"Merge PR 1340"

externals: Update fmt to master

.ci/scripts/merge/apply-patches-by-label-private.py

@@ -0,0 +1,41 @@
+# Download all pull requests as patches that match a specific label
+# Usage: python download-patches-by-label.py <Label to Match> <Root Path Folder to DL to>
+
+import requests, sys, json, urllib3.request, shutil, subprocess, os, traceback
+
+org = os.getenv("PrivateMergeOrg".upper(), "yuzu-emu")
+repo = os.getenv("PrivateMergeRepo".upper(), "yuzu-private")
+tagline = os.getenv("MergeTaglinePrivate".upper(), "")
+user = sys.argv[1]
+
+http = urllib3.PoolManager()
+dl_list = {}
+
+def check_individual(repo_id, pr_id):
+    url = 'https://%sdev.azure.com/%s/%s/_apis/git/repositories/%s/pullRequests/%s/labels?api-version=5.1-preview.1' % (user, org, repo, repo_id, pr_id)
+    response = requests.get(url)
+    if (response.ok):
+        j = json.loads(response.content)
+        for tg in j['value']:
+            if (tg['name'] == sys.argv[2]):
+                return True
+    return False
+
+try:
+    url = 'https://%sdev.azure.com/%s/%s/_apis/git/pullrequests?api-version=5.1' % (user, org, repo)
+    response = requests.get(url)
+    if (response.ok):
+        j = json.loads(response.content)
+        for pr in j["value"]:
+            repo_id = pr['repository']['id']
+            pr_id = pr['pullRequestId']
+            if (check_individual(repo_id, pr_id)):
+                pn = pr_id
+                ref  = pr['sourceRefName']
+                print("Matched PR# %s" % pn)
+                print(subprocess.check_output(["git", "fetch", "https://%sdev.azure.com/%s/_git/%s" % (user, org, repo), ref, "-f"]))
+                print(subprocess.check_output(["git", "merge", "--squash", 'origin/' + ref.replace('refs/heads/','')]))
+                print(subprocess.check_output(["git", "commit", "-m\"Merge %s PR %s\"" % (tagline, pn)]))
+except:
+    traceback.print_exc(file=sys.stdout)
+    sys.exit(-1)

.ci/scripts/merge/apply-patches-by-label.py

@@ -1,7 +1,9 @@
 # Download all pull requests as patches that match a specific label
 # Usage: python download-patches-by-label.py <Label to Match> <Root Path Folder to DL to>
 
-import requests, sys, json, urllib3.request, shutil, subprocess
+import requests, sys, json, urllib3.request, shutil, subprocess, os
+
+tagline = os.getenv("MergeTaglinePublic".upper(), "")
 
 http = urllib3.PoolManager()
 dl_list = {}
@@ -23,6 +25,6 @@ try:
                 print("Matched PR# %s" % pn)
                 print(subprocess.check_output(["git", "fetch", "https://github.com/yuzu-emu/yuzu.git", "pull/%s/head:pr-%s" % (pn, pn), "-f"]))
                 print(subprocess.check_output(["git", "merge", "--squash", "pr-%s" % pn]))
-                print(subprocess.check_output(["git", "commit", "-m\"Merge PR %s\"" % pn]))
+                print(subprocess.check_output(["git", "commit", "-m\"Merge %s PR %s\"" % (tagline, pn)]))
 except:
     sys.exit(-1)

.ci/templates/merge-private.yml

@@ -0,0 +1,47 @@
+jobs:
+- job: merge
+  displayName: 'pull requests'
+  steps:
+  - checkout: self
+    submodules: recursive
+  - template: ./mergebot-private.yml
+    parameters:
+      matchLabel: '$(BuildName)-merge'
+      matchLabelPublic: '$(PublicBuildName)-merge'
+  - task: ArchiveFiles@2
+    displayName: 'Package Source'
+    inputs:
+      rootFolderOrFile: '$(System.DefaultWorkingDirectory)'
+      includeRootFolder: false
+      archiveType: '7z'
+      archiveFile: '$(Build.ArtifactStagingDirectory)/yuzu-$(BuildName)-source.7z'
+  - task: PublishPipelineArtifact@1
+    displayName: 'Upload Artifacts'
+    inputs:
+      targetPath: '$(Build.ArtifactStagingDirectory)/yuzu-$(BuildName)-source.7z'
+      artifact: 'yuzu-$(BuildName)-source'
+      replaceExistingArchive: true
+- job: upload_source
+  displayName: 'upload'
+  dependsOn: merge
+  steps:
+  - template: ./sync-source.yml
+    parameters:
+      artifactSource: 'true'
+      needSubmodules: 'true'
+  - script: chmod a+x $(System.DefaultWorkingDirectory)/.ci/scripts/merge/yuzubot-git-config.sh && $(System.DefaultWorkingDirectory)/.ci/scripts/merge/yuzubot-git-config.sh
+    displayName: 'Apply Git Configuration'
+  - script: git tag -a $(BuildName)-$(Build.BuildId) -m "yuzu $(BuildName) $(Build.BuildNumber) $(Build.DefinitionName)"
+    displayName: 'Tag Source'
+  - script: git remote add other $(GitRepoPushChangesURL)
+    displayName: 'Register Repository'
+  - script: git push --follow-tags --force other HEAD:$(GitPushBranch)
+    displayName: 'Update Code'
+  - script: git rev-list -n 1 $(BuildName)-$(Build.BuildId) > $(Build.ArtifactStagingDirectory)/tag-commit.sha
+    displayName: 'Calculate Release Point'
+  - task: PublishPipelineArtifact@1
+    displayName: 'Upload Release Point'
+    inputs:
+      targetPath: '$(Build.ArtifactStagingDirectory)/tag-commit.sha'
+      artifact: 'yuzu-$(BuildName)-release-point'
+      replaceExistingArchive: true

.ci/templates/mergebot-private.yml

@@ -0,0 +1,23 @@
+parameters:
+  matchLabel: 'dummy-merge'
+  matchLabelPublic: 'dummy-merge'
+
+steps:
+  - script: mkdir $(System.DefaultWorkingDirectory)/patches && pip install requests urllib3
+    displayName: 'Prepare Environment'
+  - script: chmod a+x $(System.DefaultWorkingDirectory)/.ci/scripts/merge/yuzubot-git-config.sh && $(System.DefaultWorkingDirectory)/.ci/scripts/merge/yuzubot-git-config.sh
+    displayName: 'Apply Git Configuration'
+  - task: PythonScript@0
+    displayName: 'Discover, Download, and Apply Patches'
+    inputs:
+      scriptSource: 'filePath'
+      scriptPath: '.ci/scripts/merge/apply-patches-by-label.py'
+      arguments: '${{ parameters.matchLabelPublic }} patches-public'
+      workingDirectory: '$(System.DefaultWorkingDirectory)'
+  - task: PythonScript@0
+    displayName: 'Discover, Download, and Apply Patches'
+    inputs:
+      scriptSource: 'filePath'
+      scriptPath: '.ci/scripts/merge/apply-patches-by-label-private.py'
+      arguments: '$(PrivateMergeUser) ${{ parameters.matchLabel }} patches-private'
+      workingDirectory: '$(System.DefaultWorkingDirectory)'

.ci/yuzu-patreon-step1.yml

@@ -0,0 +1,8 @@
+trigger:
+- master
+
+stages:
+- stage: merge
+  displayName: 'merge'
+  jobs:
+  - template: ./templates/merge-private.yml

.ci/yuzu-patreon-step2.yml

@@ -0,0 +1,28 @@
+trigger:
+- master
+
+stages:
+- stage: format
+  displayName: 'format'
+  jobs:
+  - job: format
+    displayName: 'clang'
+    pool:
+      vmImage: ubuntu-latest
+    steps:
+    - template: ./templates/format-check.yml
+- stage: build
+  dependsOn: format
+  displayName: 'build'
+  jobs:
+  - template: ./templates/build-standard.yml
+    parameters:
+      cache: 'true'
+- stage: release
+  displayName: 'release'
+  dependsOn: build
+  jobs:
+  - job: azure
+    displayName: 'azure'
+    steps:
+    - template: ./templates/release-universal.yml

.ci/yuzu-patreon.yml

@@ -1,19 +0,0 @@
-# Starter pipeline
-# Start with a minimal pipeline that you can customize to build and deploy your code.
-# Add steps that build, run tests, deploy, and more:
-# https://aka.ms/yaml
-
-trigger:
-- master
-
-pool:
-  vmImage: 'ubuntu-latest'
-
-steps:
-- script: echo Hello, world!
-  displayName: 'Run a one-line script'
-
-- script: |
-    echo Add other tasks to build, test, and deploy your project.
-    echo See https://aka.ms/yaml
-  displayName: 'Run a multi-line script'

src/common/assert.h

@@ -28,18 +28,14 @@ __declspec(noinline, noreturn)
 }
 
 #define ASSERT(_a_)                                                                                \
-    do                                                                                             \
-        if (!(_a_)) {                                                                              \
-            assert_noinline_call([] { LOG_CRITICAL(Debug, "Assertion Failed!"); });                \
-        }                                                                                          \
-    while (0)
+    if (!(_a_)) {                                                                                  \
+        LOG_CRITICAL(Debug, "Assertion Failed!");                                                  \
+    }
 
 #define ASSERT_MSG(_a_, ...)                                                                       \
-    do                                                                                             \
-        if (!(_a_)) {                                                                              \
-            assert_noinline_call([&] { LOG_CRITICAL(Debug, "Assertion Failed!\n" __VA_ARGS__); }); \
-        }                                                                                          \
-    while (0)
+    if (!(_a_)) {                                                                                  \
+        LOG_CRITICAL(Debug, "Assertion Failed! " __VA_ARGS__);                                     \
+    }
 
 #define UNREACHABLE() ASSERT_MSG(false, "Unreachable code!")
 #define UNREACHABLE_MSG(...) ASSERT_MSG(false, __VA_ARGS__)

src/core/arm/dynarmic/arm_dynarmic.cpp

@@ -248,6 +248,9 @@ void ARM_Dynarmic::LoadContext(const ThreadContext& ctx) {
 }
 
 void ARM_Dynarmic::PrepareReschedule() {
+    if (jit == nullptr)
+        return;
+
     jit->HaltExecution();
 }
 

src/core/core.cpp

@@ -399,6 +399,12 @@ void System::PrepareReschedule() {
     CurrentCpuCore().PrepareReschedule();
 }
 
