl_buffer[0].data());
     glPixelStorei(GL_PACK_ROW_LENGTH, 0);
-    ConvertFormatAsNeeded_FlushGLBuffer(gl_buffer[0], params.pixel_format, params.width,
-                                        params.height);
+    Tegra::Texture::ConvertFromHostToGuest(gl_buffer[0].data(), params.pixel_format, params.width,
+                                           params.height, params.depth, true, true);
     const u8* const texture_src_data = Memory::GetPointer(params.addr);
     ASSERT(texture_src_data);
     if (params.is_tiled) {
@@ -884,20 +816,22 @@ void CachedSurface::UploadGLMipmapTexture(u32 mip_map, GLuint read_fb_handle,
     glPixelStorei(GL_UNPACK_ROW_LENGTH, 0);
 }
 
-void CachedSurface::EnsureTextureView() {
-    if (texture_view.handle != 0)
+void CachedSurface::EnsureTextureDiscrepantView() {
+    if (discrepant_view.handle != 0)
         return;
 
-    const GLenum target{TargetLayer()};
+    const GLenum target{GetArrayDiscrepantTarget(params.target)};
+    ASSERT(target != GL_NONE);
+
     const GLuint num_layers{target == GL_TEXTURE_CUBE_MAP_ARRAY ? 6u : 1u};
     constexpr GLuint min_layer = 0;
     constexpr GLuint min_level = 0;
 
-    glGenTextures(1, &texture_view.handle);
-    glTextureView(texture_view.handle, target, texture.handle, gl_internal_format, min_level,
+    glGenTextures(1, &discrepant_view.handle);
+    glTextureView(discrepant_view.handle, target, texture.handle, gl_internal_format, min_level,
                   params.max_mip_level, min_layer, num_layers);
-    ApplyTextureDefaults(texture_view.handle, params.max_mip_level);
-    glTextureParameteriv(texture_view.handle, GL_TEXTURE_SWIZZLE_RGBA,
+    ApplyTextureDefaults(discrepant_view.handle, params.max_mip_level);
+    glTextureParameteriv(discrepant_view.handle, GL_TEXTURE_SWIZZLE_RGBA,
                          reinterpret_cast<const GLint*>(swizzle.data()));
 }
 
@@ -923,8 +857,8 @@ void CachedSurface::UpdateSwizzle(Tegra::Texture::SwizzleSource swizzle_x,
     swizzle = {new_x, new_y, new_z, new_w};
     const auto swizzle_data = reinterpret_cast<const GLint*>(swizzle.data());
     glTextureParameteriv(texture.handle, GL_TEXTURE_SWIZZLE_RGBA, swizzle_data);
-    if (texture_view.handle != 0) {
-        glTextureParameteriv(texture_view.handle, GL_TEXTURE_SWIZZLE_RGBA, swizzle_data);
+    if (discrepant_view.handle != 0) {
+        glTextureParameteriv(discrepant_view.handle, GL_TEXTURE_SWIZZLE_RGBA, swizzle_data);
     }
 }
 

src/video_core/renderer_opengl/gl_rasterizer_cache.h

@@ -367,31 +367,19 @@ public:
         return texture;
     }
 
-    const OGLTexture& TextureLayer() {
-        if (params.is_array) {
-            return Texture();
+    const OGLTexture& Texture(bool as_array) {
+        if (params.is_array == as_array) {
+            return texture;
+        } else {
+            EnsureTextureDiscrepantView();
+            return discrepant_view;
         }
-        EnsureTextureView();
-        return texture_view;
     }
 
     GLenum Target() const {
         return gl_target;
     }
 
-    GLenum TargetLayer() const {
-        using VideoCore::Surface::SurfaceTarget;
-        switch (params.target) {
-        case SurfaceTarget::Texture1D:
-            return GL_TEXTURE_1D_ARRAY;
-        case SurfaceTarget::Texture2D:
-            return GL_TEXTURE_2D_ARRAY;
-        case SurfaceTarget::TextureCubemap:
-            return GL_TEXTURE_CUBE_MAP_ARRAY;
-        }
-        return Target();
-    }
-
     const SurfaceParams& GetSurfaceParams() const {
         return params;
     }
@@ -431,10 +419,10 @@ public:
 private:
     void UploadGLMipmapTexture(u32 mip_map, GLuint read_fb_handle, GLuint draw_fb_handle);
 
-    void EnsureTextureView();
+    void EnsureTextureDiscrepantView();
 
     OGLTexture texture;
-    OGLTexture texture_view;
+    OGLTexture discrepant_view;
     std::vector<std::vector<u8>> gl_buffer;
     SurfaceParams params{};
     GLenum gl_target{};

src/video_core/renderer_opengl/gl_shader_decompiler.cpp

@@ -5,7 +5,9 @@
 #include <array>
 #include <string>
 #include <string_view>
+#include <utility>
 #include <variant>
+#include <vector>
 
 #include <fmt/format.h>
 
@@ -717,7 +719,7 @@ private:
     }
 
     std::string GenerateTexture(Operation operation, const std::string& func,
-                                bool is_extra_int = false) {
+                                const std::vector<std::pair<Type, Node>>& extras) {
         constexpr std::array<const char*, 4> coord_constructors = {"float", "vec2", "vec3", "vec4"};
 
         const auto meta = std::get_if<MetaTexture>(&operation.GetMeta());
@@ -738,36 +740,47 @@ private:
             expr += Visit(operation[i]);
 
             const std::size_t next = i + 1;
-            if (next < count || has_array || has_shadow)
+            if (next < count)
                 expr += ", ";
         }
         if (has_array) {
-            expr += "float(ftoi(" + Visit(meta->array) + "))";
+            expr += ", float(ftoi(" + Visit(meta->array) + "))";
         }
         if (has_shadow) {
-            if (has_array)
-                expr += ", ";
-            expr += Visit(meta->depth_compare);
+            expr += ", " + Visit(meta->depth_compare);
         }
         expr += ')';
 
-        for (const Node extra : meta->extras) {
+        for (const auto& extra_pair : extras) {
+            const auto [type, operand] = extra_pair;
+            if (operand == nullptr) {
+                continue;
+            }
             expr += ", ";
-            if (is_extra_int) {
-                if (const auto immediate = std::get_if<ImmediateNode>(extra)) {
+
+            switch (type) {
+            case Type::Int:
+                if (const auto immediate = std::get_if<ImmediateNode>(operand)) {
                     // Inline the string as an immediate integer in GLSL (some extra arguments are
                     // required to be constant)
                     expr += std::to_string(static_cast<s32>(immediate->GetValue()));
                 } else {
-                    expr += "ftoi(" + Visit(extra) + ')';
+                    expr += "ftoi(" + Visit(operand) + ')';
                 }
-            } else {
-                expr += Visit(extra);
+                break;
+            case Type::Float:
+                expr += Visit(operand);
+                break;
+            default: {
+                const auto type_int = static_cast<u32>(type);
+                UNIMPLEMENTED_MSG("Unimplemented extra type={}", type_int);
+                expr += '0';
+                break;
+            }
             }
         }
 
-        expr += ')';
-        return expr;
+        return expr + ')';
     }
 
     std::string Assign(Operation operation) {
@@ -1146,7 +1159,7 @@ private:
         const auto meta = std::get_if<MetaTexture>(&operation.GetMeta());
         ASSERT(meta);
 
-        std::string expr = GenerateTexture(operation, "texture");
+        std::string expr = GenerateTexture(operation, "texture", {{Type::Float, meta->bias}});
         if (meta->sampler.IsShadow()) {
             expr = "vec4(" + expr + ')';
         }
@@ -1157,7 +1170,7 @@ private:
         const auto meta = std::get_if<MetaTexture>(&operation.GetMeta());
         ASSERT(meta);
 
-        std::string expr = GenerateTexture(operation, "textureLod");
+        std::string expr = GenerateTexture(operation, "textureLod", {{Type::Float, meta->lod}});
         if (meta->sampler.IsShadow()) {
             expr = "vec4(" + expr + ')';
         }
@@ -1168,7 +1181,8 @@ private:
         const auto meta = std::get_if<MetaTexture>(&operation.GetMeta());
         ASSERT(meta);
 
-        return GenerateTexture(operation, "textureGather", !meta->sampler.IsShadow()) +
+        const auto type = meta->sampler.IsShadow() ? Type::Float : Type::Int;
+        return GenerateTexture(operation, "textureGather", {{type, meta->component}}) +
                GetSwizzle(meta->element);
     }
 
@@ -1197,8 +1211,8 @@ private:
         ASSERT(meta);
 
         if (meta->element < 2) {
-            return "itof(int((" + GenerateTexture(operation, "textureQueryLod") + " * vec2(256))" +
-                   GetSwizzle(meta->element) + "))";
+            return "itof(int((" + GenerateTexture(operation, "textureQueryLod", {}) +
+                   " * vec2(256))" + GetSwizzle(meta->element) + "))";
         }
         return "0";
     }
@@ -1224,9 +1238,9 @@ private:
             else if (next < count)
                 expr += ", ";
         }
-        for (std::size_t i = 0; i < meta->extras.size(); ++i) {
+        if (meta->lod) {
             expr += ", ";
-            expr += CastOperand(Visit(meta->extras.at(i)), Type::Int);
+            expr += CastOperand(Visit(meta->lod), Type::Int);
         }
         expr += ')';
 

src/video_core/renderer_opengl/gl_state.cpp

@@ -461,7 +461,7 @@ void OpenGLState::ApplyTextures() const {
 
     if (has_delta) {
         glBindTextures(static_cast<GLuint>(first), static_cast<GLsizei>(last - first + 1),
-                       textures.data());
+                       textures.data() + first);
     }
 }
 
@@ -482,7 +482,7 @@ void OpenGLState::ApplySamplers() const {
     }
     if (has_delta) {
         glBindSamplers(static_cast<GLuint>(first), static_cast<GLsizei>(last - first + 1),
-                       samplers.data());
+                       samplers.data() + first);
     }
 }
 

src/video_core/renderer_opengl/renderer_opengl.cpp

@@ -244,6 +244,21 @@ void RendererOpenGL::InitOpenGLObjects() {
     LoadColorToActiveGLTexture(0, 0, 0, 0, screen_info.texture);
 }
 
+void RendererOpenGL::AddTelemetryFields() {
+    const char* const gl_version{reinterpret_cast<char const*>(glGetString(GL_VERSION))};
+    const char* const gpu_vendor{reinterpret_cast<char const*>(glGetString(GL_VENDOR))};
+    const char* const gpu_model{reinterpret_cast<char const*>(glGetString(GL_RENDERER))};
+
+    LOG_INFO(Render_OpenGL, "GL_VERSION: {}", gl_version);
+    LOG_INFO(Render_OpenGL, "GL_VENDOR: {}", gpu_vendor);
+    LOG_INFO(Render_OpenGL, "GL_RENDERER: {}", gpu_model);
+
+    auto& telemetry_session = system.TelemetrySession();
+    telemetry_session.AddField(Telemetry::FieldType::UserSystem, "GPU_Vendor", gpu_vendor);
+    telemetry_session.AddField(Telemetry::FieldType::UserSystem, "GPU_Model", gpu_model);
+    telemetry_session.AddField(Telemetry::FieldType::UserSystem, "GPU_OpenGL_Version", gl_version);
+}
+
 void RendererOpenGL::CreateRasterizer() {
     if (rasterizer) {
         return;
@@ -466,17 +481,7 @@ bool RendererOpenGL::Init() {
         glDebugMessageCallback(DebugHandler, nullptr);
     }
 
-    const char* gl_version{reinterpret_cast<char const*>(glGetString(GL_VERSION))};
-    const char* gpu_vendor{reinterpret_cast<char const*>(glGetString(GL_VENDOR))};
-    const char* gpu_model{reinterpret_cast<char const*>(glGetString(GL_RENDERER))};
-
-    LOG_INFO(Render_OpenGL, "GL_VERSION: {}", gl_version);
-    LOG_INFO(Render_OpenGL, "GL_VENDOR: {}", gpu_vendor);
-    LOG_INFO(Render_OpenGL, "GL_RENDERER: {}", gpu_model);
-
-    Core::Telemetry().AddField(Telemetry::FieldType::UserSystem, "GPU_Vendor", gpu_vendor);
-    Core::Telemetry().AddField(Telemetry::FieldType::UserSystem, "GPU_Model", gpu_model);
-    Core::Telemetry().AddField(Telemetry::FieldType::UserSystem, "GPU_OpenGL_Version", gl_version);
+    AddTelemetryFields();
 
     if (!GLAD_GL_VERSION_4_3) {
         return false;

src/video_core/renderer_opengl/renderer_opengl.h

@@ -60,6 +60,7 @@ public:
 
 private:
     void InitOpenGLObjects();
+    void AddTelemetryFields();
     void CreateRasterizer();
 