+void System::PrepareReschedule(const u32 core_index) {
+    if (core_index < GlobalScheduler().CpuCoresCount()) {
+        CpuCore(core_index).PrepareReschedule();
+    }
+}
+
 PerfStatsResults System::GetAndResetPerfStats() {
     return impl->GetAndResetPerfStats();
 }
@@ -439,6 +445,16 @@ const Kernel::Scheduler& System::Scheduler(std::size_t core_index) const {
     return CpuCore(core_index).Scheduler();
 }
 
+/// Gets the global scheduler
+Kernel::GlobalScheduler& System::GlobalScheduler() {
+    return impl->kernel.GlobalScheduler();
+}
+
+/// Gets the global scheduler
+const Kernel::GlobalScheduler& System::GlobalScheduler() const {
+    return impl->kernel.GlobalScheduler();
+}
+
 Kernel::Process* System::CurrentProcess() {
     return impl->kernel.CurrentProcess();
 }

src/core/core.h

@@ -25,6 +25,7 @@ class VfsFilesystem;
 } // namespace FileSys
 
 namespace Kernel {
+class GlobalScheduler;
 class KernelCore;
 class Process;
 class Scheduler;
@@ -179,6 +180,9 @@ public:
     /// Prepare the core emulation for a reschedule
     void PrepareReschedule();
 
+    /// Prepare the core emulation for a reschedule
+    void PrepareReschedule(u32 core_index);
+
     /// Gets and resets core performance statistics
     PerfStatsResults GetAndResetPerfStats();
 
@@ -233,6 +237,12 @@ public:
     /// Gets the scheduler for the CPU core with the specified index
     const Kernel::Scheduler& Scheduler(std::size_t core_index) const;
 
+    /// Gets the global scheduler
+    Kernel::GlobalScheduler& GlobalScheduler();
+
+    /// Gets the global scheduler
+    const Kernel::GlobalScheduler& GlobalScheduler() const;
+
     /// Provides a pointer to the current process
     Kernel::Process* CurrentProcess();
 

src/core/core_cpu.cpp

@@ -52,7 +52,8 @@ bool CpuBarrier::Rendezvous() {
 
 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} {
+    : cpu_barrier{cpu_barrier}, global_scheduler{system.GlobalScheduler()},
+      core_timing{system.CoreTiming()}, core_index{core_index} {
 #ifdef ARCHITECTURE_x86_64
     arm_interface = std::make_unique<ARM_Dynarmic>(system, exclusive_monitor, core_index);
 #else
@@ -60,7 +61,7 @@ Cpu::Cpu(System& system, ExclusiveMonitor& exclusive_monitor, CpuBarrier& cpu_ba
     LOG_WARNING(Core, "CPU JIT requested, but Dynarmic not available");
 #endif
 
-    scheduler = std::make_unique<Kernel::Scheduler>(system, *arm_interface);
+    scheduler = std::make_unique<Kernel::Scheduler>(system, *arm_interface, core_index);
 }
 
 Cpu::~Cpu() = default;
@@ -81,6 +82,8 @@ void Cpu::RunLoop(bool tight_loop) {
         return;
     }
 
+    Reschedule();
+
     // If we don't have a currently active thread then don't execute instructions,
     // instead advance to the next event and try to yield to the next thread
     if (Kernel::GetCurrentThread() == nullptr) {
@@ -92,7 +95,6 @@ void Cpu::RunLoop(bool tight_loop) {
             core_timing.Advance();
         }
 
-        PrepareReschedule();
     } else {
         if (IsMainCore()) {
             core_timing.Advance();
@@ -114,18 +116,14 @@ void Cpu::SingleStep() {
 
 void Cpu::PrepareReschedule() {
     arm_interface->PrepareReschedule();
-    reschedule_pending = true;
 }
 
 void Cpu::Reschedule() {
-    if (!reschedule_pending) {
-        return;
-    }
-
-    reschedule_pending = false;
     // Lock the global kernel mutex when we manipulate the HLE state
-    std::lock_guard lock{HLE::g_hle_lock};
-    scheduler->Reschedule();
+    std::lock_guard lock(HLE::g_hle_lock);
+
+    global_scheduler.SelectThread(core_index);
+    scheduler->TryDoContextSwitch();
 }
 
 } // namespace Core

src/core/core_cpu.h

@@ -12,8 +12,9 @@
 #include "common/common_types.h"
 
 namespace Kernel {
+class GlobalScheduler;
 class Scheduler;
-}
+} // namespace Kernel
 
 namespace Core {
 class System;
@@ -90,6 +91,7 @@ private:
 
     std::unique_ptr<ARM_Interface> arm_interface;
     CpuBarrier& cpu_barrier;
+    Kernel::GlobalScheduler& global_scheduler;
     std::unique_ptr<Kernel::Scheduler> scheduler;
     Timing::CoreTiming& core_timing;
 

src/core/gdbstub/gdbstub.cpp

@@ -202,13 +202,11 @@ void RegisterModule(std::string name, VAddr beg, VAddr end, bool add_elf_ext) {
 }
 
 static Kernel::Thread* FindThreadById(s64 id) {
-    for (u32 core = 0; core < Core::NUM_CPU_CORES; core++) {
-        const auto& threads = Core::System::GetInstance().Scheduler(core).GetThreadList();
-        for (auto& thread : threads) {
-            if (thread->GetThreadID() == static_cast<u64>(id)) {
-                current_core = core;
-                return thread.get();
-            }
+    const auto& threads = Core::System::GetInstance().GlobalScheduler().GetThreadList();
+    for (auto& thread : threads) {
+        if (thread->GetThreadID() == static_cast<u64>(id)) {
+            current_core = thread->GetProcessorID();
+            return thread.get();
         }
     }
     return nullptr;
@@ -646,11 +644,9 @@ static void HandleQuery() {
         SendReply(buffer.c_str());
     } else if (strncmp(query, "fThreadInfo", strlen("fThreadInfo")) == 0) {
         std::string val = "m";
-        for (u32 core = 0; core < Core::NUM_CPU_CORES; core++) {
-            const auto& threads = Core::System::GetInstance().Scheduler(core).GetThreadList();
-            for (const auto& thread : threads) {
-                val += fmt::format("{:x},", thread->GetThreadID());
-            }
+        const auto& threads = Core::System::GetInstance().GlobalScheduler().GetThreadList();
+        for (const auto& thread : threads) {
+            val += fmt::format("{:x},", thread->GetThreadID());
         }
         val.pop_back();
         SendReply(val.c_str());
@@ -660,13 +656,11 @@ static void HandleQuery() {
         std::string buffer;
         buffer += "l<?xml version=\"1.0\"?>";
         buffer += "<threads>";
-        for (u32 core = 0; core < Core::NUM_CPU_CORES; core++) {
-            const auto& threads = Core::System::GetInstance().Scheduler(core).GetThreadList();
-            for (const auto& thread : threads) {
-                buffer +=
-                    fmt::format(R"*(<thread id="{:x}" core="{:d}" name="Thread {:x}"></thread>)*",
-                                thread->GetThreadID(), core, thread->GetThreadID());
-            }
+        const auto& threads = Core::System::GetInstance().GlobalScheduler().GetThreadList();
+        for (const auto& thread : threads) {
+            buffer +=
+                fmt::format(R"*(<thread id="{:x}" core="{:d}" name="Thread {:x}"></thread>)*",
+                            thread->GetThreadID(), thread->GetProcessorID(), thread->GetThreadID());
         }
         buffer += "</threads>";
         SendReply(buffer.c_str());

src/core/hle/kernel/address_arbiter.cpp

@@ -22,6 +22,7 @@ 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) {
+    auto& system = Core::System::GetInstance();
     // Only process up to 'target' threads, unless 'target' is <= 0, in which case process
     // them all.
     std::size_t last = waiting_threads.size();
@@ -35,6 +36,7 @@ void WakeThreads(const std::vector<SharedPtr<Thread>>& waiting_threads, s32 num_
         waiting_threads[i]->SetWaitSynchronizationResult(RESULT_SUCCESS);
         waiting_threads[i]->SetArbiterWaitAddress(0);
         waiting_threads[i]->ResumeFromWait();
+        system.PrepareReschedule(waiting_threads[i]->GetProcessorID());
     }
 }
 } // Anonymous namespace
@@ -169,30 +171,22 @@ ResultCode AddressArbiter::WaitForAddressImpl(VAddr address, s64 timeout) {
 
     current_thread->WakeAfterDelay(timeout);
 
-    system.CpuCore(current_thread->GetProcessorID()).PrepareReschedule();
+    system.PrepareReschedule(current_thread->GetProcessorID());
     return RESULT_TIMEOUT;
 }
 
 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) {
-                waiting_threads.push_back(thread);
-            }
-        }
-    };
 
     // Retrieve all threads that are waiting for this address.
     std::vector<SharedPtr<Thread>> threads;
-    RetrieveWaitingThreads(0, threads, address);
-    RetrieveWaitingThreads(1, threads, address);
-    RetrieveWaitingThreads(2, threads, address);
-    RetrieveWaitingThreads(3, threads, address);
+    const auto& scheduler = system.GlobalScheduler();
+    const auto& thread_list = scheduler.GetThreadList();
+
+    for (const auto& thread : thread_list) {
+        if (thread->GetArbiterWaitAddress() == address) {
+            threads.push_back(thread);
+        }
+    }
 
     // Sort them by priority, such that the highest priority ones come first.
     std::sort(threads.begin(), threads.end(),

src/core/hle/kernel/kernel.cpp

@@ -18,6 +18,7 @@
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/process.h"
 #include "core/hle/kernel/resource_limit.h"
+#include "core/hle/kernel/scheduler.h"
 #include "core/hle/kernel/thread.h"
 #include "core/hle/lock.h"
 #include "core/hle/result.h"
@@ -88,7 +89,7 @@ static void ThreadWakeupCallback(u64 thread_handle, [[maybe_unused]] s64 cycles_
 }
 
 struct KernelCore::Impl {
-    explicit Impl(Core::System& system) : system{system} {}
+    explicit Impl(Core::System& system) : system{system}, global_scheduler{system} {}
 
     void Initialize(KernelCore& kernel) {
         Shutdown();
@@ -140,6 +141,7 @@ struct KernelCore::Impl {
     // Lists all processes that exist in the current session.
     std::vector<SharedPtr<Process>> process_list;
     Process* current_process = nullptr;
+    Kernel::GlobalScheduler global_scheduler;
 
     SharedPtr<ResourceLimit> system_resource_limit;
 
@@ -203,6 +205,14 @@ const std::vector<SharedPtr<Process>>& KernelCore::GetProcessList() const {
     return impl->process_list;
 }
 
+Kernel::GlobalScheduler& KernelCore::GlobalScheduler() {
+    return impl->global_scheduler;
+}
+
+const Kernel::GlobalScheduler& KernelCore::GlobalScheduler() const {
+    return impl->global_scheduler;
+}
+
 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

@@ -25,6 +25,7 @@ class HandleTable;
 class Process;
 class ResourceLimit;
 class Thread;
+class GlobalScheduler;
 
 /// Represents a single instance of the kernel.
 class KernelCore {
@@ -75,6 +76,12 @@ public:
     /// Retrieves the list of processes.
     const std::vector<SharedPtr<Process>>& GetProcessList() const;
 
+    /// Gets the sole instance of the global scheduler
+    Kernel::GlobalScheduler& GlobalScheduler();
+
+    /// Gets the sole instance of the global scheduler
+    const Kernel::GlobalScheduler& GlobalScheduler() const;
+
     /// Adds a port to the named port table
     void AddNamedPort(std::string name, SharedPtr<ClientPort> port);
 

src/core/hle/kernel/mutex.cpp

@@ -7,6 +7,7 @@
 
 #include "common/assert.h"
 #include "core/core.h"
+#include "core/core_cpu.h"
 #include "core/hle/kernel/errors.h"
 #include "core/hle/kernel/handle_table.h"
 #include "core/hle/kernel/kernel.h"
@@ -78,7 +79,7 @@ ResultCode Mutex::TryAcquire(VAddr address, Handle holding_thread_handle,
     // thread.
     ASSERT(requesting_thread == current_thread);
 
-    const u32 addr_value = Memory::Read32(address);
+    u32 addr_value = Memory::Read32(address);
 
     // If the mutex isn't being held, just return success.
     if (addr_value != (holding_thread_handle | Mutex::MutexHasWaitersFlag)) {
@@ -89,6 +90,20 @@ ResultCode Mutex::TryAcquire(VAddr address, Handle holding_thread_handle,
         return ERR_INVALID_HANDLE;
     }
 
+    // This a workaround where an unknown bug writes the mutex value to give ownership to a cond var
+    // waiting thread.
+    if (holding_thread->GetStatus() == ThreadStatus::WaitCondVar) {
+        if (holding_thread->GetMutexWaitAddress() == address) {
+            Release(address, holding_thread.get());
+            addr_value = Memory::Read32(address);
+            if (addr_value == 0)
+                return RESULT_SUCCESS;
+            else {
+                holding_thread = handle_table.Get<Thread>(addr_value & Mutex::MutexOwnerMask);
+            }
+        }
+    }
+
     // Wait until the mutex is released
     current_thread->SetMutexWaitAddress(address);
     current_thread->SetWaitHandle(requesting_thread_handle);
@@ -104,14 +119,13 @@ ResultCode Mutex::TryAcquire(VAddr address, Handle holding_thread_handle,
     return RESULT_SUCCESS;
 }
 
-ResultCode Mutex::Release(VAddr address) {
+ResultCode Mutex::Release(VAddr address, Thread* holding_thread) {
     // The mutex address must be 4-byte aligned
     if ((address % sizeof(u32)) != 0) {
         return ERR_INVALID_ADDRESS;
     }
 
-    auto* const current_thread = system.CurrentScheduler().GetCurrentThread();
-    auto [thread, num_waiters] = GetHighestPriorityMutexWaitingThread(current_thread, address);
+    auto [thread, num_waiters] = GetHighestPriorityMutexWaitingThread(holding_thread, address);
 
     // There are no more threads waiting for the mutex, release it completely.
     if (thread == nullptr) {
@@ -120,7 +134,7 @@ ResultCode Mutex::Release(VAddr address) {
     }
 
     // Transfer the ownership of the mutex from the previous owner to the new one.
-    TransferMutexOwnership(address, current_thread, thread);
+    TransferMutexOwnership(address, holding_thread, thread);
 
     u32 mutex_value = thread->GetWaitHandle();
 
@@ -140,6 +154,11 @@ ResultCode Mutex::Release(VAddr address) {
     thread->SetMutexWaitAddress(0);
     thread->SetWaitHandle(0);
 
+    if (thread->GetProcessorID() >= 0)
+        system.CpuCore(thread->GetProcessorID()).PrepareReschedule();
+    if (holding_thread->GetProcessorID() >= 0)
+        system.CpuCore(holding_thread->GetProcessorID()).PrepareReschedule();
+
     return RESULT_SUCCESS;
 }
 } // namespace Kernel

src/core/hle/kernel/mutex.h

@@ -29,7 +29,7 @@ public:
                           Handle requesting_thread_handle);
 
     /// Releases the mutex at the specified address.
-    ResultCode Release(VAddr address);
+    ResultCode Release(VAddr address, Thread* holding_thread);
 
 private:
     Core::System& system;

src/core/hle/kernel/process.cpp

@@ -213,10 +213,7 @@ void Process::PrepareForTermination() {
         }
     };
 
-    stop_threads(system.Scheduler(0).GetThreadList());
-    stop_threads(system.Scheduler(1).GetThreadList());
-    stop_threads(system.Scheduler(2).GetThreadList());
-    stop_threads(system.Scheduler(3).GetThreadList());
+    stop_threads(system.GlobalScheduler().GetThreadList());
 
     FreeTLSRegion(tls_region_address);
     tls_region_address = 0;

src/core/hle/kernel/scheduler.cpp

@@ -1,8 +1,13 @@
 // Copyright 2018 yuzu emulator team
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
+//
+// SelectThreads, Yield functions originally by TuxSH.
+// licensed under GPLv2 or later under exception provided by the author.
 
 #include <algorithm>
+#include <set>
+#include <unordered_set>
 #include <utility>
 
 #include "common/assert.h"
@@ -17,56 +22,277 @@
 
 namespace Kernel {
 
-std::mutex Scheduler::scheduler_mutex;
+GlobalScheduler::GlobalScheduler(Core::System& system) : system{system} {
+    reselection_pending = false;
+}
 
-Scheduler::Scheduler(Core::System& system, Core::ARM_Interface& cpu_core)
-    : cpu_core{cpu_core}, system{system} {}
+void GlobalScheduler::AddThread(SharedPtr<Thread> thread) {
+    thread_list.push_back(std::move(thread));
+}
 