     void ConfigureFramebufferTexture(TextureInfo& texture,

src/video_core/renderer_vulkan/maxwell_to_vk.cpp

@@ -0,0 +1,483 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "common/logging/log.h"
+#include "video_core/engines/maxwell_3d.h"
+#include "video_core/renderer_vulkan/declarations.h"
+#include "video_core/renderer_vulkan/maxwell_to_vk.h"
+#include "video_core/renderer_vulkan/vk_device.h"
+#include "video_core/surface.h"
+
+namespace Vulkan::MaxwellToVK {
+
+namespace Sampler {
+
+vk::Filter Filter(Tegra::Texture::TextureFilter filter) {
+    switch (filter) {
+    case Tegra::Texture::TextureFilter::Linear:
+        return vk::Filter::eLinear;
+    case Tegra::Texture::TextureFilter::Nearest:
+        return vk::Filter::eNearest;
+    }
+    UNIMPLEMENTED_MSG("Unimplemented sampler filter={}", static_cast<u32>(filter));
+    return {};
+}
+
+vk::SamplerMipmapMode MipmapMode(Tegra::Texture::TextureMipmapFilter mipmap_filter) {
+    switch (mipmap_filter) {
+    case Tegra::Texture::TextureMipmapFilter::None:
+        // TODO(Rodrigo): None seems to be mapped to OpenGL's mag and min filters without mipmapping
+        // (e.g. GL_NEAREST and GL_LINEAR). Vulkan doesn't have such a thing, find out if we have to
+        // use an image view with a single mipmap level to emulate this.
+        return vk::SamplerMipmapMode::eLinear;
+    case Tegra::Texture::TextureMipmapFilter::Linear:
+        return vk::SamplerMipmapMode::eLinear;
+    case Tegra::Texture::TextureMipmapFilter::Nearest:
+        return vk::SamplerMipmapMode::eNearest;
+    }
+    UNIMPLEMENTED_MSG("Unimplemented sampler mipmap mode={}", static_cast<u32>(mipmap_filter));
+    return {};
+}
+
+vk::SamplerAddressMode WrapMode(Tegra::Texture::WrapMode wrap_mode) {
+    switch (wrap_mode) {
+    case Tegra::Texture::WrapMode::Wrap:
+        return vk::SamplerAddressMode::eRepeat;
+    case Tegra::Texture::WrapMode::Mirror:
+        return vk::SamplerAddressMode::eMirroredRepeat;
+    case Tegra::Texture::WrapMode::ClampToEdge:
+        return vk::SamplerAddressMode::eClampToEdge;
+    case Tegra::Texture::WrapMode::Border:
+        return vk::SamplerAddressMode::eClampToBorder;
+    case Tegra::Texture::WrapMode::ClampOGL:
+        // TODO(Rodrigo): GL_CLAMP was removed as of OpenGL 3.1, to implement GL_CLAMP, we can use
+        // eClampToBorder to get the border color of the texture, and then sample the edge to
+        // manually mix them. However the shader part of this is not yet implemented.
+        return vk::SamplerAddressMode::eClampToBorder;
+    case Tegra::Texture::WrapMode::MirrorOnceClampToEdge:
+        return vk::SamplerAddressMode::eMirrorClampToEdge;
+    case Tegra::Texture::WrapMode::MirrorOnceBorder:
+        UNIMPLEMENTED();
+        return vk::SamplerAddressMode::eMirrorClampToEdge;
+    }
+    UNIMPLEMENTED_MSG("Unimplemented wrap mode={}", static_cast<u32>(wrap_mode));
+    return {};
+}
+
+vk::CompareOp DepthCompareFunction(Tegra::Texture::DepthCompareFunc depth_compare_func) {
+    switch (depth_compare_func) {
+    case Tegra::Texture::DepthCompareFunc::Never:
+        return vk::CompareOp::eNever;
+    case Tegra::Texture::DepthCompareFunc::Less:
+        return vk::CompareOp::eLess;
+    case Tegra::Texture::DepthCompareFunc::LessEqual:
+        return vk::CompareOp::eLessOrEqual;
+    case Tegra::Texture::DepthCompareFunc::Equal:
+        return vk::CompareOp::eEqual;
+    case Tegra::Texture::DepthCompareFunc::NotEqual:
+        return vk::CompareOp::eNotEqual;
+    case Tegra::Texture::DepthCompareFunc::Greater:
+        return vk::CompareOp::eGreater;
+    case Tegra::Texture::DepthCompareFunc::GreaterEqual:
+        return vk::CompareOp::eGreaterOrEqual;
+    case Tegra::Texture::DepthCompareFunc::Always:
+        return vk::CompareOp::eAlways;
+    }
+    UNIMPLEMENTED_MSG("Unimplemented sampler depth compare function={}",
+                      static_cast<u32>(depth_compare_func));
+    return {};
+}
+
+} // namespace Sampler
+
+struct FormatTuple {
+    vk::Format format;            ///< Vulkan format
+    ComponentType component_type; ///< Abstracted component type
+    bool attachable;              ///< True when this format can be used as an attachment
+};
+
+static constexpr std::array<FormatTuple, VideoCore::Surface::MaxPixelFormat> tex_format_tuples = {{
+    {vk::Format::eA8B8G8R8UnormPack32, ComponentType::UNorm, true},    // ABGR8U
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // ABGR8S
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // ABGR8UI
+    {vk::Format::eB5G6R5UnormPack16, ComponentType::UNorm, false},     // B5G6R5U
+    {vk::Format::eA2B10G10R10UnormPack32, ComponentType::UNorm, true}, // A2B10G10R10U
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // A1B5G5R5U
+    {vk::Format::eR8Unorm, ComponentType::UNorm, true},                // R8U
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // R8UI
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // RGBA16F
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // RGBA16U
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // RGBA16UI
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // R11FG11FB10F
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // RGBA32UI
+    {vk::Format::eBc1RgbaUnormBlock, ComponentType::UNorm, false},     // DXT1
+    {vk::Format::eBc2UnormBlock, ComponentType::UNorm, false},         // DXT23
+    {vk::Format::eBc3UnormBlock, ComponentType::UNorm, false},         // DXT45
+    {vk::Format::eBc4UnormBlock, ComponentType::UNorm, false},         // DXN1
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // DXN2UNORM
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // DXN2SNORM
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // BC7U
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // BC6H_UF16
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // BC6H_SF16
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // ASTC_2D_4X4
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // BGRA8
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // RGBA32F
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // RG32F
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // R32F
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // R16F
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // R16U
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // R16S
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // R16UI
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // R16I
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // RG16
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // RG16F
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // RG16UI
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // RG16I
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // RG16S
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // RGB32F
+    {vk::Format::eA8B8G8R8SrgbPack32, ComponentType::UNorm, true},     // RGBA8_SRGB
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // RG8U
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // RG8S
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // RG32UI
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // R32UI
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // ASTC_2D_8X8
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // ASTC_2D_8X5
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // ASTC_2D_5X4
+
+    // Compressed sRGB formats
+    {vk::Format::eUndefined, ComponentType::Invalid, false}, // BGRA8_SRGB
+    {vk::Format::eUndefined, ComponentType::Invalid, false}, // DXT1_SRGB
+    {vk::Format::eUndefined, ComponentType::Invalid, false}, // DXT23_SRGB
+    {vk::Format::eUndefined, ComponentType::Invalid, false}, // DXT45_SRGB
+    {vk::Format::eUndefined, ComponentType::Invalid, false}, // BC7U_SRGB
+    {vk::Format::eUndefined, ComponentType::Invalid, false}, // ASTC_2D_4X4_SRGB
+    {vk::Format::eUndefined, ComponentType::Invalid, false}, // ASTC_2D_8X8_SRGB
+    {vk::Format::eUndefined, ComponentType::Invalid, false}, // ASTC_2D_8X5_SRGB
+    {vk::Format::eUndefined, ComponentType::Invalid, false}, // ASTC_2D_5X4_SRGB
+    {vk::Format::eUndefined, ComponentType::Invalid, false}, // ASTC_2D_5X5
+    {vk::Format::eUndefined, ComponentType::Invalid, false}, // ASTC_2D_5X5_SRGB
+    {vk::Format::eUndefined, ComponentType::Invalid, false}, // ASTC_2D_10X8
+    {vk::Format::eUndefined, ComponentType::Invalid, false}, // ASTC_2D_10X8_SRGB
+
+    // Depth formats
+    {vk::Format::eD32Sfloat, ComponentType::Float, true}, // Z32F
+    {vk::Format::eD16Unorm, ComponentType::UNorm, true},  // Z16
+
+    // DepthStencil formats
+    {vk::Format::eD24UnormS8Uint, ComponentType::UNorm, true}, // Z24S8
+    {vk::Format::eD24UnormS8Uint, ComponentType::UNorm, true}, // S8Z24 (emulated)
+    {vk::Format::eUndefined, ComponentType::Invalid, false},   // Z32FS8
+}};
+
+static constexpr bool IsZetaFormat(PixelFormat pixel_format) {
+    return pixel_format >= PixelFormat::MaxColorFormat &&
+           pixel_format < PixelFormat::MaxDepthStencilFormat;
+}
+
+std::pair<vk::Format, bool> SurfaceFormat(const VKDevice& device, FormatType format_type,
+                                          PixelFormat pixel_format, ComponentType component_type) {
+    ASSERT(static_cast<std::size_t>(pixel_format) < tex_format_tuples.size());
+
+    const auto tuple = tex_format_tuples[static_cast<u32>(pixel_format)];
+    UNIMPLEMENTED_IF_MSG(tuple.format == vk::Format::eUndefined,
+                         "Unimplemented texture format with pixel format={} and component type={}",
+                         static_cast<u32>(pixel_format), static_cast<u32>(component_type));
+    ASSERT_MSG(component_type == tuple.component_type, "Component type mismatch");
+
+    auto usage = vk::FormatFeatureFlagBits::eSampledImage |
+                 vk::FormatFeatureFlagBits::eTransferDst | vk::FormatFeatureFlagBits::eTransferSrc;
+    if (tuple.attachable) {
+        usage |= IsZetaFormat(pixel_format) ? vk::FormatFeatureFlagBits::eDepthStencilAttachment
+                                            : vk::FormatFeatureFlagBits::eColorAttachment;
+    }
+    return {device.GetSupportedFormat(tuple.format, usage, format_type), tuple.attachable};
+}
+
+vk::ShaderStageFlagBits ShaderStage(Maxwell::ShaderStage stage) {
+    switch (stage) {
+    case Maxwell::ShaderStage::Vertex:
+        return vk::ShaderStageFlagBits::eVertex;
+    case Maxwell::ShaderStage::TesselationControl:
+        return vk::ShaderStageFlagBits::eTessellationControl;
+    case Maxwell::ShaderStage::TesselationEval:
+        return vk::ShaderStageFlagBits::eTessellationEvaluation;
+    case Maxwell::ShaderStage::Geometry:
+        return vk::ShaderStageFlagBits::eGeometry;
+    case Maxwell::ShaderStage::Fragment:
+        return vk::ShaderStageFlagBits::eFragment;
+    }
+    UNIMPLEMENTED_MSG("Unimplemented shader stage={}", static_cast<u32>(stage));
+    return {};
+}
+
+vk::PrimitiveTopology PrimitiveTopology(Maxwell::PrimitiveTopology topology) {
+    switch (topology) {
+    case Maxwell::PrimitiveTopology::Points:
+        return vk::PrimitiveTopology::ePointList;
+    case Maxwell::PrimitiveTopology::Lines:
+        return vk::PrimitiveTopology::eLineList;
+    case Maxwell::PrimitiveTopology::LineStrip:
+        return vk::PrimitiveTopology::eLineStrip;
+    case Maxwell::PrimitiveTopology::Triangles:
+        return vk::PrimitiveTopology::eTriangleList;
+    case Maxwell::PrimitiveTopology::TriangleStrip:
+        return vk::PrimitiveTopology::eTriangleStrip;
+    }
+    UNIMPLEMENTED_MSG("Unimplemented topology={}", static_cast<u32>(topology));
+    return {};
+}
+
+vk::Format VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttribute::Size size) {
+    switch (type) {
+    case Maxwell::VertexAttribute::Type::SignedNorm:
+        break;
+    case Maxwell::VertexAttribute::Type::UnsignedNorm:
+        switch (size) {
+        case Maxwell::VertexAttribute::Size::Size_8_8_8_8:
+            return vk::Format::eR8G8B8A8Unorm;
+        default:
+            break;
+        }
+        break;
+    case Maxwell::VertexAttribute::Type::SignedInt:
+        break;
+    case Maxwell::VertexAttribute::Type::UnsignedInt:
+        switch (size) {
+        case Maxwell::VertexAttribute::Size::Size_32:
+            return vk::Format::eR32Uint;
+        default:
+            break;
+        }
+    case Maxwell::VertexAttribute::Type::UnsignedScaled:
+    case Maxwell::VertexAttribute::Type::SignedScaled:
+        break;
+    case Maxwell::VertexAttribute::Type::Float:
+        switch (size) {
+        case Maxwell::VertexAttribute::Size::Size_32_32_32_32:
+            return vk::Format::eR32G32B32A32Sfloat;
+        case Maxwell::VertexAttribute::Size::Size_32_32_32:
+            return vk::Format::eR32G32B32Sfloat;
+        case Maxwell::VertexAttribute::Size::Size_32_32:
+            return vk::Format::eR32G32Sfloat;
+        case Maxwell::VertexAttribute::Size::Size_32:
+            return vk::Format::eR32Sfloat;
+        default:
+            break;
+        }
+        break;
+    }
+    UNIMPLEMENTED_MSG("Unimplemented vertex format of type={} and size={}", static_cast<u32>(type),
+                      static_cast<u32>(size));
+    return {};
+}
+
+vk::CompareOp ComparisonOp(Maxwell::ComparisonOp comparison) {
+    switch (comparison) {
+    case Maxwell::ComparisonOp::Never:
+    case Maxwell::ComparisonOp::NeverOld:
+        return vk::CompareOp::eNever;
+    case Maxwell::ComparisonOp::Less:
+    case Maxwell::ComparisonOp::LessOld:
+        return vk::CompareOp::eLess;
+    case Maxwell::ComparisonOp::Equal:
+    case Maxwell::ComparisonOp::EqualOld:
+        return vk::CompareOp::eEqual;
+    case Maxwell::ComparisonOp::LessEqual:
+    case Maxwell::ComparisonOp::LessEqualOld:
+        return vk::CompareOp::eLessOrEqual;
+    case Maxwell::ComparisonOp::Greater:
+    case Maxwell::ComparisonOp::GreaterOld:
+        return vk::CompareOp::eGreater;
+    case Maxwell::ComparisonOp::NotEqual:
+    case Maxwell::ComparisonOp::NotEqualOld:
+        return vk::CompareOp::eNotEqual;
+    case Maxwell::ComparisonOp::GreaterEqual:
+    case Maxwell::ComparisonOp::GreaterEqualOld:
+        return vk::CompareOp::eGreaterOrEqual;
+    case Maxwell::ComparisonOp::Always:
+    case Maxwell::ComparisonOp::AlwaysOld:
+        return vk::CompareOp::eAlways;
+    }
+    UNIMPLEMENTED_MSG("Unimplemented comparison op={}", static_cast<u32>(comparison));
+    return {};
+}
+
+vk::IndexType IndexFormat(Maxwell::IndexFormat index_format) {
+    switch (index_format) {
+    case Maxwell::IndexFormat::UnsignedByte:
+        UNIMPLEMENTED_MSG("Vulkan does not support native u8 index format");
+        return vk::IndexType::eUint16;
+    case Maxwell::IndexFormat::UnsignedShort:
+        return vk::IndexType::eUint16;
+    case Maxwell::IndexFormat::UnsignedInt:
+        return vk::IndexType::eUint32;
+    }
+    UNIMPLEMENTED_MSG("Unimplemented index_format={}", static_cast<u32>(index_format));
+    return {};
+}
+
+vk::StencilOp StencilOp(Maxwell::StencilOp stencil_op) {
+    switch (stencil_op) {
+    case Maxwell::StencilOp::Keep:
+    case Maxwell::StencilOp::KeepOGL:
+        return vk::StencilOp::eKeep;
+    case Maxwell::StencilOp::Zero:
+    case Maxwell::StencilOp::ZeroOGL:
+        return vk::StencilOp::eZero;
+    case Maxwell::StencilOp::Replace:
+    case Maxwell::StencilOp::ReplaceOGL:
+        return vk::StencilOp::eReplace;
+    case Maxwell::StencilOp::Incr:
+    case Maxwell::StencilOp::IncrOGL:
+        return vk::StencilOp::eIncrementAndClamp;
+    case Maxwell::StencilOp::Decr:
+    case Maxwell::StencilOp::DecrOGL:
+        return vk::StencilOp::eDecrementAndClamp;
+    case Maxwell::StencilOp::Invert:
+    case Maxwell::StencilOp::InvertOGL:
+        return vk::StencilOp::eInvert;
+    case Maxwell::StencilOp::IncrWrap:
+    case Maxwell::StencilOp::IncrWrapOGL:
+        return vk::StencilOp::eIncrementAndWrap;
+    case Maxwell::StencilOp::DecrWrap:
+    case Maxwell::StencilOp::DecrWrapOGL:
+        return vk::StencilOp::eDecrementAndWrap;
+    }
+    UNIMPLEMENTED_MSG("Unimplemented stencil op={}", static_cast<u32>(stencil_op));
+    return {};
+}
+
+vk::BlendOp BlendEquation(Maxwell::Blend::Equation equation) {
+    switch (equation) {
+    case Maxwell::Blend::Equation::Add:
+    case Maxwell::Blend::Equation::AddGL:
+        return vk::BlendOp::eAdd;
+    case Maxwell::Blend::Equation::Subtract:
+    case Maxwell::Blend::Equation::SubtractGL:
+        return vk::BlendOp::eSubtract;
+    case Maxwell::Blend::Equation::ReverseSubtract:
+    case Maxwell::Blend::Equation::ReverseSubtractGL:
+        return vk::BlendOp::eReverseSubtract;
+    case Maxwell::Blend::Equation::Min:
+    case Maxwell::Blend::Equation::MinGL:
+        return vk::BlendOp::eMin;
+    case Maxwell::Blend::Equation::Max:
+    case Maxwell::Blend::Equation::MaxGL:
+        return vk::BlendOp::eMax;
+    }
+    UNIMPLEMENTED_MSG("Unimplemented blend equation={}", static_cast<u32>(equation));
+    return {};
+}
+
+vk::BlendFactor BlendFactor(Maxwell::Blend::Factor factor) {
+    switch (factor) {
+    case Maxwell::Blend::Factor::Zero:
+    case Maxwell::Blend::Factor::ZeroGL:
+        return vk::BlendFactor::eZero;
+    case Maxwell::Blend::Factor::One:
+    case Maxwell::Blend::Factor::OneGL:
+        return vk::BlendFactor::eOne;
+    case Maxwell::Blend::Factor::SourceColor:
+    case Maxwell::Blend::Factor::SourceColorGL:
+        return vk::BlendFactor::eSrcColor;
+    case Maxwell::Blend::Factor::OneMinusSourceColor:
+    case Maxwell::Blend::Factor::OneMinusSourceColorGL:
+        return vk::BlendFactor::eOneMinusSrcColor;
+    case Maxwell::Blend::Factor::SourceAlpha:
+    case Maxwell::Blend::Factor::SourceAlphaGL:
+        return vk::BlendFactor::eSrcAlpha;
+    case Maxwell::Blend::Factor::OneMinusSourceAlpha:
+    case Maxwell::Blend::Factor::OneMinusSourceAlphaGL:
+        return vk::BlendFactor::eOneMinusSrcAlpha;
+    case Maxwell::Blend::Factor::DestAlpha:
+    case Maxwell::Blend::Factor::DestAlphaGL:
+        return vk::BlendFactor::eDstAlpha;
+    case Maxwell::Blend::Factor::OneMinusDestAlpha:
+    case Maxwell::Blend::Factor::OneMinusDestAlphaGL:
+        return vk::BlendFactor::eOneMinusDstAlpha;
+    case Maxwell::Blend::Factor::DestColor:
+    case Maxwell::Blend::Factor::DestColorGL:
+        return vk::BlendFactor::eDstColor;
+    case Maxwell::Blend::Factor::OneMinusDestColor:
+    case Maxwell::Blend::Factor::OneMinusDestColorGL:
+        return vk::BlendFactor::eOneMinusDstColor;
+    case Maxwell::Blend::Factor::SourceAlphaSaturate:
+    case Maxwell::Blend::Factor::SourceAlphaSaturateGL:
+        return vk::BlendFactor::eSrcAlphaSaturate;
+    case Maxwell::Blend::Factor::Source1Color:
+    case Maxwell::Blend::Factor::Source1ColorGL:
+        return vk::BlendFactor::eSrc1Color;
+    case Maxwell::Blend::Factor::OneMinusSource1Color:
+    case Maxwell::Blend::Factor::OneMinusSource1ColorGL:
+        return vk::BlendFactor::eOneMinusSrc1Color;
+    case Maxwell::Blend::Factor::Source1Alpha:
+    case Maxwell::Blend::Factor::Source1AlphaGL:
+        return vk::BlendFactor::eSrc1Alpha;
+    case Maxwell::Blend::Factor::OneMinusSource1Alpha:
+    case Maxwell::Blend::Factor::OneMinusSource1AlphaGL:
+        return vk::BlendFactor::eOneMinusSrc1Alpha;
+    case Maxwell::Blend::Factor::ConstantColor:
+    case Maxwell::Blend::Factor::ConstantColorGL:
+        return vk::BlendFactor::eConstantColor;
+    case Maxwell::Blend::Factor::OneMinusConstantColor:
+    case Maxwell::Blend::Factor::OneMinusConstantColorGL:
+        return vk::BlendFactor::eOneMinusConstantColor;
+    case Maxwell::Blend::Factor::ConstantAlpha:
+    case Maxwell::Blend::Factor::ConstantAlphaGL:
+        return vk::BlendFactor::eConstantAlpha;
+    case Maxwell::Blend::Factor::OneMinusConstantAlpha:
+    case Maxwell::Blend::Factor::OneMinusConstantAlphaGL:
+        return vk::BlendFactor::eOneMinusConstantAlpha;
+    }
+    UNIMPLEMENTED_MSG("Unimplemented blend factor={}", static_cast<u32>(factor));
+    return {};
+}
+
+vk::FrontFace FrontFace(Maxwell::Cull::FrontFace front_face) {
+    switch (front_face) {
+    case Maxwell::Cull::FrontFace::ClockWise:
+        return vk::FrontFace::eClockwise;
+    case Maxwell::Cull::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) {
+    switch (cull_face) {
+    case Maxwell::Cull::CullFace::Front:
+        return vk::CullModeFlagBits::eFront;
+    case Maxwell::Cull::CullFace::Back:
+        return vk::CullModeFlagBits::eBack;
+    case Maxwell::Cull::CullFace::FrontAndBack:
+        return vk::CullModeFlagBits::eFrontAndBack;
+    }
+    UNIMPLEMENTED_MSG("Unimplemented cull face={}", static_cast<u32>(cull_face));
+    return {};
+}
+
+vk::ComponentSwizzle SwizzleSource(Tegra::Texture::SwizzleSource swizzle) {
+    switch (swizzle) {
+    case Tegra::Texture::SwizzleSource::Zero:
+        return vk::ComponentSwizzle::eZero;
+    case Tegra::Texture::SwizzleSource::R:
+        return vk::ComponentSwizzle::eR;
+    case Tegra::Texture::SwizzleSource::G:
+        return vk::ComponentSwizzle::eG;
+    case Tegra::Texture::SwizzleSource::B:
+        return vk::ComponentSwizzle::eB;
+    case Tegra::Texture::SwizzleSource::A:
+        return vk::ComponentSwizzle::eA;
+    case Tegra::Texture::SwizzleSource::OneInt:
+    case Tegra::Texture::SwizzleSource::OneFloat:
+        return vk::ComponentSwizzle::eOne;
+    }
+    UNIMPLEMENTED_MSG("Unimplemented swizzle source={}", static_cast<u32>(swizzle));
+    return {};
+}
+
+} // namespace Vulkan::MaxwellToVK

src/video_core/renderer_vulkan/maxwell_to_vk.h

@@ -0,0 +1,58 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <utility>
+#include "common/common_types.h"
+#include "video_core/engines/maxwell_3d.h"
+#include "video_core/renderer_vulkan/declarations.h"
+#include "video_core/renderer_vulkan/vk_device.h"
+#include "video_core/surface.h"
+#include "video_core/textures/texture.h"
+
+namespace Vulkan::MaxwellToVK {
+
+using Maxwell = Tegra::Engines::Maxwell3D::Regs;
+using PixelFormat = VideoCore::Surface::PixelFormat;
+using ComponentType = VideoCore::Surface::ComponentType;
+
+namespace Sampler {
+
+vk::Filter Filter(Tegra::Texture::TextureFilter filter);
+
+vk::SamplerMipmapMode MipmapMode(Tegra::Texture::TextureMipmapFilter mipmap_filter);
+
+vk::SamplerAddressMode WrapMode(Tegra::Texture::WrapMode wrap_mode);
+
+vk::CompareOp DepthCompareFunction(Tegra::Texture::DepthCompareFunc depth_compare_func);
+
+} // namespace Sampler
+
+std::pair<vk::Format, bool> SurfaceFormat(const VKDevice& device, FormatType format_type,
+                                          PixelFormat pixel_format, ComponentType component_type);
+
+vk::ShaderStageFlagBits ShaderStage(Maxwell::ShaderStage stage);
+
+vk::PrimitiveTopology PrimitiveTopology(Maxwell::PrimitiveTopology topology);
+
+vk::Format VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttribute::Size size);
+
+vk::CompareOp ComparisonOp(Maxwell::ComparisonOp comparison);
+
+vk::IndexType IndexFormat(Maxwell::IndexFormat index_format);
+
+vk::StencilOp StencilOp(Maxwell::StencilOp stencil_op);
+
+vk::BlendOp BlendEquation(Maxwell::Blend::Equation equation);
+
+vk::BlendFactor BlendFactor(Maxwell::Blend::Factor factor);
+
+vk::FrontFace FrontFace(Maxwell::Cull::FrontFace front_face);
+
+vk::CullModeFlags CullFace(Maxwell::Cull::CullFace cull_face);
+
+vk::ComponentSwizzle SwizzleSource(Tegra::Texture::SwizzleSource swizzle);
+
+} // namespace Vulkan::MaxwellToVK

src/video_core/renderer_vulkan/vk_buffer_cache.cpp

@@ -8,7 +8,7 @@
 #include <tuple>
 
 #include "common/alignment.h"
-#include "core/core.h"
+#include "common/assert.h"
 #include "core/memory.h"
 #include "video_core/renderer_vulkan/declarations.h"
 #include "video_core/renderer_vulkan/vk_buffer_cache.h"

src/video_core/renderer_vulkan/vk_device.cpp

@@ -122,8 +122,7 @@ bool VKDevice::IsFormatSupported(vk::Format wanted_format, vk::FormatFeatureFlag
                                  FormatType format_type) const {
     const auto it = format_properties.find(wanted_format);
     if (it == format_properties.end()) {
-        LOG_CRITICAL(Render_Vulkan, "Unimplemented format query={}",
-                     static_cast<u32>(wanted_format));
+        LOG_CRITICAL(Render_Vulkan, "Unimplemented format query={}", vk::to_string(wanted_format));
         UNREACHABLE();
         return true;
     }
@@ -219,11 +218,19 @@ std::map<vk::Format, vk::FormatProperties> VKDevice::GetFormatProperties(
         format_properties.emplace(format, physical.getFormatProperties(format, dldi));
     };
     AddFormatQuery(vk::Format::eA8B8G8R8UnormPack32);
-    AddFormatQuery(vk::Format::eR5G6B5UnormPack16);
+    AddFormatQuery(vk::Format::eB5G6R5UnormPack16);
+    AddFormatQuery(vk::Format::eA2B10G10R10UnormPack32);
+    AddFormatQuery(vk::Format::eR8G8B8A8Srgb);
+    AddFormatQuery(vk::Format::eR8Unorm);
     AddFormatQuery(vk::Format::eD32Sfloat);
+    AddFormatQuery(vk::Format::eD16Unorm);
     AddFormatQuery(vk::Format::eD16UnormS8Uint);
     AddFormatQuery(vk::Format::eD24UnormS8Uint);
     AddFormatQuery(vk::Format::eD32SfloatS8Uint);
+    AddFormatQuery(vk::Format::eBc1RgbaUnormBlock);
+    AddFormatQuery(vk::Format::eBc2UnormBlock);
+    AddFormatQuery(vk::Format::eBc3UnormBlock);
+    AddFormatQuery(vk::Format::eBc4UnormBlock);
 
     return format_properties;
 }

src/video_core/shader/decode.cpp

@@ -165,6 +165,7 @@ u32 ShaderIR::DecodeInstr(NodeBlock& bb, u32 pc) {
         {OpCode::Type::Hfma2, &ShaderIR::DecodeHfma2},
         {OpCode::Type::Conversion, &ShaderIR::DecodeConversion},
         {OpCode::Type::Memory, &ShaderIR::DecodeMemory},
+        {OpCode::Type::Texture, &ShaderIR::DecodeTexture},
         {OpCode::Type::FloatSetPredicate, &ShaderIR::DecodeFloatSetPredicate},
         {OpCode::Type::IntegerSetPredicate, &ShaderIR::DecodeIntegerSetPredicate},
         {OpCode::Type::HalfSetPredicate, &ShaderIR::DecodeHalfSetPredicate},