-Scheduler::~Scheduler() {
-    for (auto& thread : thread_list) {
-        thread->Stop();
+void GlobalScheduler::RemoveThread(const Thread* thread) {
+    thread_list.erase(std::remove(thread_list.begin(), thread_list.end(), thread),
+                      thread_list.end());
+}
+
+/*
+ * UnloadThread selects a core and forces it to unload its current thread's context
+ */
+void GlobalScheduler::UnloadThread(s32 core) {
+    Scheduler& sched = system.Scheduler(core);
+    sched.UnloadThread();
+}
+
+/*
+ * SelectThread takes care of selecting the new scheduled thread.
+ * It does it in 3 steps:
+ * - First a thread is selected from the top of the priority queue. If no thread
+ * is obtained then we move to step two, else we are done.
+ * - Second we try to get a suggested thread that's not assigned to any core or
+ * that is not the top thread in that core.
+ * - Third is no suggested thread is found, we do a second pass and pick a running
+ * thread in another core and swap it with its current thread.
+ */
+void GlobalScheduler::SelectThread(u32 core) {
+    const auto update_thread = [](Thread* thread, Scheduler& sched) {
+        if (thread != sched.selected_thread) {
+            if (thread == nullptr) {
+                ++sched.idle_selection_count;
+            }
+            sched.selected_thread = thread;
+        }
+        sched.context_switch_pending = sched.selected_thread != sched.current_thread;
+        std::atomic_thread_fence(std::memory_order_seq_cst);
+    };
+    Scheduler& sched = system.Scheduler(core);
+    Thread* current_thread = nullptr;
+    // Step 1: Get top thread in schedule queue.
+    current_thread = scheduled_queue[core].empty() ? nullptr : scheduled_queue[core].front();
+    if (current_thread) {
+        update_thread(current_thread, sched);
+        return;
+    }
+    // Step 2: Try selecting a suggested thread.
+    Thread* winner = nullptr;
+    std::set<s32> sug_cores;
+    for (auto thread : suggested_queue[core]) {
+        s32 this_core = thread->GetProcessorID();
+        Thread* thread_on_core = nullptr;
+        if (this_core >= 0) {
+            thread_on_core = scheduled_queue[this_core].front();
+        }
+        if (this_core < 0 || thread != thread_on_core) {
+            winner = thread;
+            break;
+        }
+        sug_cores.insert(this_core);
+    }
+    // if we got a suggested thread, select it, else do a second pass.
+    if (winner && winner->GetPriority() > 2) {
+        if (winner->IsRunning()) {
+            UnloadThread(winner->GetProcessorID());
+        }
+        TransferToCore(winner->GetPriority(), core, winner);
+        update_thread(winner, sched);
+        return;
+    }
+    // Step 3: Select a suggested thread from another core
+    for (auto& src_core : sug_cores) {
+        auto it = scheduled_queue[src_core].begin();
+        it++;
+        if (it != scheduled_queue[src_core].end()) {
+            Thread* thread_on_core = scheduled_queue[src_core].front();
+            Thread* to_change = *it;
+            if (thread_on_core->IsRunning() || to_change->IsRunning()) {
+                UnloadThread(src_core);
+            }
+            TransferToCore(thread_on_core->GetPriority(), core, thread_on_core);
+            current_thread = thread_on_core;
+            break;
+        }
+    }
+    update_thread(current_thread, sched);
+}
+
+/*
+ * YieldThread takes a thread and moves it to the back of the it's priority list
+ * This operation can be redundant and no scheduling is changed if marked as so.
+ */
+void GlobalScheduler::YieldThread(Thread* yielding_thread) {
+    // Note: caller should use critical section, etc.
+    const u32 core_id = static_cast<u32>(yielding_thread->GetProcessorID());
+    const u32 priority = yielding_thread->GetPriority();
+
+    // Yield the thread
+    ASSERT_MSG(yielding_thread == scheduled_queue[core_id].front(priority),
+               "Thread yielding without being in front");
+    scheduled_queue[core_id].yield(priority);
+
+    Thread* winner = scheduled_queue[core_id].front(priority);
+    AskForReselectionOrMarkRedundant(yielding_thread, winner);
+}
+
+/*
+ * YieldThreadAndBalanceLoad takes a thread and moves it to the back of the it's priority list.
+ * Afterwards, tries to pick a suggested thread from the suggested queue that has worse time or
+ * a better priority than the next thread in the core.
+ * This operation can be redundant and no scheduling is changed if marked as so.
+ */
+void GlobalScheduler::YieldThreadAndBalanceLoad(Thread* yielding_thread) {
+    // Note: caller should check if !thread.IsSchedulerOperationRedundant and use critical section,
+    // etc.
+    const u32 core_id = static_cast<u32>(yielding_thread->GetProcessorID());
+    const u32 priority = yielding_thread->GetPriority();
+
+    // Yield the thread
+    ASSERT_MSG(yielding_thread == scheduled_queue[core_id].front(priority),
+               "Thread yielding without being in front");
+    scheduled_queue[core_id].yield(priority);
+
+    std::array<Thread*, NUM_CPU_CORES> current_threads;
+    for (u32 i = 0; i < NUM_CPU_CORES; i++) {
+        current_threads[i] = scheduled_queue[i].empty() ? nullptr : scheduled_queue[i].front();
+    }
+
+    Thread* next_thread = scheduled_queue[core_id].front(priority);
+    Thread* winner = nullptr;
+    for (auto& thread : suggested_queue[core_id]) {
+        const s32 source_core = thread->GetProcessorID();
+        if (source_core >= 0) {
+            if (current_threads[source_core] != nullptr) {
+                if (thread == current_threads[source_core] ||
+                    current_threads[source_core]->GetPriority() < min_regular_priority) {
+                    continue;
+                }
+            }
+            if (next_thread->GetLastRunningTicks() >= thread->GetLastRunningTicks() ||
+                next_thread->GetPriority() < thread->GetPriority()) {
+                if (thread->GetPriority() <= priority) {
+                    winner = thread;
+                    break;
+                }
+            }
+        }
+    }
+
+    if (winner != nullptr) {
+        if (winner != yielding_thread) {
+            if (winner->IsRunning()) {
+                UnloadThread(winner->GetProcessorID());
+            }
+            TransferToCore(winner->GetPriority(), core_id, winner);
+        }
+    } else {
+        winner = next_thread;
+    }
+
+    AskForReselectionOrMarkRedundant(yielding_thread, winner);
+}
+
+/*
+ * YieldThreadAndWaitForLoadBalancing takes a thread and moves it out of the scheduling queue
+ * and into the suggested queue. If no thread can be squeduled afterwards in that core,
+ * a suggested thread is obtained instead.
+ * This operation can be redundant and no scheduling is changed if marked as so.
+ */
+void GlobalScheduler::YieldThreadAndWaitForLoadBalancing(Thread* yielding_thread) {
+    // Note: caller should check if !thread.IsSchedulerOperationRedundant and use critical section,
+    // etc.
+    Thread* winner = nullptr;
+    const u32 core_id = static_cast<u32>(yielding_thread->GetProcessorID());
+
+    // Remove the thread from its scheduled mlq, put it on the corresponding "suggested" one instead
+    TransferToCore(yielding_thread->GetPriority(), -1, yielding_thread);
+
+    // If the core is idle, perform load balancing, excluding the threads that have just used this
+    // function...
+    if (scheduled_queue[core_id].empty()) {
+        // Here, "current_threads" is calculated after the ""yield"", unlike yield -1
+        std::array<Thread*, NUM_CPU_CORES> current_threads;
+        for (u32 i = 0; i < NUM_CPU_CORES; i++) {
+            current_threads[i] = scheduled_queue[i].empty() ? nullptr : scheduled_queue[i].front();
+        }
+        for (auto& thread : suggested_queue[core_id]) {
+            const s32 source_core = thread->GetProcessorID();
+            if (source_core < 0 || thread == current_threads[source_core]) {
+                continue;
+            }
+            if (current_threads[source_core] == nullptr ||
+                current_threads[source_core]->GetPriority() >= min_regular_priority) {
+                winner = thread;
+            }
+            break;
+        }
+        if (winner != nullptr) {
+            if (winner != yielding_thread) {
+                if (winner->IsRunning()) {
+                    UnloadThread(winner->GetProcessorID());
+                }
+                TransferToCore(winner->GetPriority(), core_id, winner);
+            }
+        } else {
+            winner = yielding_thread;
+        }
+    }
+
+    AskForReselectionOrMarkRedundant(yielding_thread, winner);
+}
+
+void GlobalScheduler::Schedule(u32 priority, u32 core, Thread* thread) {
+    ASSERT_MSG(thread->GetProcessorID() == core, "Thread must be assigned to this core.");
+    scheduled_queue[core].add(thread, priority);
+}
+
+void GlobalScheduler::SchedulePrepend(u32 priority, u32 core, Thread* thread) {
+    ASSERT_MSG(thread->GetProcessorID() == core, "Thread must be assigned to this core.");
+    scheduled_queue[core].add(thread, priority, false);
+}
+
+void GlobalScheduler::AskForReselectionOrMarkRedundant(Thread* current_thread, Thread* winner) {
+    if (current_thread == winner) {
+        // TODO(blinkhawk): manage redundant operations, this is not implemented.
+        // as its mostly an optimization.
+        // current_thread->SetRedundantSchedulerOperation();
+    } else {
+        reselection_pending.store(true, std::memory_order_release);
     }
 }
 
+GlobalScheduler::~GlobalScheduler() = default;
+
+Scheduler::Scheduler(Core::System& system, Core::ARM_Interface& cpu_core, u32 core_id)
+    : system(system), cpu_core(cpu_core), core_id(core_id) {}
+
+Scheduler::~Scheduler() = default;
+
 bool Scheduler::HaveReadyThreads() const {
-    std::lock_guard lock{scheduler_mutex};
-    return !ready_queue.empty();
+    return system.GlobalScheduler().HaveReadyThreads(core_id);
 }
 
 Thread* Scheduler::GetCurrentThread() const {
     return current_thread.get();
 }
 
+Thread* Scheduler::GetSelectedThread() const {
+    return selected_thread.get();
+}
+
+void Scheduler::SelectThreads() {
+    system.GlobalScheduler().SelectThread(core_id);
+}
+
 u64 Scheduler::GetLastContextSwitchTicks() const {
     return last_context_switch_time;
 }
 
-Thread* Scheduler::PopNextReadyThread() {
-    Thread* next = nullptr;
-    Thread* thread = GetCurrentThread();
-
-    if (thread && thread->GetStatus() == ThreadStatus::Running) {
-        if (ready_queue.empty()) {
-            return thread;
-        }
-        // We have to do better than the current thread.
-        // This call returns null when that's not possible.
-        next = ready_queue.front();
-        if (next == nullptr || next->GetPriority() >= thread->GetPriority()) {
-            next = thread;
-        }
-    } else {
-        if (ready_queue.empty()) {
-            return nullptr;
-        }
-        next = ready_queue.front();
+void Scheduler::TryDoContextSwitch() {
+    if (context_switch_pending) {
+        SwitchContext();
     }
-
-    return next;
 }
 
-void Scheduler::SwitchContext(Thread* new_thread) {
-    Thread* previous_thread = GetCurrentThread();
+void Scheduler::UnloadThread() {
+    Thread* const previous_thread = GetCurrentThread();
     Process* const previous_process = system.Kernel().CurrentProcess();
 
     UpdateLastContextSwitchTime(previous_thread, previous_process);
@@ -80,23 +306,52 @@ void Scheduler::SwitchContext(Thread* new_thread) {
         if (previous_thread->GetStatus() == ThreadStatus::Running) {
             // This is only the case when a reschedule is triggered without the current thread
             // yielding execution (i.e. an event triggered, system core time-sliced, etc)
-            ready_queue.add(previous_thread, previous_thread->GetPriority(), false);
             previous_thread->SetStatus(ThreadStatus::Ready);
         }
+        previous_thread->SetIsRunning(false);
+    }
+    current_thread = nullptr;
+}
+
+void Scheduler::SwitchContext() {
+    Thread* const previous_thread = GetCurrentThread();
+    Thread* const new_thread = GetSelectedThread();
+
+    context_switch_pending = false;
+    if (new_thread == previous_thread) {
+        return;
+    }
+
+    Process* const previous_process = system.Kernel().CurrentProcess();
+
+    UpdateLastContextSwitchTime(previous_thread, previous_process);
+
+    // Save context for previous thread
+    if (previous_thread) {
+        cpu_core.SaveContext(previous_thread->GetContext());
+        // Save the TPIDR_EL0 system register in case it was modified.
+        previous_thread->SetTPIDR_EL0(cpu_core.GetTPIDR_EL0());
+
+        if (previous_thread->GetStatus() == ThreadStatus::Running) {
+            // This is only the case when a reschedule is triggered without the current thread
+            // yielding execution (i.e. an event triggered, system core time-sliced, etc)
+            previous_thread->SetStatus(ThreadStatus::Ready);
+        }
+        previous_thread->SetIsRunning(false);
     }
 
     // Load context of new thread
     if (new_thread) {
+        ASSERT_MSG(new_thread->GetProcessorID() == this->core_id,
+                   "Thread must be assigned to this core.");
         ASSERT_MSG(new_thread->GetStatus() == ThreadStatus::Ready,
                    "Thread must be ready to become running.");
 