src/video_core/shader/decode/memory.cpp

@@ -17,24 +17,6 @@ using Tegra::Shader::Attribute;
 using Tegra::Shader::Instruction;
 using Tegra::Shader::OpCode;
 using Tegra::Shader::Register;
-using Tegra::Shader::TextureMiscMode;
-using Tegra::Shader::TextureProcessMode;
-using Tegra::Shader::TextureType;
-
-static std::size_t GetCoordCount(TextureType texture_type) {
-    switch (texture_type) {
-    case TextureType::Texture1D:
-        return 1;
-    case TextureType::Texture2D:
-        return 2;
-    case TextureType::Texture3D:
-    case TextureType::TextureCube:
-        return 3;
-    default:
-        UNIMPLEMENTED_MSG("Unhandled texture type: {}", static_cast<u32>(texture_type));
-        return 0;
-    }
-}
 
 u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
     const Instruction instr = {program_code[pc]};
@@ -247,194 +229,6 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
         }
         break;
     }
-    case OpCode::Id::TEX: {
-        UNIMPLEMENTED_IF_MSG(instr.tex.UsesMiscMode(TextureMiscMode::AOFFI),
-                             "AOFFI is not implemented");
-
-        if (instr.tex.UsesMiscMode(TextureMiscMode::NODEP)) {
-            LOG_WARNING(HW_GPU, "TEX.NODEP implementation is incomplete");
-        }
-
-        const TextureType texture_type{instr.tex.texture_type};
-        const bool is_array = instr.tex.array != 0;
-        const bool depth_compare = instr.tex.UsesMiscMode(TextureMiscMode::DC);
-        const auto process_mode = instr.tex.GetTextureProcessMode();
-        WriteTexInstructionFloat(
-            bb, instr, GetTexCode(instr, texture_type, process_mode, depth_compare, is_array));
-        break;
-    }
-    case OpCode::Id::TEXS: {
-        const TextureType texture_type{instr.texs.GetTextureType()};
-        const bool is_array{instr.texs.IsArrayTexture()};
-        const bool depth_compare = instr.texs.UsesMiscMode(TextureMiscMode::DC);
-        const auto process_mode = instr.texs.GetTextureProcessMode();
-
-        if (instr.texs.UsesMiscMode(TextureMiscMode::NODEP)) {
-            LOG_WARNING(HW_GPU, "TEXS.NODEP implementation is incomplete");
-        }
-
-        const Node4 components =
-            GetTexsCode(instr, texture_type, process_mode, depth_compare, is_array);
-
-        if (instr.texs.fp32_flag) {
-            WriteTexsInstructionFloat(bb, instr, components);
-        } else {
-            WriteTexsInstructionHalfFloat(bb, instr, components);
-        }
-        break;
-    }
-    case OpCode::Id::TLD4: {
-        ASSERT(instr.tld4.array == 0);
-        UNIMPLEMENTED_IF_MSG(instr.tld4.UsesMiscMode(TextureMiscMode::AOFFI),
-                             "AOFFI is not implemented");
-        UNIMPLEMENTED_IF_MSG(instr.tld4.UsesMiscMode(TextureMiscMode::NDV),
-                             "NDV is not implemented");
-        UNIMPLEMENTED_IF_MSG(instr.tld4.UsesMiscMode(TextureMiscMode::PTP),
-                             "PTP is not implemented");
-
-        if (instr.tld4.UsesMiscMode(TextureMiscMode::NODEP)) {
-            LOG_WARNING(HW_GPU, "TLD4.NODEP implementation is incomplete");
-        }
-
-        const auto texture_type = instr.tld4.texture_type.Value();
-        const bool depth_compare = instr.tld4.UsesMiscMode(TextureMiscMode::DC);
-        const bool is_array = instr.tld4.array != 0;
-        WriteTexInstructionFloat(bb, instr,
-                                 GetTld4Code(instr, texture_type, depth_compare, is_array));
-        break;
-    }
-    case OpCode::Id::TLD4S: {
-        UNIMPLEMENTED_IF_MSG(instr.tld4s.UsesMiscMode(TextureMiscMode::AOFFI),
-                             "AOFFI is not implemented");
-        if (instr.tld4s.UsesMiscMode(TextureMiscMode::NODEP)) {
-            LOG_WARNING(HW_GPU, "TLD4S.NODEP implementation is incomplete");
-        }
-
-        const bool depth_compare = instr.tld4s.UsesMiscMode(TextureMiscMode::DC);
-        const Node op_a = GetRegister(instr.gpr8);
-        const Node op_b = GetRegister(instr.gpr20);
-
-        // TODO(Subv): Figure out how the sampler type is encoded in the TLD4S instruction.
-        std::vector<Node> coords;
-        if (depth_compare) {
-            // Note: TLD4S coordinate encoding works just like TEXS's
-            const Node op_y = GetRegister(instr.gpr8.Value() + 1);
-            coords.push_back(op_a);
-            coords.push_back(op_y);
-            coords.push_back(op_b);
-        } else {
-            coords.push_back(op_a);
-            coords.push_back(op_b);
-        }
-        std::vector<Node> extras;
-        extras.push_back(Immediate(static_cast<u32>(instr.tld4s.component)));
-
-        const auto& sampler =
-            GetSampler(instr.sampler, TextureType::Texture2D, false, depth_compare);
-
-        Node4 values;
-        for (u32 element = 0; element < values.size(); ++element) {
-            auto coords_copy = coords;
-            MetaTexture meta{sampler, {}, {}, extras, element};
-            values[element] = Operation(OperationCode::TextureGather, meta, std::move(coords_copy));
-        }
-
-        WriteTexsInstructionFloat(bb, instr, values);
-        break;
-    }
-    case OpCode::Id::TXQ: {
-        if (instr.txq.UsesMiscMode(TextureMiscMode::NODEP)) {
-            LOG_WARNING(HW_GPU, "TXQ.NODEP implementation is incomplete");
-        }
-
-        // TODO: The new commits on the texture refactor, change the way samplers work.
-        // Sadly, not all texture instructions specify the type of texture their sampler
-        // uses. This must be fixed at a later instance.
-        const auto& sampler =
-            GetSampler(instr.sampler, Tegra::Shader::TextureType::Texture2D, false, false);
-
-        u32 indexer = 0;
-        switch (instr.txq.query_type) {
-        case Tegra::Shader::TextureQueryType::Dimension: {
-            for (u32 element = 0; element < 4; ++element) {
-                if (!instr.txq.IsComponentEnabled(element)) {
-                    continue;
-                }
-                MetaTexture meta{sampler, {}, {}, {}, element};
-                const Node value =
-                    Operation(OperationCode::TextureQueryDimensions, meta, GetRegister(instr.gpr8));
-                SetTemporal(bb, indexer++, value);
-            }
-            for (u32 i = 0; i < indexer; ++i) {
-                SetRegister(bb, instr.gpr0.Value() + i, GetTemporal(i));
-            }
-            break;
-        }
-        default:
-            UNIMPLEMENTED_MSG("Unhandled texture query type: {}",
-                              static_cast<u32>(instr.txq.query_type.Value()));
-        }
-        break;
-    }
-    case OpCode::Id::TMML: {
-        UNIMPLEMENTED_IF_MSG(instr.tmml.UsesMiscMode(Tegra::Shader::TextureMiscMode::NDV),
-                             "NDV is not implemented");
-
-        if (instr.tmml.UsesMiscMode(TextureMiscMode::NODEP)) {
-            LOG_WARNING(HW_GPU, "TMML.NODEP implementation is incomplete");
-        }
-
-        auto texture_type = instr.tmml.texture_type.Value();
-        const bool is_array = instr.tmml.array != 0;
-        const auto& sampler = GetSampler(instr.sampler, texture_type, is_array, false);
-
-        std::vector<Node> coords;
-
-        // TODO: Add coordinates for different samplers once other texture types are implemented.
-        switch (texture_type) {
-        case TextureType::Texture1D:
-            coords.push_back(GetRegister(instr.gpr8));
-            break;
-        case TextureType::Texture2D:
-            coords.push_back(GetRegister(instr.gpr8.Value() + 0));
-            coords.push_back(GetRegister(instr.gpr8.Value() + 1));
-            break;
-        default:
-            UNIMPLEMENTED_MSG("Unhandled texture type {}", static_cast<u32>(texture_type));
-
-            // Fallback to interpreting as a 2D texture for now
-            coords.push_back(GetRegister(instr.gpr8.Value() + 0));
-            coords.push_back(GetRegister(instr.gpr8.Value() + 1));
-            texture_type = TextureType::Texture2D;
-        }
-
-        for (u32 element = 0; element < 2; ++element) {
-            auto params = coords;
-            MetaTexture meta{sampler, {}, {}, {}, element};
-            const Node value = Operation(OperationCode::TextureQueryLod, meta, std::move(params));
-            SetTemporal(bb, element, value);
-        }
-        for (u32 element = 0; element < 2; ++element) {
-            SetRegister(bb, instr.gpr0.Value() + element, GetTemporal(element));
-        }
-
-        break;
-    }
-    case OpCode::Id::TLDS: {
-        const Tegra::Shader::TextureType texture_type{instr.tlds.GetTextureType()};
-        const bool is_array{instr.tlds.IsArrayTexture()};
-
-        UNIMPLEMENTED_IF_MSG(instr.tlds.UsesMiscMode(TextureMiscMode::AOFFI),
-                             "AOFFI is not implemented");
-        UNIMPLEMENTED_IF_MSG(instr.tlds.UsesMiscMode(TextureMiscMode::MZ), "MZ is not implemented");
-
-        if (instr.tlds.UsesMiscMode(TextureMiscMode::NODEP)) {
-            LOG_WARNING(HW_GPU, "TLDS.NODEP implementation is incomplete");
-        }
-
-        WriteTexsInstructionFloat(bb, instr, GetTldsCode(instr, texture_type, is_array));
-        break;
-    }
     default:
         UNIMPLEMENTED_MSG("Unhandled memory instruction: {}", opcode->get().GetName());
     }
@@ -442,291 +236,4 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
     return pc;
 }
 
-const Sampler& ShaderIR::GetSampler(const Tegra::Shader::Sampler& sampler, TextureType type,
-                                    bool is_array, bool is_shadow) {
-    const auto offset = static_cast<std::size_t>(sampler.index.Value());
-
-    // If this sampler has already been used, return the existing mapping.
-    const auto itr =
-        std::find_if(used_samplers.begin(), used_samplers.end(),
-                     [&](const Sampler& entry) { return entry.GetOffset() == offset; });
-    if (itr != used_samplers.end()) {
-        ASSERT(itr->GetType() == type && itr->IsArray() == is_array &&
-               itr->IsShadow() == is_shadow);
-        return *itr;
-    }
-
-    // Otherwise create a new mapping for this sampler
-    const std::size_t next_index = used_samplers.size();
-    const Sampler entry{offset, next_index, type, is_array, is_shadow};
-    return *used_samplers.emplace(entry).first;
-}
-
-void ShaderIR::WriteTexInstructionFloat(NodeBlock& bb, Instruction instr, const Node4& components) {
-    u32 dest_elem = 0;
-    for (u32 elem = 0; elem < 4; ++elem) {
-        if (!instr.tex.IsComponentEnabled(elem)) {
-            // Skip disabled components
-            continue;
-        }
-        SetTemporal(bb, dest_elem++, components[elem]);
-    }
-    // After writing values in temporals, move them to the real registers
-    for (u32 i = 0; i < dest_elem; ++i) {
-        SetRegister(bb, instr.gpr0.Value() + i, GetTemporal(i));
-    }
-}
-
-void ShaderIR::WriteTexsInstructionFloat(NodeBlock& bb, Instruction instr,
-                                         const Node4& components) {
-    // TEXS has two destination registers and a swizzle. The first two elements in the swizzle
-    // go into gpr0+0 and gpr0+1, and the rest goes into gpr28+0 and gpr28+1
-
-    u32 dest_elem = 0;
-    for (u32 component = 0; component < 4; ++component) {
-        if (!instr.texs.IsComponentEnabled(component))
-            continue;
-        SetTemporal(bb, dest_elem++, components[component]);
-    }
-
-    for (u32 i = 0; i < dest_elem; ++i) {
-        if (i < 2) {
-            // Write the first two swizzle components to gpr0 and gpr0+1
-            SetRegister(bb, instr.gpr0.Value() + i % 2, GetTemporal(i));
-        } else {
-            ASSERT(instr.texs.HasTwoDestinations());
-            // Write the rest of the swizzle components to gpr28 and gpr28+1
-            SetRegister(bb, instr.gpr28.Value() + i % 2, GetTemporal(i));
-        }
-    }
-}
-
-void ShaderIR::WriteTexsInstructionHalfFloat(NodeBlock& bb, Instruction instr,
-                                             const Node4& components) {
-    // TEXS.F16 destionation registers are packed in two registers in pairs (just like any half
-    // float instruction).
-
-    Node4 values;
-    u32 dest_elem = 0;
-    for (u32 component = 0; component < 4; ++component) {
-        if (!instr.texs.IsComponentEnabled(component))
-            continue;
-        values[dest_elem++] = components[component];
-    }
-    if (dest_elem == 0)
-        return;
-
-    std::generate(values.begin() + dest_elem, values.end(), [&]() { return Immediate(0); });
-
-    const Node first_value = Operation(OperationCode::HPack2, values[0], values[1]);
-    if (dest_elem <= 2) {
-        SetRegister(bb, instr.gpr0, first_value);
-        return;
-    }
-
-    SetTemporal(bb, 0, first_value);
-    SetTemporal(bb, 1, Operation(OperationCode::HPack2, values[2], values[3]));
-
-    SetRegister(bb, instr.gpr0, GetTemporal(0));
-    SetRegister(bb, instr.gpr28, GetTemporal(1));
-}
-
-Node4 ShaderIR::GetTextureCode(Instruction instr, TextureType texture_type,
-                               TextureProcessMode process_mode, std::vector<Node> coords,
-                               Node array, Node depth_compare, u32 bias_offset) {
-    const bool is_array = array;
-    const bool is_shadow = depth_compare;
-
-    UNIMPLEMENTED_IF_MSG((texture_type == TextureType::Texture3D && (is_array || is_shadow)) ||
-                             (texture_type == TextureType::TextureCube && is_array && is_shadow),
-                         "This method is not supported.");
-
-    const auto& sampler = GetSampler(instr.sampler, texture_type, is_array, is_shadow);
-
-    const bool lod_needed = process_mode == TextureProcessMode::LZ ||
-                            process_mode == TextureProcessMode::LL ||
-                            process_mode == TextureProcessMode::LLA;
-
-    // LOD selection (either via bias or explicit textureLod) not supported in GL for
-    // sampler2DArrayShadow and samplerCubeArrayShadow.
-    const bool gl_lod_supported =
-        !((texture_type == Tegra::Shader::TextureType::Texture2D && is_array && is_shadow) ||
-          (texture_type == Tegra::Shader::TextureType::TextureCube && is_array && is_shadow));
-
-    const OperationCode read_method =
-        lod_needed && gl_lod_supported ? OperationCode::TextureLod : OperationCode::Texture;
-
-    UNIMPLEMENTED_IF(process_mode != TextureProcessMode::None && !gl_lod_supported);
-
-    std::vector<Node> extras;
-    if (process_mode != TextureProcessMode::None && gl_lod_supported) {
-        if (process_mode == TextureProcessMode::LZ) {
-            extras.push_back(Immediate(0.0f));
-        } else {
-            // If present, lod or bias are always stored in the register indexed by the gpr20
-            // field with an offset depending on the usage of the other registers
-            extras.push_back(GetRegister(instr.gpr20.Value() + bias_offset));
-        }
-    }
-
-    Node4 values;
-    for (u32 element = 0; element < values.size(); ++element) {
-        auto copy_coords = coords;
-        MetaTexture meta{sampler, array, depth_compare, extras, element};
-        values[element] = Operation(read_method, meta, std::move(copy_coords));
-    }
-
-    return values;
-}
-
-Node4 ShaderIR::GetTexCode(Instruction instr, TextureType texture_type,
-                           TextureProcessMode process_mode, bool depth_compare, bool is_array) {
-    const bool lod_bias_enabled =
-        (process_mode != TextureProcessMode::None && process_mode != TextureProcessMode::LZ);
-
-    const auto [coord_count, total_coord_count] = ValidateAndGetCoordinateElement(
-        texture_type, depth_compare, is_array, lod_bias_enabled, 4, 5);
-    // If enabled arrays index is always stored in the gpr8 field
-    const u64 array_register = instr.gpr8.Value();
-    // First coordinate index is the gpr8 or gpr8 + 1 when arrays are used
-    const u64 coord_register = array_register + (is_array ? 1 : 0);
-
-    std::vector<Node> coords;
-    for (std::size_t i = 0; i < coord_count; ++i) {
-        coords.push_back(GetRegister(coord_register + i));
-    }
-    // 1D.DC in OpenGL the 2nd component is ignored.
-    if (depth_compare && !is_array && texture_type == TextureType::Texture1D) {
-        coords.push_back(Immediate(0.0f));
-    }
-
-    const Node array = is_array ? GetRegister(array_register) : nullptr;
-
-    Node dc{};
-    if (depth_compare) {
-        // Depth is always stored in the register signaled by gpr20 or in the next register if lod
-        // or bias are used
-        const u64 depth_register = instr.gpr20.Value() + (lod_bias_enabled ? 1 : 0);
-        dc = GetRegister(depth_register);
-    }
-
-    return GetTextureCode(instr, texture_type, process_mode, coords, array, dc, 0);
-}
-
-Node4 ShaderIR::GetTexsCode(Instruction instr, TextureType texture_type,
-                            TextureProcessMode process_mode, bool depth_compare, bool is_array) {
-    const bool lod_bias_enabled =
-        (process_mode != TextureProcessMode::None && process_mode != TextureProcessMode::LZ);
-
-    const auto [coord_count, total_coord_count] = ValidateAndGetCoordinateElement(
-        texture_type, depth_compare, is_array, lod_bias_enabled, 4, 4);
-    // If enabled arrays index is always stored in the gpr8 field
-    const u64 array_register = instr.gpr8.Value();
-    // First coordinate index is stored in gpr8 field or (gpr8 + 1) when arrays are used
-    const u64 coord_register = array_register + (is_array ? 1 : 0);
-    const u64 last_coord_register =
-        (is_array || !(lod_bias_enabled || depth_compare) || (coord_count > 2))
-            ? static_cast<u64>(instr.gpr20.Value())
-            : coord_register + 1;
-    const u32 bias_offset = coord_count > 2 ? 1 : 0;
-
-    std::vector<Node> coords;
-    for (std::size_t i = 0; i < coord_count; ++i) {
-        const bool last = (i == (coord_count - 1)) && (coord_count > 1);
-        coords.push_back(GetRegister(last ? last_coord_register : coord_register + i));
-    }
-
-    const Node array = is_array ? GetRegister(array_register) : nullptr;
-
-    Node dc{};
-    if (depth_compare) {
-        // Depth is always stored in the register signaled by gpr20 or in the next register if lod
-        // or bias are used
-        const u64 depth_register = instr.gpr20.Value() + (lod_bias_enabled ? 1 : 0);
-        dc = GetRegister(depth_register);
-    }
-
-    return GetTextureCode(instr, texture_type, process_mode, coords, array, dc, bias_offset);
-}
-
-Node4 ShaderIR::GetTld4Code(Instruction instr, TextureType texture_type, bool depth_compare,
-                            bool is_array) {
-    const std::size_t coord_count = GetCoordCount(texture_type);
-    const std::size_t total_coord_count = coord_count + (is_array ? 1 : 0);
-    const std::size_t total_reg_count = total_coord_count + (depth_compare ? 1 : 0);
-
-    // If enabled arrays index is always stored in the gpr8 field
-    const u64 array_register = instr.gpr8.Value();
-    // First coordinate index is the gpr8 or gpr8 + 1 when arrays are used
-    const u64 coord_register = array_register + (is_array ? 1 : 0);
-
-    std::vector<Node> coords;
-    for (size_t i = 0; i < coord_count; ++i)
-        coords.push_back(GetRegister(coord_register + i));
-
-    const auto& sampler = GetSampler(instr.sampler, texture_type, is_array, depth_compare);
-
-    Node4 values;
-    for (u32 element = 0; element < values.size(); ++element) {
-        auto coords_copy = coords;
-        MetaTexture meta{sampler, GetRegister(array_register), {}, {}, element};
-        values[element] = Operation(OperationCode::TextureGather, meta, std::move(coords_copy));
-    }
-
-    return values;
-}
-
-Node4 ShaderIR::GetTldsCode(Instruction instr, TextureType texture_type, bool is_array) {
-    const std::size_t type_coord_count = GetCoordCount(texture_type);
-    const bool lod_enabled = instr.tlds.GetTextureProcessMode() == TextureProcessMode::LL;
-
-    // If enabled arrays index is always stored in the gpr8 field
-    const u64 array_register = instr.gpr8.Value();
-    // if is array gpr20 is used
-    const u64 coord_register = is_array ? instr.gpr20.Value() : instr.gpr8.Value();
-
-    const u64 last_coord_register =
-        ((type_coord_count > 2) || (type_coord_count == 2 && !lod_enabled)) && !is_array
-            ? static_cast<u64>(instr.gpr20.Value())
-            : coord_register + 1;
-
-    std::vector<Node> coords;
-    for (std::size_t i = 0; i < type_coord_count; ++i) {
-        const bool last = (i == (type_coord_count - 1)) && (type_coord_count > 1);
-        coords.push_back(GetRegister(last ? last_coord_register : coord_register + i));
-    }
-
-    const Node array = is_array ? GetRegister(array_register) : nullptr;
-    // When lod is used always is in gpr20
-    const Node lod = lod_enabled ? GetRegister(instr.gpr20) : Immediate(0);
-
-    const auto& sampler = GetSampler(instr.sampler, texture_type, is_array, false);
-
-    Node4 values;
-    for (u32 element = 0; element < values.size(); ++element) {
-        auto coords_copy = coords;
-        MetaTexture meta{sampler, array, {}, {lod}, element};
-        values[element] = Operation(OperationCode::TexelFetch, meta, std::move(coords_copy));
-    }
-    return values;
-}
-
-std::tuple<std::size_t, std::size_t> ShaderIR::ValidateAndGetCoordinateElement(
-    TextureType texture_type, bool depth_compare, bool is_array, bool lod_bias_enabled,
-    std::size_t max_coords, std::size_t max_inputs) {
-    const std::size_t coord_count = GetCoordCount(texture_type);
-
-    std::size_t total_coord_count = coord_count + (is_array ? 1 : 0) + (depth_compare ? 1 : 0);
-    const std::size_t total_reg_count = total_coord_count + (lod_bias_enabled ? 1 : 0);
-    if (total_coord_count > max_coords || total_reg_count > max_inputs) {
-        UNIMPLEMENTED_MSG("Unsupported Texture operation");
-        total_coord_count = std::min(total_coord_count, max_coords);
-    }
-    // 1D.DC OpenGL is using a vec3 but 2nd component is ignored later.
-    total_coord_count +=
-        (depth_compare && !is_array && texture_type == TextureType::Texture1D) ? 1 : 0;
-
-    return {coord_count, total_coord_count};
-}
-
 } // namespace VideoCommon::Shader