         // Cancel any outstanding wakeup events for this thread
         new_thread->CancelWakeupTimer();
-
         current_thread = new_thread;
-
-        ready_queue.remove(new_thread, new_thread->GetPriority());
         new_thread->SetStatus(ThreadStatus::Running);
+        new_thread->SetIsRunning(true);
 
         auto* const thread_owner_process = current_thread->GetOwnerProcess();
         if (previous_process != thread_owner_process) {
@@ -130,124 +385,4 @@ void Scheduler::UpdateLastContextSwitchTime(Thread* thread, Process* process) {
     last_context_switch_time = most_recent_switch_ticks;
 }
 
-void Scheduler::Reschedule() {
-    std::lock_guard lock{scheduler_mutex};
-
-    Thread* cur = GetCurrentThread();
-    Thread* next = PopNextReadyThread();
-
-    if (cur && next) {
-        LOG_TRACE(Kernel, "context switch {} -> {}", cur->GetObjectId(), next->GetObjectId());
-    } else if (cur) {
-        LOG_TRACE(Kernel, "context switch {} -> idle", cur->GetObjectId());
-    } else if (next) {
-        LOG_TRACE(Kernel, "context switch idle -> {}", next->GetObjectId());
-    }
-
-    SwitchContext(next);
-}
-
-void Scheduler::AddThread(SharedPtr<Thread> thread) {
-    std::lock_guard lock{scheduler_mutex};
-
-    thread_list.push_back(std::move(thread));
-}
-
-void Scheduler::RemoveThread(Thread* thread) {
-    std::lock_guard lock{scheduler_mutex};
-
-    thread_list.erase(std::remove(thread_list.begin(), thread_list.end(), thread),
-                      thread_list.end());
-}
-
-void Scheduler::ScheduleThread(Thread* thread, u32 priority) {
-    std::lock_guard lock{scheduler_mutex};
-
-    ASSERT(thread->GetStatus() == ThreadStatus::Ready);
-    ready_queue.add(thread, priority);
-}
-
-void Scheduler::UnscheduleThread(Thread* thread, u32 priority) {
-    std::lock_guard lock{scheduler_mutex};
-
-    ASSERT(thread->GetStatus() == ThreadStatus::Ready);
-    ready_queue.remove(thread, priority);
-}
-
-void Scheduler::SetThreadPriority(Thread* thread, u32 priority) {
-    std::lock_guard lock{scheduler_mutex};
-    if (thread->GetPriority() == priority) {
-        return;
-    }
-
-    // If thread was ready, adjust queues
-    if (thread->GetStatus() == ThreadStatus::Ready)
-        ready_queue.adjust(thread, thread->GetPriority(), priority);
-}
-
-Thread* Scheduler::GetNextSuggestedThread(u32 core, u32 maximum_priority) const {
-    std::lock_guard lock{scheduler_mutex};
-
-    const u32 mask = 1U << core;
-    for (auto* thread : ready_queue) {
-        if ((thread->GetAffinityMask() & mask) != 0 && thread->GetPriority() < maximum_priority) {
-            return thread;
-        }
-    }
-    return nullptr;
-}
-
-void Scheduler::YieldWithoutLoadBalancing(Thread* thread) {
-    ASSERT(thread != nullptr);
-    // Avoid yielding if the thread isn't even running.
-    ASSERT(thread->GetStatus() == ThreadStatus::Running);
-
-    // Sanity check that the priority is valid
-    ASSERT(thread->GetPriority() < THREADPRIO_COUNT);
-
-    // Yield this thread -- sleep for zero time and force reschedule to different thread
-    GetCurrentThread()->Sleep(0);
-}
-
-void Scheduler::YieldWithLoadBalancing(Thread* thread) {
-    ASSERT(thread != nullptr);
-    const auto priority = thread->GetPriority();
-    const auto core = static_cast<u32>(thread->GetProcessorID());
-
-    // Avoid yielding if the thread isn't even running.
-    ASSERT(thread->GetStatus() == ThreadStatus::Running);
-
-    // Sanity check that the priority is valid
-    ASSERT(priority < THREADPRIO_COUNT);
-
-    // Sleep for zero time to be able to force reschedule to different thread
-    GetCurrentThread()->Sleep(0);
-
-    Thread* suggested_thread = nullptr;
-
-    // Search through all of the cpu cores (except this one) for a suggested thread.
-    // Take the first non-nullptr one
-    for (unsigned cur_core = 0; cur_core < Core::NUM_CPU_CORES; ++cur_core) {
-        const auto res =
-            system.CpuCore(cur_core).Scheduler().GetNextSuggestedThread(core, priority);
-
-        // If scheduler provides a suggested thread
-        if (res != nullptr) {
-            // And its better than the current suggested thread (or is the first valid one)
-            if (suggested_thread == nullptr ||
-                suggested_thread->GetPriority() > res->GetPriority()) {
-                suggested_thread = res;
-            }
-        }
-    }
-
-    // If a suggested thread was found, queue that for this core
-    if (suggested_thread != nullptr)
-        suggested_thread->ChangeCore(core, suggested_thread->GetAffinityMask());
-}
-
-void Scheduler::YieldAndWaitForLoadBalancing(Thread* thread) {
-    UNIMPLEMENTED_MSG("Wait for load balancing thread yield type is not implemented!");
-}
-
 } // namespace Kernel

src/core/hle/kernel/scheduler.h

@@ -20,124 +20,178 @@ namespace Kernel {
 
 class Process;
 
-class Scheduler final {
+class GlobalScheduler final {
 public:
-    explicit Scheduler(Core::System& system, Core::ARM_Interface& cpu_core);
-    ~Scheduler();
-
-    /// Returns whether there are any threads that are ready to run.
-    bool HaveReadyThreads() const;
-
-    /// Reschedules to the next available thread (call after current thread is suspended)
-    void Reschedule();
-
-    /// Gets the current running thread
-    Thread* GetCurrentThread() const;
-
-    /// Gets the timestamp for the last context switch in ticks.
-    u64 GetLastContextSwitchTicks() const;
+    static constexpr u32 NUM_CPU_CORES = 4;
 
+    explicit GlobalScheduler(Core::System& system);
+    ~GlobalScheduler();
     /// Adds a new thread to the scheduler
     void AddThread(SharedPtr<Thread> thread);
 
     /// Removes a thread from the scheduler
-    void RemoveThread(Thread* thread);
-
-    /// Schedules a thread that has become "ready"
-    void ScheduleThread(Thread* thread, u32 priority);
-
-    /// Unschedules a thread that was already scheduled
-    void UnscheduleThread(Thread* thread, u32 priority);
-
-    /// Sets the priority of a thread in the scheduler
-    void SetThreadPriority(Thread* thread, u32 priority);
-
-    /// Gets the next suggested thread for load balancing
-    Thread* GetNextSuggestedThread(u32 core, u32 minimum_priority) const;
-
-    /**
-     * YieldWithoutLoadBalancing -- analogous to normal yield on a system
-     * Moves the thread to the end of the ready queue for its priority, and then reschedules the
-     * system to the new head of the queue.
-     *
-     * Example (Single Core -- but can be extrapolated to multi):
-     * ready_queue[prio=0]: ThreadA, ThreadB, ThreadC (->exec order->)
-     * Currently Running: ThreadR
-     *
-     * ThreadR calls YieldWithoutLoadBalancing
-     *
-     * ThreadR is moved to the end of ready_queue[prio=0]:
-     * ready_queue[prio=0]: ThreadA, ThreadB, ThreadC, ThreadR (->exec order->)
-     * Currently Running: Nothing
-     *
-     * System is rescheduled (ThreadA is popped off of queue):
-     * ready_queue[prio=0]: ThreadB, ThreadC, ThreadR (->exec order->)
-     * Currently Running: ThreadA
-     *
-     * If the queue is empty at time of call, no yielding occurs. This does not cross between cores
-     * or priorities at all.
-     */
-    void YieldWithoutLoadBalancing(Thread* thread);
-
-    /**
-     * YieldWithLoadBalancing -- yield but with better selection of the new running thread
-     * Moves the current thread to the end of the ready queue for its priority, then selects a
-     * 'suggested thread' (a thread on a different core that could run on this core) from the
-     * scheduler, changes its core, and reschedules the current core to that thread.
-     *
-     * Example (Dual Core -- can be extrapolated to Quad Core, this is just normal yield if it were
-     * single core):
-     * ready_queue[core=0][prio=0]: ThreadA, ThreadB (affinities not pictured as irrelevant
-     * ready_queue[core=1][prio=0]: ThreadC[affinity=both], ThreadD[affinity=core1only]
-     * Currently Running: ThreadQ on Core 0 || ThreadP on Core 1
-     *
-     * ThreadQ calls YieldWithLoadBalancing
-     *
-     * ThreadQ is moved to the end of ready_queue[core=0][prio=0]:
-     * ready_queue[core=0][prio=0]: ThreadA, ThreadB
-     * ready_queue[core=1][prio=0]: ThreadC[affinity=both], ThreadD[affinity=core1only]
-     * Currently Running: ThreadQ on Core 0 || ThreadP on Core 1
-     *
-     * A list of suggested threads for each core is compiled
-     * Suggested Threads: {ThreadC on Core 1}
-     * If this were quad core (as the switch is), there could be between 0 and 3 threads in this
-     * list. If there are more than one, the thread is selected by highest prio.
-     *
-     * ThreadC is core changed to Core 0:
-     * ready_queue[core=0][prio=0]: ThreadC, ThreadA, ThreadB, ThreadQ
-     * ready_queue[core=1][prio=0]: ThreadD
-     * Currently Running: None on Core 0 || ThreadP on Core 1
-     *
-     * System is rescheduled (ThreadC is popped off of queue):
-     * ready_queue[core=0][prio=0]: ThreadA, ThreadB, ThreadQ
-     * ready_queue[core=1][prio=0]: ThreadD
-     * Currently Running: ThreadC on Core 0 || ThreadP on Core 1
-     *
-     * If no suggested threads can be found this will behave just as normal yield. If there are
-     * multiple candidates for the suggested thread on a core, the highest prio is taken.
-     */
-    void YieldWithLoadBalancing(Thread* thread);
-
-    /// Currently unknown -- asserts as unimplemented on call
-    void YieldAndWaitForLoadBalancing(Thread* thread);
+    void RemoveThread(const Thread* thread);
 