src/video_core/shader/decode/texture.cpp

@@ -0,0 +1,534 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include <vector>
+#include <fmt/format.h>
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "video_core/engines/shader_bytecode.h"
+#include "video_core/shader/shader_ir.h"
+
+namespace VideoCommon::Shader {
+
+using Tegra::Shader::Instruction;
+using Tegra::Shader::OpCode;
+using Tegra::Shader::Register;
+using Tegra::Shader::TextureMiscMode;
+using Tegra::Shader::TextureProcessMode;
+using Tegra::Shader::TextureType;
+
+static std::size_t GetCoordCount(TextureType texture_type) {
+    switch (texture_type) {
+    case TextureType::Texture1D:
+        return 1;
+    case TextureType::Texture2D:
+        return 2;
+    case TextureType::Texture3D:
+    case TextureType::TextureCube:
+        return 3;
+    default:
+        UNIMPLEMENTED_MSG("Unhandled texture type: {}", static_cast<u32>(texture_type));
+        return 0;
+    }
+}
+
+u32 ShaderIR::DecodeTexture(NodeBlock& bb, u32 pc) {
+    const Instruction instr = {program_code[pc]};
+    const auto opcode = OpCode::Decode(instr);
+
+    switch (opcode->get().GetId()) {
+    case OpCode::Id::TEX: {
+        UNIMPLEMENTED_IF_MSG(instr.tex.UsesMiscMode(TextureMiscMode::AOFFI),
+                             "AOFFI is not implemented");
+
+        if (instr.tex.UsesMiscMode(TextureMiscMode::NODEP)) {
+            LOG_WARNING(HW_GPU, "TEX.NODEP implementation is incomplete");
+        }
+
+        const TextureType texture_type{instr.tex.texture_type};
+        const bool is_array = instr.tex.array != 0;
+        const bool depth_compare = instr.tex.UsesMiscMode(TextureMiscMode::DC);
+        const auto process_mode = instr.tex.GetTextureProcessMode();
+        WriteTexInstructionFloat(
+            bb, instr, GetTexCode(instr, texture_type, process_mode, depth_compare, is_array));
+        break;
+    }
+    case OpCode::Id::TEXS: {
+        const TextureType texture_type{instr.texs.GetTextureType()};
+        const bool is_array{instr.texs.IsArrayTexture()};
+        const bool depth_compare = instr.texs.UsesMiscMode(TextureMiscMode::DC);
+        const auto process_mode = instr.texs.GetTextureProcessMode();
+
+        if (instr.texs.UsesMiscMode(TextureMiscMode::NODEP)) {
+            LOG_WARNING(HW_GPU, "TEXS.NODEP implementation is incomplete");
+        }
+
+        const Node4 components =
+            GetTexsCode(instr, texture_type, process_mode, depth_compare, is_array);
+
+        if (instr.texs.fp32_flag) {
+            WriteTexsInstructionFloat(bb, instr, components);
+        } else {
+            WriteTexsInstructionHalfFloat(bb, instr, components);
+        }
+        break;
+    }
+    case OpCode::Id::TLD4: {
+        ASSERT(instr.tld4.array == 0);
+        UNIMPLEMENTED_IF_MSG(instr.tld4.UsesMiscMode(TextureMiscMode::AOFFI),
+                             "AOFFI is not implemented");
+        UNIMPLEMENTED_IF_MSG(instr.tld4.UsesMiscMode(TextureMiscMode::NDV),
+                             "NDV is not implemented");
+        UNIMPLEMENTED_IF_MSG(instr.tld4.UsesMiscMode(TextureMiscMode::PTP),
+                             "PTP is not implemented");
+
+        if (instr.tld4.UsesMiscMode(TextureMiscMode::NODEP)) {
+            LOG_WARNING(HW_GPU, "TLD4.NODEP implementation is incomplete");
+        }
+
+        const auto texture_type = instr.tld4.texture_type.Value();
+        const bool depth_compare = instr.tld4.UsesMiscMode(TextureMiscMode::DC);
+        const bool is_array = instr.tld4.array != 0;
+        WriteTexInstructionFloat(bb, instr,
+                                 GetTld4Code(instr, texture_type, depth_compare, is_array));
+        break;
+    }
+    case OpCode::Id::TLD4S: {
+        UNIMPLEMENTED_IF_MSG(instr.tld4s.UsesMiscMode(TextureMiscMode::AOFFI),
+                             "AOFFI is not implemented");
+        if (instr.tld4s.UsesMiscMode(TextureMiscMode::NODEP)) {
+            LOG_WARNING(HW_GPU, "TLD4S.NODEP implementation is incomplete");
+        }
+
+        const bool depth_compare = instr.tld4s.UsesMiscMode(TextureMiscMode::DC);
+        const Node op_a = GetRegister(instr.gpr8);
+        const Node op_b = GetRegister(instr.gpr20);
+
+        // TODO(Subv): Figure out how the sampler type is encoded in the TLD4S instruction.
+        std::vector<Node> coords;
+        if (depth_compare) {
+            // Note: TLD4S coordinate encoding works just like TEXS's
+            const Node op_y = GetRegister(instr.gpr8.Value() + 1);
+            coords.push_back(op_a);
+            coords.push_back(op_y);
+            coords.push_back(op_b);
+        } else {
+            coords.push_back(op_a);
+            coords.push_back(op_b);
+        }
+        const Node component = Immediate(static_cast<u32>(instr.tld4s.component));
+
+        const auto& sampler =
+            GetSampler(instr.sampler, TextureType::Texture2D, false, depth_compare);
+
+        Node4 values;
+        for (u32 element = 0; element < values.size(); ++element) {
+            auto coords_copy = coords;
+            MetaTexture meta{sampler, {}, {}, {}, {}, component, element};
+            values[element] = Operation(OperationCode::TextureGather, meta, std::move(coords_copy));
+        }
+
+        WriteTexsInstructionFloat(bb, instr, values);
+        break;
+    }
+    case OpCode::Id::TXQ: {
+        if (instr.txq.UsesMiscMode(TextureMiscMode::NODEP)) {
+            LOG_WARNING(HW_GPU, "TXQ.NODEP implementation is incomplete");
+        }
+
+        // TODO: The new commits on the texture refactor, change the way samplers work.
+        // Sadly, not all texture instructions specify the type of texture their sampler
+        // uses. This must be fixed at a later instance.
+        const auto& sampler =
+            GetSampler(instr.sampler, Tegra::Shader::TextureType::Texture2D, false, false);
+
+        u32 indexer = 0;
+        switch (instr.txq.query_type) {
+        case Tegra::Shader::TextureQueryType::Dimension: {
+            for (u32 element = 0; element < 4; ++element) {
+                if (!instr.txq.IsComponentEnabled(element)) {
+                    continue;
+                }
+                MetaTexture meta{sampler, {}, {}, {}, {}, {}, element};
+                const Node value =
+                    Operation(OperationCode::TextureQueryDimensions, meta, GetRegister(instr.gpr8));
+                SetTemporal(bb, indexer++, value);
+            }
+            for (u32 i = 0; i < indexer; ++i) {
+                SetRegister(bb, instr.gpr0.Value() + i, GetTemporal(i));
+            }
+            break;
+        }
+        default:
+            UNIMPLEMENTED_MSG("Unhandled texture query type: {}",
+                              static_cast<u32>(instr.txq.query_type.Value()));
+        }
+        break;
+    }
+    case OpCode::Id::TMML: {
+        UNIMPLEMENTED_IF_MSG(instr.tmml.UsesMiscMode(Tegra::Shader::TextureMiscMode::NDV),
+                             "NDV is not implemented");
+
+        if (instr.tmml.UsesMiscMode(TextureMiscMode::NODEP)) {
+            LOG_WARNING(HW_GPU, "TMML.NODEP implementation is incomplete");
+        }
+
+        auto texture_type = instr.tmml.texture_type.Value();
+        const bool is_array = instr.tmml.array != 0;
+        const auto& sampler = GetSampler(instr.sampler, texture_type, is_array, false);
+
+        std::vector<Node> coords;
+
+        // TODO: Add coordinates for different samplers once other texture types are implemented.
+        switch (texture_type) {
+        case TextureType::Texture1D:
+            coords.push_back(GetRegister(instr.gpr8));
+            break;
+        case TextureType::Texture2D:
+            coords.push_back(GetRegister(instr.gpr8.Value() + 0));
+            coords.push_back(GetRegister(instr.gpr8.Value() + 1));
+            break;
+        default:
+            UNIMPLEMENTED_MSG("Unhandled texture type {}", static_cast<u32>(texture_type));
+
+            // Fallback to interpreting as a 2D texture for now
+            coords.push_back(GetRegister(instr.gpr8.Value() + 0));
+            coords.push_back(GetRegister(instr.gpr8.Value() + 1));
+            texture_type = TextureType::Texture2D;
+        }
+
+        for (u32 element = 0; element < 2; ++element) {
+            auto params = coords;
+            MetaTexture meta{sampler, {}, {}, {}, {}, {}, element};
+            const Node value = Operation(OperationCode::TextureQueryLod, meta, std::move(params));
+            SetTemporal(bb, element, value);
+        }
+        for (u32 element = 0; element < 2; ++element) {
+            SetRegister(bb, instr.gpr0.Value() + element, GetTemporal(element));
+        }
+
+        break;
+    }
+    case OpCode::Id::TLDS: {
+        const Tegra::Shader::TextureType texture_type{instr.tlds.GetTextureType()};
+        const bool is_array{instr.tlds.IsArrayTexture()};
+
+        UNIMPLEMENTED_IF_MSG(instr.tlds.UsesMiscMode(TextureMiscMode::AOFFI),
+                             "AOFFI is not implemented");
+        UNIMPLEMENTED_IF_MSG(instr.tlds.UsesMiscMode(TextureMiscMode::MZ), "MZ is not implemented");
+
+        if (instr.tlds.UsesMiscMode(TextureMiscMode::NODEP)) {
+            LOG_WARNING(HW_GPU, "TLDS.NODEP implementation is incomplete");
+        }
+
+        WriteTexsInstructionFloat(bb, instr, GetTldsCode(instr, texture_type, is_array));
+        break;
+    }
+    default:
+        UNIMPLEMENTED_MSG("Unhandled memory instruction: {}", opcode->get().GetName());
+    }
+
+    return pc;
+}
+
+const Sampler& ShaderIR::GetSampler(const Tegra::Shader::Sampler& sampler, TextureType type,
+                                    bool is_array, bool is_shadow) {
+    const auto offset = static_cast<std::size_t>(sampler.index.Value());
+
+    // If this sampler has already been used, return the existing mapping.
+    const auto itr =
+        std::find_if(used_samplers.begin(), used_samplers.end(),
+                     [&](const Sampler& entry) { return entry.GetOffset() == offset; });
+    if (itr != used_samplers.end()) {
+        ASSERT(itr->GetType() == type && itr->IsArray() == is_array &&
+               itr->IsShadow() == is_shadow);
+        return *itr;
+    }
+
+    // Otherwise create a new mapping for this sampler
+    const std::size_t next_index = used_samplers.size();
+    const Sampler entry{offset, next_index, type, is_array, is_shadow};
+    return *used_samplers.emplace(entry).first;
+}
+
+void ShaderIR::WriteTexInstructionFloat(NodeBlock& bb, Instruction instr, const Node4& components) {
+    u32 dest_elem = 0;
+    for (u32 elem = 0; elem < 4; ++elem) {
+        if (!instr.tex.IsComponentEnabled(elem)) {
+            // Skip disabled components
+            continue;
+        }
+        SetTemporal(bb, dest_elem++, components[elem]);
+    }
+    // After writing values in temporals, move them to the real registers
+    for (u32 i = 0; i < dest_elem; ++i) {
+        SetRegister(bb, instr.gpr0.Value() + i, GetTemporal(i));
+    }
+}
+
+void ShaderIR::WriteTexsInstructionFloat(NodeBlock& bb, Instruction instr,
+                                         const Node4& components) {
+    // TEXS has two destination registers and a swizzle. The first two elements in the swizzle
+    // go into gpr0+0 and gpr0+1, and the rest goes into gpr28+0 and gpr28+1
+
+    u32 dest_elem = 0;
+    for (u32 component = 0; component < 4; ++component) {
+        if (!instr.texs.IsComponentEnabled(component))
+            continue;
+        SetTemporal(bb, dest_elem++, components[component]);
+    }
+
+    for (u32 i = 0; i < dest_elem; ++i) {
+        if (i < 2) {
+            // Write the first two swizzle components to gpr0 and gpr0+1
+            SetRegister(bb, instr.gpr0.Value() + i % 2, GetTemporal(i));
+        } else {
+            ASSERT(instr.texs.HasTwoDestinations());
+            // Write the rest of the swizzle components to gpr28 and gpr28+1
+            SetRegister(bb, instr.gpr28.Value() + i % 2, GetTemporal(i));
+        }
+    }
+}
+
+void ShaderIR::WriteTexsInstructionHalfFloat(NodeBlock& bb, Instruction instr,
+                                             const Node4& components) {
+    // TEXS.F16 destionation registers are packed in two registers in pairs (just like any half
+    // float instruction).
+
+    Node4 values;
+    u32 dest_elem = 0;
+    for (u32 component = 0; component < 4; ++component) {
+        if (!instr.texs.IsComponentEnabled(component))
+            continue;
+        values[dest_elem++] = components[component];
+    }
+    if (dest_elem == 0)
+        return;
+
+    std::generate(values.begin() + dest_elem, values.end(), [&]() { return Immediate(0); });
+
+    const Node first_value = Operation(OperationCode::HPack2, values[0], values[1]);
+    if (dest_elem <= 2) {
+        SetRegister(bb, instr.gpr0, first_value);
+        return;
+    }
+
+    SetTemporal(bb, 0, first_value);
+    SetTemporal(bb, 1, Operation(OperationCode::HPack2, values[2], values[3]));
+
+    SetRegister(bb, instr.gpr0, GetTemporal(0));
+    SetRegister(bb, instr.gpr28, GetTemporal(1));
+}
+
+Node4 ShaderIR::GetTextureCode(Instruction instr, TextureType texture_type,
+                               TextureProcessMode process_mode, std::vector<Node> coords,
+                               Node array, Node depth_compare, u32 bias_offset) {
+    const bool is_array = array;
+    const bool is_shadow = depth_compare;
+
+    UNIMPLEMENTED_IF_MSG((texture_type == TextureType::Texture3D && (is_array || is_shadow)) ||
+                             (texture_type == TextureType::TextureCube && is_array && is_shadow),
+                         "This method is not supported.");
+
+    const auto& sampler = GetSampler(instr.sampler, texture_type, is_array, is_shadow);
+
+    const bool lod_needed = process_mode == TextureProcessMode::LZ ||
+                            process_mode == TextureProcessMode::LL ||
+                            process_mode == TextureProcessMode::LLA;
+
+    // LOD selection (either via bias or explicit textureLod) not supported in GL for
+    // sampler2DArrayShadow and samplerCubeArrayShadow.
+    const bool gl_lod_supported =
+        !((texture_type == Tegra::Shader::TextureType::Texture2D && is_array && is_shadow) ||
+          (texture_type == Tegra::Shader::TextureType::TextureCube && is_array && is_shadow));
+
+    const OperationCode read_method =
+        (lod_needed && gl_lod_supported) ? OperationCode::TextureLod : OperationCode::Texture;
+
+    UNIMPLEMENTED_IF(process_mode != TextureProcessMode::None && !gl_lod_supported);
+
+    Node bias = {};
+    Node lod = {};
+    if (process_mode != TextureProcessMode::None && gl_lod_supported) {
+        switch (process_mode) {
+        case TextureProcessMode::LZ:
+            lod = Immediate(0.0f);
+            break;
+        case TextureProcessMode::LB:
+            // If present, lod or bias are always stored in the register indexed by the gpr20
+            // field with an offset depending on the usage of the other registers
+            bias = GetRegister(instr.gpr20.Value() + bias_offset);
+            break;
+        case TextureProcessMode::LL:
+            lod = GetRegister(instr.gpr20.Value() + bias_offset);
+            break;
+        default:
+            UNIMPLEMENTED_MSG("Unimplemented process mode={}", static_cast<u32>(process_mode));
+            break;
+        }
+    }
+
+    Node4 values;
+    for (u32 element = 0; element < values.size(); ++element) {
+        auto copy_coords = coords;
+        MetaTexture meta{sampler, array, depth_compare, bias, lod, {}, element};
+        values[element] = Operation(read_method, meta, std::move(copy_coords));
+    }
+
+    return values;
+}
+
+Node4 ShaderIR::GetTexCode(Instruction instr, TextureType texture_type,
+                           TextureProcessMode process_mode, bool depth_compare, bool is_array) {
+    const bool lod_bias_enabled =
+        (process_mode != TextureProcessMode::None && process_mode != TextureProcessMode::LZ);
+
+    const auto [coord_count, total_coord_count] = ValidateAndGetCoordinateElement(
+        texture_type, depth_compare, is_array, lod_bias_enabled, 4, 5);
+    // If enabled arrays index is always stored in the gpr8 field
+    const u64 array_register = instr.gpr8.Value();
+    // First coordinate index is the gpr8 or gpr8 + 1 when arrays are used
+    const u64 coord_register = array_register + (is_array ? 1 : 0);
+
+    std::vector<Node> coords;
+    for (std::size_t i = 0; i < coord_count; ++i) {
+        coords.push_back(GetRegister(coord_register + i));
+    }
+    // 1D.DC in OpenGL the 2nd component is ignored.
+    if (depth_compare && !is_array && texture_type == TextureType::Texture1D) {
+        coords.push_back(Immediate(0.0f));
+    }
+
+    const Node array = is_array ? GetRegister(array_register) : nullptr;
+
+    Node dc{};
+    if (depth_compare) {
+        // Depth is always stored in the register signaled by gpr20 or in the next register if lod
+        // or bias are used
+        const u64 depth_register = instr.gpr20.Value() + (lod_bias_enabled ? 1 : 0);
+        dc = GetRegister(depth_register);
+    }
+
+    return GetTextureCode(instr, texture_type, process_mode, coords, array, dc, 0);
+}
+
+Node4 ShaderIR::GetTexsCode(Instruction instr, TextureType texture_type,
+                            TextureProcessMode process_mode, bool depth_compare, bool is_array) {
+    const bool lod_bias_enabled =
+        (process_mode != TextureProcessMode::None && process_mode != TextureProcessMode::LZ);
+
+    const auto [coord_count, total_coord_count] = ValidateAndGetCoordinateElement(
+        texture_type, depth_compare, is_array, lod_bias_enabled, 4, 4);
+    // If enabled arrays index is always stored in the gpr8 field
+    const u64 array_register = instr.gpr8.Value();
+    // First coordinate index is stored in gpr8 field or (gpr8 + 1) when arrays are used
+    const u64 coord_register = array_register + (is_array ? 1 : 0);
+    const u64 last_coord_register =
+        (is_array || !(lod_bias_enabled || depth_compare) || (coord_count > 2))
+            ? static_cast<u64>(instr.gpr20.Value())
+            : coord_register + 1;
+    const u32 bias_offset = coord_count > 2 ? 1 : 0;
+
+    std::vector<Node> coords;
+    for (std::size_t i = 0; i < coord_count; ++i) {
+        const bool last = (i == (coord_count - 1)) && (coord_count > 1);
+        coords.push_back(GetRegister(last ? last_coord_register : coord_register + i));
+    }
+
+    const Node array = is_array ? GetRegister(array_register) : nullptr;
+
+    Node dc{};
+    if (depth_compare) {
+        // Depth is always stored in the register signaled by gpr20 or in the next register if lod
+        // or bias are used
+        const u64 depth_register = instr.gpr20.Value() + (lod_bias_enabled ? 1 : 0);
+        dc = GetRegister(depth_register);
+    }
+
+    return GetTextureCode(instr, texture_type, process_mode, coords, array, dc, bias_offset);
+}
+
+Node4 ShaderIR::GetTld4Code(Instruction instr, TextureType texture_type, bool depth_compare,
+                            bool is_array) {
+    const std::size_t coord_count = GetCoordCount(texture_type);
+    const std::size_t total_coord_count = coord_count + (is_array ? 1 : 0);
+    const std::size_t total_reg_count = total_coord_count + (depth_compare ? 1 : 0);
+
+    // If enabled arrays index is always stored in the gpr8 field
+    const u64 array_register = instr.gpr8.Value();
+    // First coordinate index is the gpr8 or gpr8 + 1 when arrays are used
+    const u64 coord_register = array_register + (is_array ? 1 : 0);
+
+    std::vector<Node> coords;
+    for (size_t i = 0; i < coord_count; ++i)
+        coords.push_back(GetRegister(coord_register + i));
+
+    const auto& sampler = GetSampler(instr.sampler, texture_type, is_array, depth_compare);
+
+    Node4 values;
+    for (u32 element = 0; element < values.size(); ++element) {
+        auto coords_copy = coords;
+        MetaTexture meta{sampler, GetRegister(array_register), {}, {}, {}, {}, element};
+        values[element] = Operation(OperationCode::TextureGather, meta, std::move(coords_copy));
+    }
+
+    return values;
+}
+
+Node4 ShaderIR::GetTldsCode(Instruction instr, TextureType texture_type, bool is_array) {
+    const std::size_t type_coord_count = GetCoordCount(texture_type);
+    const bool lod_enabled = instr.tlds.GetTextureProcessMode() == TextureProcessMode::LL;
+
+    // If enabled arrays index is always stored in the gpr8 field
+    const u64 array_register = instr.gpr8.Value();
+    // if is array gpr20 is used
+    const u64 coord_register = is_array ? instr.gpr20.Value() : instr.gpr8.Value();
+
+    const u64 last_coord_register =
+        ((type_coord_count > 2) || (type_coord_count == 2 && !lod_enabled)) && !is_array
+            ? static_cast<u64>(instr.gpr20.Value())
+            : coord_register + 1;
+
+    std::vector<Node> coords;
+    for (std::size_t i = 0; i < type_coord_count; ++i) {
+        const bool last = (i == (type_coord_count - 1)) && (type_coord_count > 1);
+        coords.push_back(GetRegister(last ? last_coord_register : coord_register + i));
+    }
+
+    const Node array = is_array ? GetRegister(array_register) : nullptr;
+    // When lod is used always is in gpr20
+    const Node lod = lod_enabled ? GetRegister(instr.gpr20) : Immediate(0);
+
+    const auto& sampler = GetSampler(instr.sampler, texture_type, is_array, false);
+
+    Node4 values;
+    for (u32 element = 0; element < values.size(); ++element) {
+        auto coords_copy = coords;
+        MetaTexture meta{sampler, array, {}, {}, lod, {}, element};
+        values[element] = Operation(OperationCode::TexelFetch, meta, std::move(coords_copy));
+    }
+    return values;
+}
+
+std::tuple<std::size_t, std::size_t> ShaderIR::ValidateAndGetCoordinateElement(
+    TextureType texture_type, bool depth_compare, bool is_array, bool lod_bias_enabled,
+    std::size_t max_coords, std::size_t max_inputs) {
+    const std::size_t coord_count = GetCoordCount(texture_type);
+
+    std::size_t total_coord_count = coord_count + (is_array ? 1 : 0) + (depth_compare ? 1 : 0);
+    const std::size_t total_reg_count = total_coord_count + (lod_bias_enabled ? 1 : 0);
+    if (total_coord_count > max_coords || total_reg_count > max_inputs) {
+        UNIMPLEMENTED_MSG("Unsupported Texture operation");
+        total_coord_count = std::min(total_coord_count, max_coords);
+    }
+    // 1D.DC OpenGL is using a vec3 but 2nd component is ignored later.
+    total_coord_count +=
+        (depth_compare && !is_array && texture_type == TextureType::Texture1D) ? 1 : 0;
+
+    return {coord_count, total_coord_count};
+}
+
+} // namespace VideoCommon::Shader