     /// Returns a list of all threads managed by the scheduler
     const std::vector<SharedPtr<Thread>>& GetThreadList() const {
         return thread_list;
     }
 
-private:
-    /**
-     * Pops and returns the next thread from the thread queue
-     * @return A pointer to the next ready thread
-     */
-    Thread* PopNextReadyThread();
+    // Add a thread to the suggested queue of a cpu core. Suggested threads may be
+    // picked if no thread is scheduled to run on the core.
+    void Suggest(u32 priority, u32 core, Thread* thread) {
+        suggested_queue[core].add(thread, priority);
+    }
 
+    // Remove a thread to the suggested queue of a cpu core. Suggested threads may be
+    // picked if no thread is scheduled to run on the core.
+    void Unsuggest(u32 priority, u32 core, Thread* thread) {
+        suggested_queue[core].remove(thread, priority);
+    }
+
+    // Add a thread to the scheduling queue of a cpu core. The thread is added at the
+    // back the queue in its priority level
+    void Schedule(u32 priority, u32 core, Thread* thread);
+
+    // Add a thread to the scheduling queue of a cpu core. The thread is added at the
+    // front the queue in its priority level
+    void SchedulePrepend(u32 priority, u32 core, Thread* thread);
+
+    // Reschedule an already scheduled thread based on a new priority
+    void Reschedule(u32 priority, u32 core, Thread* thread) {
+        scheduled_queue[core].remove(thread, priority);
+        scheduled_queue[core].add(thread, priority);
+    }
+
+    // Unschedule a thread.
+    void Unschedule(u32 priority, u32 core, Thread* thread) {
+        scheduled_queue[core].remove(thread, priority);
+    }
+
+    // Transfers a thread into an specific core. If the destination_core is -1
+    // it will be unscheduled from its source code and added into its suggested
+    // queue.
+    void TransferToCore(u32 priority, s32 destination_core, Thread* thread) {
+        const bool schedulable = thread->GetPriority() < THREADPRIO_COUNT;
+        const s32 source_core = thread->GetProcessorID();
+        if (source_core == destination_core || !schedulable) {
+            return;
+        }
+        thread->SetProcessorID(destination_core);
+        if (source_core >= 0) {
+            Unschedule(priority, source_core, thread);
+        }
+        if (destination_core >= 0) {
+            Unsuggest(priority, destination_core, thread);
+            Schedule(priority, destination_core, thread);
+        }
+        if (source_core >= 0) {
+            Suggest(priority, source_core, thread);
+        }
+    }
+
+    /*
+     * UnloadThread selects a core and forces it to unload its current thread's context
+     */
+    void UnloadThread(s32 core);
+
+    /*
+     * SelectThread takes care of selecting the new scheduled thread.
+     * It does it in 3 steps:
+     * - First a thread is selected from the top of the priority queue. If no thread
+     * is obtained then we move to step two, else we are done.
+     * - Second we try to get a suggested thread that's not assigned to any core or
+     * that is not the top thread in that core.
+     * - Third is no suggested thread is found, we do a second pass and pick a running
+     * thread in another core and swap it with its current thread.
+     */
+    void SelectThread(u32 core);
+
+    bool HaveReadyThreads(u32 core_id) const {
+        return !scheduled_queue[core_id].empty();
+    }
+
+    /*
+     * YieldThread takes a thread and moves it to the back of the it's priority list
+     * This operation can be redundant and no scheduling is changed if marked as so.
+     */
+    void YieldThread(Thread* thread);
+
+    /*
+     * YieldThreadAndBalanceLoad takes a thread and moves it to the back of the it's priority list.
+     * Afterwards, tries to pick a suggested thread from the suggested queue that has worse time or
+     * a better priority than the next thread in the core.
+     * This operation can be redundant and no scheduling is changed if marked as so.
+     */
+    void YieldThreadAndBalanceLoad(Thread* thread);
+
+    /*
+     * YieldThreadAndWaitForLoadBalancing takes a thread and moves it out of the scheduling queue
+     * and into the suggested queue. If no thread can be squeduled afterwards in that core,
+     * a suggested thread is obtained instead.
+     * This operation can be redundant and no scheduling is changed if marked as so.
+     */
+    void YieldThreadAndWaitForLoadBalancing(Thread* thread);
+
+    u32 CpuCoresCount() const {
+        return NUM_CPU_CORES;
+    }
+
+    void SetReselectionPending() {
+        reselection_pending.store(true, std::memory_order_release);
+    }
+
+    bool IsReselectionPending() const {
+        return reselection_pending.load();
+    }
+
+private:
+    void AskForReselectionOrMarkRedundant(Thread* current_thread, Thread* winner);
+
+    static constexpr u32 min_regular_priority = 2;
+    std::array<Common::MultiLevelQueue<Thread*, THREADPRIO_COUNT>, NUM_CPU_CORES> scheduled_queue;
+    std::array<Common::MultiLevelQueue<Thread*, THREADPRIO_COUNT>, NUM_CPU_CORES> suggested_queue;
+    std::atomic<bool> reselection_pending;
+
+    /// Lists all thread ids that aren't deleted/etc.
+    std::vector<SharedPtr<Thread>> thread_list;
+    Core::System& system;
+};
+
+class Scheduler final {
+public:
+    explicit Scheduler(Core::System& system, Core::ARM_Interface& cpu_core, const u32 core_id);
+    ~Scheduler();
+
+    /// Returns whether there are any threads that are ready to run.
+    bool HaveReadyThreads() const;
+
+    /// Reschedules to the next available thread (call after current thread is suspended)
+    void TryDoContextSwitch();
+
+    void UnloadThread();
+
+    void SelectThreads();
+
+    /// Gets the current running thread
+    Thread* GetCurrentThread() const;
+
+    Thread* GetSelectedThread() const;
+
+    /// Gets the timestamp for the last context switch in ticks.
+    u64 GetLastContextSwitchTicks() const;
+
+    bool ContextSwitchPending() const {
+        return context_switch_pending;
+    }
+
+private:
+    friend class GlobalScheduler;
     /**
      * Switches the CPU's active thread context to that of the specified thread
      * @param new_thread The thread to switch to
      */
-    void SwitchContext(Thread* new_thread);
+    void SwitchContext();
 
     /**
      * Called on every context switch to update the internal timestamp
@@ -152,19 +206,16 @@ private:
      */
     void UpdateLastContextSwitchTime(Thread* thread, Process* process);
 
-    /// Lists all thread ids that aren't deleted/etc.
-    std::vector<SharedPtr<Thread>> thread_list;
-
-    /// Lists only ready thread ids.
-    Common::MultiLevelQueue<Thread*, THREADPRIO_LOWEST + 1> ready_queue;
-
     SharedPtr<Thread> current_thread = nullptr;
-
-    Core::ARM_Interface& cpu_core;
-    u64 last_context_switch_time = 0;
+    SharedPtr<Thread> selected_thread = nullptr;
 
     Core::System& system;
-    static std::mutex scheduler_mutex;
+    Core::ARM_Interface& cpu_core;
+    u64 last_context_switch_time = 0;
+    u64 idle_selection_count = 0;
+    const u32 core_id;
+
+    bool context_switch_pending = false;
 };
 
 } // namespace Kernel

src/core/hle/kernel/svc.cpp

@@ -516,7 +516,7 @@ static ResultCode WaitSynchronization(Core::System& system, Handle* index, VAddr
     thread->WakeAfterDelay(nano_seconds);
     thread->SetWakeupCallback(DefaultThreadWakeupCallback);
 
-    system.CpuCore(thread->GetProcessorID()).PrepareReschedule();
+    system.PrepareReschedule(thread->GetProcessorID());
 
     return RESULT_TIMEOUT;
 }
@@ -534,6 +534,7 @@ static ResultCode CancelSynchronization(Core::System& system, Handle thread_hand
     }
 
     thread->CancelWait();
+    system.PrepareReschedule(thread->GetProcessorID());
     return RESULT_SUCCESS;
 }
 
@@ -577,7 +578,8 @@ static ResultCode ArbitrateUnlock(Core::System& system, VAddr mutex_addr) {
     }
 
     auto* const current_process = system.Kernel().CurrentProcess();
-    return current_process->GetMutex().Release(mutex_addr);
+    return current_process->GetMutex().Release(mutex_addr,
+                                               system.CurrentScheduler().GetCurrentThread());
 }
 
 enum class BreakType : u32 {
@@ -1066,6 +1068,8 @@ static ResultCode SetThreadActivity(Core::System& system, Handle handle, u32 act
     }
 
     thread->SetActivity(static_cast<ThreadActivity>(activity));
+
+    system.PrepareReschedule(thread->GetProcessorID());
     return RESULT_SUCCESS;
 }
 
@@ -1147,7 +1151,7 @@ static ResultCode SetThreadPriority(Core::System& system, Handle handle, u32 pri
 
     thread->SetPriority(priority);
 
-    system.CpuCore(thread->GetProcessorID()).PrepareReschedule();
+    system.PrepareReschedule(thread->GetProcessorID());
     return RESULT_SUCCESS;
 }
 
@@ -1503,7 +1507,7 @@ static ResultCode CreateThread(Core::System& system, Handle* out_handle, VAddr e
     thread->SetName(
         fmt::format("thread[entry_point={:X}, handle={:X}]", entry_point, *new_thread_handle));
 
-    system.CpuCore(thread->GetProcessorID()).PrepareReschedule();
+    system.PrepareReschedule(thread->GetProcessorID());
 
     return RESULT_SUCCESS;
 }
@@ -1525,7 +1529,7 @@ static ResultCode StartThread(Core::System& system, Handle thread_handle) {
     thread->ResumeFromWait();
 
     if (thread->GetStatus() == ThreadStatus::Ready) {
-        system.CpuCore(thread->GetProcessorID()).PrepareReschedule();
+        system.PrepareReschedule(thread->GetProcessorID());
     }
 
     return RESULT_SUCCESS;
@@ -1537,7 +1541,7 @@ static void ExitThread(Core::System& system) {
 
     auto* const current_thread = system.CurrentScheduler().GetCurrentThread();
     current_thread->Stop();
-    system.CurrentScheduler().RemoveThread(current_thread);
+    system.GlobalScheduler().RemoveThread(current_thread);
     system.PrepareReschedule();
 }
 