src/video_core/shader/shader_ir.h

@@ -290,7 +290,9 @@ struct MetaTexture {
     const Sampler& sampler;
     Node array{};
     Node depth_compare{};
-    std::vector<Node> extras;
+    Node bias{};
+    Node lod{};
+    Node component{};
     u32 element{};
 };
 
@@ -614,6 +616,7 @@ private:
     u32 DecodeHfma2(NodeBlock& bb, u32 pc);
     u32 DecodeConversion(NodeBlock& bb, u32 pc);
     u32 DecodeMemory(NodeBlock& bb, u32 pc);
+    u32 DecodeTexture(NodeBlock& bb, u32 pc);
     u32 DecodeFloatSetPredicate(NodeBlock& bb, u32 pc);
     u32 DecodeIntegerSetPredicate(NodeBlock& bb, u32 pc);
     u32 DecodeHalfSetPredicate(NodeBlock& bb, u32 pc);

src/video_core/surface.cpp

@@ -89,8 +89,6 @@ PixelFormat PixelFormatFromDepthFormat(Tegra::DepthFormat format) {
 
 PixelFormat PixelFormatFromRenderTargetFormat(Tegra::RenderTargetFormat format) {
     switch (format) {
-        // TODO (Hexagon12): Converting SRGBA to RGBA is a hack and doesn't completely correct the
-        // gamma.
     case Tegra::RenderTargetFormat::RGBA8_SRGB:
         return PixelFormat::RGBA8_SRGB;
     case Tegra::RenderTargetFormat::RGBA8_UNORM:

src/video_core/textures/astc.cpp

@@ -23,28 +23,12 @@
 
 #include "video_core/textures/astc.h"
 
-class BitStream {
+class InputBitStream {
 public:
-    explicit BitStream(unsigned char* ptr, int nBits = 0, int start_offset = 0)
+    explicit InputBitStream(const unsigned char* ptr, int nBits = 0, int start_offset = 0)
         : m_NumBits(nBits), m_CurByte(ptr), m_NextBit(start_offset % 8) {}
 
-    ~BitStream() = default;
-
-    int GetBitsWritten() const {
-        return m_BitsWritten;
-    }
-
-    void WriteBitsR(unsigned int val, unsigned int nBits) {
-        for (unsigned int i = 0; i < nBits; i++) {
-            WriteBit((val >> (nBits - i - 1)) & 1);
-        }
-    }
-
-    void WriteBits(unsigned int val, unsigned int nBits) {
-        for (unsigned int i = 0; i < nBits; i++) {
-            WriteBit((val >> i) & 1);
-        }
-    }
+    ~InputBitStream() = default;
 
     int GetBitsRead() const {
         return m_BitsRead;
@@ -70,6 +54,38 @@ public:
         return ret;
     }
 
+private:
+    const int m_NumBits;
+    const unsigned char* m_CurByte;
+    int m_NextBit = 0;
+    int m_BitsRead = 0;
+
+    bool done = false;
+};
+
+class OutputBitStream {
+public:
+    explicit OutputBitStream(unsigned char* ptr, int nBits = 0, int start_offset = 0)
+        : m_NumBits(nBits), m_CurByte(ptr), m_NextBit(start_offset % 8) {}
+
+    ~OutputBitStream() = default;
+
+    int GetBitsWritten() const {
+        return m_BitsWritten;
+    }
+
+    void WriteBitsR(unsigned int val, unsigned int nBits) {
+        for (unsigned int i = 0; i < nBits; i++) {
+            WriteBit((val >> (nBits - i - 1)) & 1);
+        }
+    }
+
+    void WriteBits(unsigned int val, unsigned int nBits) {
+        for (unsigned int i = 0; i < nBits; i++) {
+            WriteBit((val >> i) & 1);
+        }
+    }
+
 private:
     void WriteBit(int b) {
 
@@ -238,8 +254,8 @@ public:
     // Fills result with the values that are encoded in the given
     // bitstream. We must know beforehand what the maximum possible
     // value is, and how many values we're decoding.
-    static void DecodeIntegerSequence(std::vector<IntegerEncodedValue>& result, BitStream& bits,
-                                      uint32_t maxRange, uint32_t nValues) {
+    static void DecodeIntegerSequence(std::vector<IntegerEncodedValue>& result,
+                                      InputBitStream& bits, uint32_t maxRange, uint32_t nValues) {
         // Determine encoding parameters
         IntegerEncodedValue val = IntegerEncodedValue::CreateEncoding(maxRange);
 
@@ -267,7 +283,7 @@ public:
     }
 
 private:
-    static void DecodeTritBlock(BitStream& bits, std::vector<IntegerEncodedValue>& result,
+    static void DecodeTritBlock(InputBitStream& bits, std::vector<IntegerEncodedValue>& result,
                                 uint32_t nBitsPerValue) {
         // Implement the algorithm in section C.2.12
         uint32_t m[5];
@@ -327,7 +343,7 @@ private:
         }
     }
 
-    static void DecodeQuintBlock(BitStream& bits, std::vector<IntegerEncodedValue>& result,
+    static void DecodeQuintBlock(InputBitStream& bits, std::vector<IntegerEncodedValue>& result,
                                  uint32_t nBitsPerValue) {
         // Implement the algorithm in section C.2.12
         uint32_t m[3];
@@ -406,7 +422,7 @@ struct TexelWeightParams {
     }
 };
 
-static TexelWeightParams DecodeBlockInfo(BitStream& strm) {
+static TexelWeightParams DecodeBlockInfo(InputBitStream& strm) {
     TexelWeightParams params;
 
     // Read the entire block mode all at once
@@ -605,7 +621,7 @@ static TexelWeightParams DecodeBlockInfo(BitStream& strm) {
     return params;
 }
 
-static void FillVoidExtentLDR(BitStream& strm, uint32_t* const outBuf, uint32_t blockWidth,
+static void FillVoidExtentLDR(InputBitStream& strm, uint32_t* const outBuf, uint32_t blockWidth,
                               uint32_t blockHeight) {
     // Don't actually care about the void extent, just read the bits...
     for (int i = 0; i < 4; ++i) {
@@ -821,7 +837,7 @@ static void DecodeColorValues(uint32_t* out, uint8_t* data, const uint32_t* mode
 
     // We now have enough to decode our integer sequence.
     std::vector<IntegerEncodedValue> decodedColorValues;
-    BitStream colorStream(data);
+    InputBitStream colorStream(data);
     IntegerEncodedValue::DecodeIntegerSequence(decodedColorValues, colorStream, range, nValues);
 
     // Once we have the decoded values, we need to dequantize them to the 0-255 range
@@ -1365,9 +1381,9 @@ static void ComputeEndpoints(Pixel& ep1, Pixel& ep2, const uint32_t*& colorValue
 #undef READ_INT_VALUES
 }
 
-static void DecompressBlock(uint8_t inBuf[16], const uint32_t blockWidth,
+static void DecompressBlock(const uint8_t inBuf[16], const uint32_t blockWidth,
                             const uint32_t blockHeight, uint32_t* outBuf) {
-    BitStream strm(inBuf);
+    InputBitStream strm(inBuf);
     TexelWeightParams weightParams = DecodeBlockInfo(strm);
 
     // Was there an error?
@@ -1421,7 +1437,7 @@ static void DecompressBlock(uint8_t inBuf[16], const uint32_t blockWidth,
     // Define color data.
     uint8_t colorEndpointData[16];
     memset(colorEndpointData, 0, sizeof(colorEndpointData));
-    BitStream colorEndpointStream(colorEndpointData, 16 * 8, 0);
+    OutputBitStream colorEndpointStream(colorEndpointData, 16 * 8, 0);
 
     // Read extra config data...
     uint32_t baseCEM = 0;
@@ -1549,7 +1565,7 @@ static void DecompressBlock(uint8_t inBuf[16], const uint32_t blockWidth,
     memset(texelWeightData + clearByteStart, 0, 16 - clearByteStart);
 
     std::vector<IntegerEncodedValue> texelWeightValues;
-    BitStream weightStream(texelWeightData);
+    InputBitStream weightStream(texelWeightData);
 
     IntegerEncodedValue::DecodeIntegerSequence(texelWeightValues, weightStream,
                                                weightParams.m_MaxWeight,
@@ -1597,7 +1613,7 @@ static void DecompressBlock(uint8_t inBuf[16], const uint32_t blockWidth,
 
 namespace Tegra::Texture::ASTC {
 
-std::vector<uint8_t> Decompress(std::vector<uint8_t>& data, uint32_t width, uint32_t height,
+std::vector<uint8_t> Decompress(const uint8_t* data, uint32_t width, uint32_t height,
                                 uint32_t depth, uint32_t block_width, uint32_t block_height) {
     uint32_t blockIdx = 0;
     std::vector<uint8_t> outData(height * width * depth * 4);
@@ -1605,7 +1621,7 @@ std::vector<uint8_t> Decompress(std::vector<uint8_t>& data, uint32_t width, uint
         for (uint32_t j = 0; j < height; j += block_height) {
             for (uint32_t i = 0; i < width; i += block_width) {
 
-                uint8_t* blockPtr = data.data() + blockIdx * 16;
+                const uint8_t* blockPtr = data + blockIdx * 16;
 
                 // Blocks can be at most 12x12
                 uint32_t uncompData[144];

src/video_core/textures/astc.h

@@ -9,7 +9,7 @@
 
 namespace Tegra::Texture::ASTC {
 
-std::vector<uint8_t> Decompress(std::vector<uint8_t>& data, uint32_t width, uint32_t height,
+std::vector<uint8_t> Decompress(const uint8_t* data, uint32_t width, uint32_t height,
                                 uint32_t depth, uint32_t block_width, uint32_t block_height);
 