@@ -1557,13 +1561,13 @@ static void SleepThread(Core::System& system, s64 nanoseconds) {
     if (nanoseconds <= 0) {
         switch (static_cast<SleepType>(nanoseconds)) {
         case SleepType::YieldWithoutLoadBalancing:
-            scheduler.YieldWithoutLoadBalancing(current_thread);
+            current_thread->YieldSimple();
             break;
         case SleepType::YieldWithLoadBalancing:
-            scheduler.YieldWithLoadBalancing(current_thread);
+            current_thread->YieldAndBalanceLoad();
             break;
         case SleepType::YieldAndWaitForLoadBalancing:
-            scheduler.YieldAndWaitForLoadBalancing(current_thread);
+            current_thread->YieldAndWaitForLoadBalancing();
             break;
         default:
             UNREACHABLE_MSG("Unimplemented sleep yield type '{:016X}'!", nanoseconds);
@@ -1572,10 +1576,7 @@ static void SleepThread(Core::System& system, s64 nanoseconds) {
         current_thread->Sleep(nanoseconds);
     }
 
-    // Reschedule all CPU cores
-    for (std::size_t i = 0; i < Core::NUM_CPU_CORES; ++i) {
-        system.CpuCore(i).PrepareReschedule();
-    }
+    system.PrepareReschedule(current_thread->GetProcessorID());
 }
 
 /// Wait process wide key atomic
@@ -1606,12 +1607,14 @@ static ResultCode WaitProcessWideKeyAtomic(Core::System& system, VAddr mutex_add
     SharedPtr<Thread> thread = handle_table.Get<Thread>(thread_handle);
     ASSERT(thread);
 
-    const auto release_result = current_process->GetMutex().Release(mutex_addr);
+    SharedPtr<Thread> current_thread = system.CurrentScheduler().GetCurrentThread();
+
+    const auto release_result =
+        current_process->GetMutex().Release(mutex_addr, current_thread.get());
     if (release_result.IsError()) {
         return release_result;
     }
 
-    SharedPtr<Thread> current_thread = system.CurrentScheduler().GetCurrentThread();
     current_thread->SetCondVarWaitAddress(condition_variable_addr);
     current_thread->SetMutexWaitAddress(mutex_addr);
     current_thread->SetWaitHandle(thread_handle);
@@ -1622,7 +1625,7 @@ static ResultCode WaitProcessWideKeyAtomic(Core::System& system, VAddr mutex_add
 
     // Note: Deliberately don't attempt to inherit the lock owner's priority.
 
-    system.CpuCore(current_thread->GetProcessorID()).PrepareReschedule();
+    system.PrepareReschedule(current_thread->GetProcessorID());
     return RESULT_SUCCESS;
 }
 
@@ -1632,24 +1635,17 @@ static ResultCode SignalProcessWideKey(Core::System& system, VAddr condition_var
     LOG_TRACE(Kernel_SVC, "called, condition_variable_addr=0x{:X}, target=0x{:08X}",
               condition_variable_addr, target);
 
-    const auto RetrieveWaitingThreads = [&system](std::size_t core_index,
-                                                  std::vector<SharedPtr<Thread>>& waiting_threads,
-                                                  VAddr condvar_addr) {
-        const auto& scheduler = system.Scheduler(core_index);
-        const auto& thread_list = scheduler.GetThreadList();
-
-        for (const auto& thread : thread_list) {
-            if (thread->GetCondVarWaitAddress() == condvar_addr)
-                waiting_threads.push_back(thread);
-        }
-    };
-
     // Retrieve a list of all threads that are waiting for this condition variable.
     std::vector<SharedPtr<Thread>> waiting_threads;
-    RetrieveWaitingThreads(0, waiting_threads, condition_variable_addr);
-    RetrieveWaitingThreads(1, waiting_threads, condition_variable_addr);
-    RetrieveWaitingThreads(2, waiting_threads, condition_variable_addr);
-    RetrieveWaitingThreads(3, waiting_threads, condition_variable_addr);
+    const auto& scheduler = system.GlobalScheduler();
+    const auto& thread_list = scheduler.GetThreadList();
+
+    for (const auto& thread : thread_list) {
+        if (thread->GetCondVarWaitAddress() == condition_variable_addr) {
+            waiting_threads.push_back(thread);
+        }
+    }
+
     // Sort them by priority, such that the highest priority ones come first.
     std::sort(waiting_threads.begin(), waiting_threads.end(),
               [](const SharedPtr<Thread>& lhs, const SharedPtr<Thread>& rhs) {
@@ -1704,7 +1700,7 @@ static ResultCode SignalProcessWideKey(Core::System& system, VAddr condition_var
             thread->SetLockOwner(nullptr);
             thread->SetMutexWaitAddress(0);
             thread->SetWaitHandle(0);
-            system.CpuCore(thread->GetProcessorID()).PrepareReschedule();
+            system.PrepareReschedule(thread->GetProcessorID());
         } else {
             // Atomically signal that the mutex now has a waiting thread.
             do {
@@ -1728,6 +1724,7 @@ static ResultCode SignalProcessWideKey(Core::System& system, VAddr condition_var
             thread->SetStatus(ThreadStatus::WaitMutex);
 
             owner->AddMutexWaiter(thread);
+            system.PrepareReschedule(thread->GetProcessorID());
         }
     }
 
@@ -1754,7 +1751,12 @@ static ResultCode WaitForAddress(Core::System& system, VAddr address, u32 type,
 
     const auto arbitration_type = static_cast<AddressArbiter::ArbitrationType>(type);
     auto& address_arbiter = system.Kernel().CurrentProcess()->GetAddressArbiter();
-    return address_arbiter.WaitForAddress(address, arbitration_type, value, timeout);
+    const ResultCode result =
+        address_arbiter.WaitForAddress(address, arbitration_type, value, timeout);
+    if (result == RESULT_SUCCESS) {
+        system.PrepareReschedule();
+    }
+    return result;
 }
 
 // Signals to an address (via Address Arbiter)
@@ -2040,7 +2042,10 @@ static ResultCode SetThreadCoreMask(Core::System& system, Handle thread_handle,
         return ERR_INVALID_HANDLE;
     }
 
+    system.PrepareReschedule(thread->GetProcessorID());
     thread->ChangeCore(core, affinity_mask);
+    system.PrepareReschedule(thread->GetProcessorID());
+
     return RESULT_SUCCESS;
 }
 
@@ -2151,6 +2156,7 @@ static ResultCode SignalEvent(Core::System& system, Handle handle) {
     }
 
     writable_event->Signal();
+    system.PrepareReschedule();
     return RESULT_SUCCESS;
 }
 

src/core/hle/kernel/thread.cpp

@@ -45,15 +45,7 @@ void Thread::Stop() {
                                                              callback_handle);
     kernel.ThreadWakeupCallbackHandleTable().Close(callback_handle);
     callback_handle = 0;
-
-    // Clean up thread from ready queue
-    // This is only needed when the thread is terminated forcefully (SVC TerminateProcess)
-    if (status == ThreadStatus::Ready || status == ThreadStatus::Paused) {
-        scheduler->UnscheduleThread(this, current_priority);
-    }
-
-    status = ThreadStatus::Dead;
-
+    SetStatus(ThreadStatus::Dead);
     WakeupAllWaitingThreads();
 
     // Clean up any dangling references in objects that this thread was waiting for
@@ -132,13 +124,11 @@ void Thread::ResumeFromWait() {
     wakeup_callback = nullptr;
 
     if (activity == ThreadActivity::Paused) {
-        status = ThreadStatus::Paused;
+        SetStatus(ThreadStatus::Paused);
         return;
     }
 
-    status = ThreadStatus::Ready;
-
-    ChangeScheduler();
+    SetStatus(ThreadStatus::Ready);
 }
 
 void Thread::CancelWait() {
@@ -205,9 +195,9 @@ ResultVal<SharedPtr<Thread>> Thread::Create(KernelCore& kernel, std::string name
     thread->name = std::move(name);
     thread->callback_handle = kernel.ThreadWakeupCallbackHandleTable().Create(thread).Unwrap();
     thread->owner_process = &owner_process;
+    auto& scheduler = kernel.GlobalScheduler();
+    scheduler.AddThread(thread);
     thread->tls_address = thread->owner_process->CreateTLSRegion();
-    thread->scheduler = &system.Scheduler(processor_id);
-    thread->scheduler->AddThread(thread);
 
     thread->owner_process->RegisterThread(thread.get());
 
@@ -250,6 +240,22 @@ void Thread::SetStatus(ThreadStatus new_status) {
         return;
     }
 
+    switch (new_status) {
+    case ThreadStatus::Ready:
+    case ThreadStatus::Running:
+        SetSchedulingStatus(ThreadSchedStatus::Runnable);
+        break;
+    case ThreadStatus::Dormant:
+        SetSchedulingStatus(ThreadSchedStatus::None);
+        break;
+    case ThreadStatus::Dead:
+        SetSchedulingStatus(ThreadSchedStatus::Exited);
+        break;
+    default:
+        SetSchedulingStatus(ThreadSchedStatus::Paused);
+        break;
+    }
+
     if (status == ThreadStatus::Running) {
         last_running_ticks = Core::System::GetInstance().CoreTiming().GetTicks();
     }
@@ -311,8 +317,7 @@ void Thread::UpdatePriority() {
         return;
     }
 