 } // namespace Tegra::Texture::ASTC

src/video_core/textures/convert.cpp

@@ -0,0 +1,92 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include <cstring>
+#include <tuple>
+#include <vector>
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "common/logging/log.h"
+#include "video_core/textures/astc.h"
+#include "video_core/textures/convert.h"
+
+namespace Tegra::Texture {
+
+using VideoCore::Surface::PixelFormat;
+
+template <bool reverse>
+void SwapS8Z24ToZ24S8(u8* data, u32 width, u32 height) {
+    union S8Z24 {
+        BitField<0, 24, u32> z24;
+        BitField<24, 8, u32> s8;
+    };
+    static_assert(sizeof(S8Z24) == 4, "S8Z24 is incorrect size");
+
+    union Z24S8 {
+        BitField<0, 8, u32> s8;
+        BitField<8, 24, u32> z24;
+    };
+    static_assert(sizeof(Z24S8) == 4, "Z24S8 is incorrect size");
+
+    S8Z24 s8z24_pixel{};
+    Z24S8 z24s8_pixel{};
+    constexpr auto bpp{
+        VideoCore::Surface::GetBytesPerPixel(VideoCore::Surface::PixelFormat::S8Z24)};
+    for (std::size_t y = 0; y < height; ++y) {
+        for (std::size_t x = 0; x < width; ++x) {
+            const std::size_t offset{bpp * (y * width + x)};
+            if constexpr (reverse) {
+                std::memcpy(&z24s8_pixel, &data[offset], sizeof(Z24S8));
+                s8z24_pixel.s8.Assign(z24s8_pixel.s8);
+                s8z24_pixel.z24.Assign(z24s8_pixel.z24);
+                std::memcpy(&data[offset], &s8z24_pixel, sizeof(S8Z24));
+            } else {
+                std::memcpy(&s8z24_pixel, &data[offset], sizeof(S8Z24));
+                z24s8_pixel.s8.Assign(s8z24_pixel.s8);
+                z24s8_pixel.z24.Assign(s8z24_pixel.z24);
+                std::memcpy(&data[offset], &z24s8_pixel, sizeof(Z24S8));
+            }
+        }
+    }
+}
+
+static void ConvertS8Z24ToZ24S8(u8* data, u32 width, u32 height) {
+    SwapS8Z24ToZ24S8<false>(data, width, height);
+}
+
+static void ConvertZ24S8ToS8Z24(u8* data, u32 width, u32 height) {
+    SwapS8Z24ToZ24S8<true>(data, width, height);
+}
+
+void ConvertFromGuestToHost(u8* data, PixelFormat pixel_format, u32 width, u32 height, u32 depth,
+                            bool convert_astc, bool convert_s8z24) {
+    if (convert_astc && IsPixelFormatASTC(pixel_format)) {
+        // Convert ASTC pixel formats to RGBA8, as most desktop GPUs do not support ASTC.
+        u32 block_width{};
+        u32 block_height{};
+        std::tie(block_width, block_height) = GetASTCBlockSize(pixel_format);
+        const std::vector<u8> rgba8_data =
+            Tegra::Texture::ASTC::Decompress(data, width, height, depth, block_width, block_height);
+        std::copy(rgba8_data.begin(), rgba8_data.end(), data);
+
+    } else if (convert_s8z24 && pixel_format == PixelFormat::S8Z24) {
+        Tegra::Texture::ConvertS8Z24ToZ24S8(data, width, height);
+    }
+}
+
+void ConvertFromHostToGuest(u8* data, PixelFormat pixel_format, u32 width, u32 height, u32 depth,
+                            bool convert_astc, bool convert_s8z24) {
+    if (convert_astc && IsPixelFormatASTC(pixel_format)) {
+        LOG_CRITICAL(HW_GPU, "Conversion of format {} after texture flushing is not implemented",
+                     static_cast<u32>(pixel_format));
+        UNREACHABLE();
+
+    } else if (convert_s8z24 && pixel_format == PixelFormat::S8Z24) {
+        Tegra::Texture::ConvertZ24S8ToS8Z24(data, width, height);
+    }
+}
+
+} // namespace Tegra::Texture

src/video_core/textures/convert.h

@@ -0,0 +1,18 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "common/common_types.h"
+#include "video_core/surface.h"
+
+namespace Tegra::Texture {
+
+void ConvertFromGuestToHost(u8* data, VideoCore::Surface::PixelFormat pixel_format, u32 width,
+                            u32 height, u32 depth, bool convert_astc, bool convert_s8z24);
+
+void ConvertFromHostToGuest(u8* data, VideoCore::Surface::PixelFormat pixel_format, u32 width,
+                            u32 height, u32 depth, bool convert_astc, bool convert_s8z24);
+
+} // namespace Tegra::Texture

src/video_core/textures/decoders.cpp

@@ -103,8 +103,8 @@ void FastProcessBlock(u8* const swizzled_data, u8* const unswizzled_data, const
                 const u32 swizzle_offset{y_address + table[(xb / fast_swizzle_align) % 4]};
                 const u32 out_x = xb * out_bytes_per_pixel / bytes_per_pixel;
                 const u32 pixel_index{out_x + pixel_base};
-                data_ptrs[unswizzle] = swizzled_data + swizzle_offset;
-                data_ptrs[!unswizzle] = unswizzled_data + pixel_index;
+                data_ptrs[unswizzle ? 1 : 0] = swizzled_data + swizzle_offset;
+                data_ptrs[unswizzle ? 0 : 1] = unswizzled_data + pixel_index;
                 std::memcpy(data_ptrs[0], data_ptrs[1], fast_swizzle_align);
             }
             pixel_base += stride_x;
@@ -154,7 +154,7 @@ void SwizzledData(u8* const swizzled_data, u8* const unswizzled_data, const bool
             for (u32 xb = 0; xb < blocks_on_x; xb++) {
                 const u32 x_start = xb * block_x_elements;
                 const u32 x_end = std::min(width, x_start + block_x_elements);
-                if (fast) {
+                if constexpr (fast) {
                     FastProcessBlock(swizzled_data, unswizzled_data, unswizzle, x_start, y_start,
                                      z_start, x_end, y_end, z_end, tile_offset, xy_block_size,
                                      layer_z, stride_x, bytes_per_pixel, out_bytes_per_pixel);

src/video_core/textures/decoders.h

@@ -16,16 +16,13 @@ inline std::size_t GetGOBSize() {
     return 512;
 }
 
-/**
- * Unswizzles a swizzled texture without changing its format.
- */
+/// Unswizzles a swizzled texture without changing its format.
 void UnswizzleTexture(u8* unswizzled_data, VAddr address, u32 tile_size_x, u32 tile_size_y,
                       u32 bytes_per_pixel, u32 width, u32 height, u32 depth,
                       u32 block_height = TICEntry::DefaultBlockHeight,
                       u32 block_depth = TICEntry::DefaultBlockHeight, u32 width_spacing = 0);
-/**
- * Unswizzles a swizzled texture without changing its format.
- */
+
+/// Unswizzles a swizzled texture without changing its format.
 std::vector<u8> UnswizzleTexture(VAddr address, u32 tile_size_x, u32 tile_size_y,
                                  u32 bytes_per_pixel, u32 width, u32 height, u32 depth,
                                  u32 block_height = TICEntry::DefaultBlockHeight,
@@ -37,15 +34,11 @@ void CopySwizzledData(u32 width, u32 height, u32 depth, u32 bytes_per_pixel,
                       u32 out_bytes_per_pixel, u8* swizzled_data, u8* unswizzled_data,
                       bool unswizzle, u32 block_height, u32 block_depth, u32 width_spacing);
 
-/**
- * Decodes an unswizzled texture into a A8R8G8B8 texture.
- */
+/// Decodes an unswizzled texture into a A8R8G8B8 texture.
 std::vector<u8> DecodeTexture(const std::vector<u8>& texture_data, TextureFormat format, u32 width,
                               u32 height);
 
-/**
- * This function calculates the correct size of a texture depending if it's tiled or not.
- */
+/// This function calculates the correct size of a texture depending if it's tiled or not.
 std::size_t CalculateSize(bool tiled, u32 bytes_per_pixel, u32 width, u32 height, u32 depth,
                           u32 block_height, u32 block_depth);
 
@@ -53,6 +46,7 @@ std::size_t CalculateSize(bool tiled, u32 bytes_per_pixel, u32 width, u32 height
 void SwizzleSubrect(u32 subrect_width, u32 subrect_height, u32 source_pitch, u32 swizzled_width,
                     u32 bytes_per_pixel, VAddr swizzled_data, VAddr unswizzled_data,
                     u32 block_height);
+
 /// Copies a tiled subrectangle into a linear surface.
 void UnswizzleSubrect(u32 subrect_width, u32 subrect_height, u32 dest_pitch, u32 swizzled_width,
                       u32 bytes_per_pixel, VAddr swizzled_data, VAddr unswizzled_data,

src/web_service/verify_login.h

@@ -4,8 +4,6 @@
 
 #pragma once
 
-#include <functional>
-#include <future>
 #include <string>
 
 namespace WebService {

src/web_service/web_backend.cpp

@@ -10,7 +10,6 @@
 #include "common/common_types.h"
 #include "common/logging/log.h"
 #include "common/web_result.h"
-#include "core/settings.h"
 #include "web_service/web_backend.h"
 
 namespace WebService {

src/yuzu/applets/web_browser.cpp

@@ -56,6 +56,8 @@ constexpr char NX_SHIM_INJECT_SCRIPT[] = R"(
     window.nx.endApplet = function() {
         applet_done = true;
     };
+
+    window.onkeypress = function(e) { if (e.keyCode === 13) { applet_done = true; } };
 )";
 
 QString GetNXShimInjectionScript() {

src/yuzu/bootmanager.cpp

@@ -20,10 +20,7 @@
 EmuThread::EmuThread(GRenderWindow* render_window) : render_window(render_window) {}
 
 void EmuThread::run() {
-    if (!Settings::values.use_multi_core) {
-        // Single core mode must acquire OpenGL context for entire emulation session
-        render_window->MakeCurrent();
-    }
+    render_window->MakeCurrent();
 
     MicroProfileOnThreadCreate("EmuThread");
 
@@ -38,6 +35,11 @@ void EmuThread::run() {
 
     emit LoadProgress(VideoCore::LoadCallbackStage::Complete, 0, 0);
 
+    if (Settings::values.use_asynchronous_gpu_emulation) {
+        // Release OpenGL context for the GPU thread
+        render_window->DoneCurrent();
+    }
+
     // holds whether the cpu was running during the last iteration,
     // so that the DebugModeLeft signal can be emitted before the
     // next execution step

src/yuzu/compatdb.cpp

@@ -53,8 +53,8 @@ void CompatDB::Submit() {
     case CompatDBPage::Final:
         back();
         LOG_DEBUG(Frontend, "Compatibility Rating: {}", compatibility->checkedId());
-        Core::Telemetry().AddField(Telemetry::FieldType::UserFeedback, "Compatibility",
-                                   compatibility->checkedId());
+        Core::System::GetInstance().TelemetrySession().AddField(
+            Telemetry::FieldType::UserFeedback, "Compatibility", compatibility->checkedId());
 
         button(NextButton)->setEnabled(false);
         button(NextButton)->setText(tr("Submitting"));

src/yuzu/configuration/config.cpp

@@ -374,6 +374,8 @@ void Config::ReadValues() {
         qt_config->value("use_disk_shader_cache", false).toBool();
     Settings::values.use_accurate_gpu_emulation =
         qt_config->value("use_accurate_gpu_emulation", false).toBool();
+    Settings::values.use_asynchronous_gpu_emulation =
+        qt_config->value("use_asynchronous_gpu_emulation", false).toBool();
 
     Settings::values.bg_red = qt_config->value("bg_red", 0.0).toFloat();
     Settings::values.bg_green = qt_config->value("bg_green", 0.0).toFloat();
@@ -633,6 +635,8 @@ void Config::SaveValues() {
     qt_config->setValue("frame_limit", Settings::values.frame_limit);
     qt_config->setValue("use_disk_shader_cache", Settings::values.use_disk_shader_cache);
     qt_config->setValue("use_accurate_gpu_emulation", Settings::values.use_accurate_gpu_emulation);
+    qt_config->setValue("use_asynchronous_gpu_emulation",
+                        Settings::values.use_asynchronous_gpu_emulation);
 
     // Cast to double because Qt's written float values are not human-readable
     qt_config->setValue("bg_red", (double)Settings::values.bg_red);

src/yuzu/configuration/configure_graphics.cpp

@@ -75,6 +75,8 @@ void ConfigureGraphics::setConfiguration() {
     ui->frame_limit->setValue(Settings::values.frame_limit);
     ui->use_disk_shader_cache->setChecked(Settings::values.use_disk_shader_cache);
     ui->use_accurate_gpu_emulation->setChecked(Settings::values.use_accurate_gpu_emulation);
+    ui->use_asynchronous_gpu_emulation->setEnabled(!Core::System::GetInstance().IsPoweredOn());
+    ui->use_asynchronous_gpu_emulation->setChecked(Settings::values.use_asynchronous_gpu_emulation);
     UpdateBackgroundColorButton(QColor::fromRgbF(Settings::values.bg_red, Settings::values.bg_green,
                                                  Settings::values.bg_blue));
 }
@@ -86,6 +88,8 @@ void ConfigureGraphics::applyConfiguration() {
     Settings::values.frame_limit = ui->frame_limit->value();
     Settings::values.use_disk_shader_cache = ui->use_disk_shader_cache->isChecked();
     Settings::values.use_accurate_gpu_emulation = ui->use_accurate_gpu_emulation->isChecked();
+    Settings::values.use_asynchronous_gpu_emulation =
+        ui->use_asynchronous_gpu_emulation->isChecked();
     Settings::values.bg_red = static_cast<float>(bg_color.redF());
     Settings::values.bg_green = static_cast<float>(bg_color.greenF());
     Settings::values.bg_blue = static_cast<float>(bg_color.blueF());

src/yuzu/configuration/configure_graphics.ui

@@ -63,6 +63,13 @@
           </property>
          </widget>
         </item>
+        <item>
+         <widget class="QCheckBox" name="use_asynchronous_gpu_emulation">
+          <property name="text">
+           <string>Use asynchronous GPU emulation</string>
+          </property>
+         </widget>
+        </item>
         <item>
          <layout class="QHBoxLayout" name="horizontalLayout">
           <item>