-    scheduler->SetThreadPriority(this, new_priority);
-    current_priority = new_priority;
+    SetCurrentPriority(new_priority);
 
     if (!lock_owner) {
         return;
@@ -328,47 +333,7 @@ void Thread::UpdatePriority() {
 }
 
 void Thread::ChangeCore(u32 core, u64 mask) {
-    ideal_core = core;
-    affinity_mask = mask;
-    ChangeScheduler();
-}
-
-void Thread::ChangeScheduler() {
-    if (status != ThreadStatus::Ready) {
-        return;
-    }
-
-    auto& system = Core::System::GetInstance();
-    std::optional<s32> new_processor_id{GetNextProcessorId(affinity_mask)};
-
-    if (!new_processor_id) {
-        new_processor_id = processor_id;
-    }
-    if (ideal_core != -1 && system.Scheduler(ideal_core).GetCurrentThread() == nullptr) {
-        new_processor_id = ideal_core;
-    }
-
-    ASSERT(*new_processor_id < 4);
-
-    // Add thread to new core's scheduler
-    auto& next_scheduler = system.Scheduler(*new_processor_id);
-
-    if (*new_processor_id != processor_id) {
-        // Remove thread from previous core's scheduler
-        scheduler->RemoveThread(this);
-        next_scheduler.AddThread(this);
-    }
-
-    processor_id = *new_processor_id;
-
-    // If the thread was ready, unschedule from the previous core and schedule on the new core
-    scheduler->UnscheduleThread(this, current_priority);
-    next_scheduler.ScheduleThread(this, current_priority);
-
-    // Change thread's scheduler
-    scheduler = &next_scheduler;
-
-    system.CpuCore(processor_id).PrepareReschedule();
+    SetCoreAndAffinityMask(core, mask);
 }
 
 bool Thread::AllWaitObjectsReady() const {
@@ -391,7 +356,7 @@ void Thread::SetActivity(ThreadActivity value) {
         if (status == ThreadStatus::Ready) {
             status = ThreadStatus::Paused;
         } else if (status == ThreadStatus::Running) {
-            status = ThreadStatus::Paused;
+            SetStatus(ThreadStatus::Paused);
             Core::System::GetInstance().CpuCore(processor_id).PrepareReschedule();
         }
     } else if (status == ThreadStatus::Paused) {
@@ -408,6 +373,171 @@ void Thread::Sleep(s64 nanoseconds) {
     WakeAfterDelay(nanoseconds);
 }
 
+void Thread::YieldSimple() {
+    auto& scheduler = kernel.GlobalScheduler();
+    scheduler.YieldThread(this);
+}
+
+void Thread::YieldAndBalanceLoad() {
+    auto& scheduler = kernel.GlobalScheduler();
+    scheduler.YieldThreadAndBalanceLoad(this);
+}
+
+void Thread::YieldAndWaitForLoadBalancing() {
+    auto& scheduler = kernel.GlobalScheduler();
+    scheduler.YieldThreadAndWaitForLoadBalancing(this);
+}
+
+void Thread::SetSchedulingStatus(ThreadSchedStatus new_status) {
+    const u32 old_flags = scheduling_state;
+    scheduling_state =
+        (scheduling_state & ThreadSchedMasks::HighMask) | static_cast<u32>(new_status);
+    AdjustSchedulingOnStatus(old_flags);
+}
+
+void Thread::SetCurrentPriority(u32 new_priority) {
+    u32 old_priority = std::exchange(current_priority, new_priority);
+    AdjustSchedulingOnPriority(old_priority);
+}
+
+ResultCode Thread::SetCoreAndAffinityMask(s32 new_core, u64 new_affinity_mask) {
+    const auto HighestSetCore = [](u64 mask, u32 max_cores) {
+        for (s32 core = max_cores - 1; core >= 0; core--) {
+            if (((mask >> core) & 1) != 0) {
+                return core;
+            }
+        }
+        return -1;
+    };
+
+    const bool use_override = affinity_override_count != 0;
+    // The value -3 is "do not change the ideal core".
+    if (new_core == -3) {
+        new_core = use_override ? ideal_core_override : ideal_core;
+        if ((new_affinity_mask & (1 << new_core)) == 0) {
+            return ERR_INVALID_COMBINATION;
+        }
+    }
+    if (use_override) {
+        ideal_core_override = new_core;
+        affinity_mask_override = new_affinity_mask;
+    } else {
+        const u64 old_affinity_mask = std::exchange(affinity_mask, new_affinity_mask);
+        ideal_core = new_core;
+        if (old_affinity_mask != new_affinity_mask) {
+            const s32 old_core = processor_id;
+            if (processor_id >= 0 && ((affinity_mask >> processor_id) & 1) == 0) {
+                if (ideal_core < 0) {
+                    processor_id = HighestSetCore(affinity_mask, GlobalScheduler::NUM_CPU_CORES);
+                } else {
+                    processor_id = ideal_core;
+                }
+            }
+            AdjustSchedulingOnAffinity(old_affinity_mask, old_core);
+        }
+    }
+    return RESULT_SUCCESS;
+}
+
+void Thread::AdjustSchedulingOnStatus(u32 old_flags) {
+    if (old_flags == scheduling_state) {
+        return;
+    }
+
+    auto& scheduler = kernel.GlobalScheduler();
+    if (static_cast<ThreadSchedStatus>(old_flags & ThreadSchedMasks::LowMask) ==
+        ThreadSchedStatus::Runnable) {
+        // In this case the thread was running, now it's pausing/exitting
+        if (processor_id >= 0) {
+            scheduler.Unschedule(current_priority, processor_id, this);
+        }
+
+        for (u32 core = 0; core < GlobalScheduler::NUM_CPU_CORES; core++) {
+            if (core != processor_id && ((affinity_mask >> core) & 1) != 0) {
+                scheduler.Unsuggest(current_priority, core, this);
+            }
+        }
+    } else if (GetSchedulingStatus() == ThreadSchedStatus::Runnable) {
+        // The thread is now set to running from being stopped
+        if (processor_id >= 0) {
+            scheduler.Schedule(current_priority, processor_id, this);
+        }
+
+        for (u32 core = 0; core < GlobalScheduler::NUM_CPU_CORES; core++) {
+            if (core != processor_id && ((affinity_mask >> core) & 1) != 0) {
+                scheduler.Suggest(current_priority, core, this);
+            }
+        }
+    }
+
+    scheduler.SetReselectionPending();
+}
+
+void Thread::AdjustSchedulingOnPriority(u32 old_priority) {
+    if (GetSchedulingStatus() != ThreadSchedStatus::Runnable) {
+        return;
+    }
+    auto& scheduler = Core::System::GetInstance().GlobalScheduler();
+    if (processor_id >= 0) {
+        scheduler.Unschedule(old_priority, processor_id, this);
+    }
+
+    for (u32 core = 0; core < GlobalScheduler::NUM_CPU_CORES; core++) {
+        if (core != processor_id && ((affinity_mask >> core) & 1) != 0) {
+            scheduler.Unsuggest(old_priority, core, this);
+        }
+    }
+
+    // Add thread to the new priority queues.
+    Thread* current_thread = GetCurrentThread();
+
+    if (processor_id >= 0) {
+        if (current_thread == this) {
+            scheduler.SchedulePrepend(current_priority, processor_id, this);
+        } else {
+            scheduler.Schedule(current_priority, processor_id, this);
+        }
+    }
+
+    for (u32 core = 0; core < GlobalScheduler::NUM_CPU_CORES; core++) {
+        if (core != processor_id && ((affinity_mask >> core) & 1) != 0) {
+            scheduler.Suggest(current_priority, core, this);
+        }
+    }
+
+    scheduler.SetReselectionPending();
+}
+
+void Thread::AdjustSchedulingOnAffinity(u64 old_affinity_mask, s32 old_core) {
+    auto& scheduler = Core::System::GetInstance().GlobalScheduler();
+    if (GetSchedulingStatus() != ThreadSchedStatus::Runnable ||
+        current_priority >= THREADPRIO_COUNT) {
+        return;
+    }
+
+    for (u32 core = 0; core < GlobalScheduler::NUM_CPU_CORES; core++) {
+        if (((old_affinity_mask >> core) & 1) != 0) {
+            if (core == old_core) {
+                scheduler.Unschedule(current_priority, core, this);
+            } else {
+                scheduler.Unsuggest(current_priority, core, this);
+            }
+        }
+    }
+
+    for (u32 core = 0; core < GlobalScheduler::NUM_CPU_CORES; core++) {
+        if (((affinity_mask >> core) & 1) != 0) {
+            if (core == processor_id) {
+                scheduler.Schedule(current_priority, core, this);
+            } else {
+                scheduler.Suggest(current_priority, core, this);
+            }
+        }
+    }
+
+    scheduler.SetReselectionPending();
+}
+
 ////////////////////////////////////////////////////////////////////////////////////////////////////
 
 /**

src/core/hle/kernel/thread.h

@@ -75,6 +75,26 @@ enum class ThreadActivity : u32 {
     Paused = 1,
 };
 
+enum class ThreadSchedStatus : u32 {
+    None = 0,
+    Paused = 1,
+    Runnable = 2,
+    Exited = 3,
+};
+
+enum ThreadSchedFlags : u32 {
+    ProcessPauseFlag = 1 << 4,
+    ThreadPauseFlag = 1 << 5,
+    ProcessDebugPauseFlag = 1 << 6,
+    KernelInitPauseFlag = 1 << 8,
+};
+
+enum ThreadSchedMasks : u32 {
+    LowMask = 0x000f,
+    HighMask = 0xfff0,
+    ForcePauseMask = 0x0070,
+};
+
 class Thread final : public WaitObject {
 public:
     using MutexWaitingThreads = std::vector<SharedPtr<Thread>>;
@@ -278,6 +298,10 @@ public:
         return processor_id;
     }
 
+    void SetProcessorID(s32 new_core) {
+        processor_id = new_core;
+    }
+
     Process* GetOwnerProcess() {
         return owner_process;
     }
@@ -383,11 +407,38 @@ public:
     /// Sleeps this thread for the given amount of nanoseconds.
     void Sleep(s64 nanoseconds);
 
+    /// Yields this thread without rebalancing loads.
+    void YieldSimple();
+
+    /// Yields this thread and does a load rebalancing.
+    void YieldAndBalanceLoad();
+
+    /// Yields this thread and if the core is left idle, loads are rebalanced
+    void YieldAndWaitForLoadBalancing();
+
+    ThreadSchedStatus GetSchedulingStatus() const {
+        return static_cast<ThreadSchedStatus>(scheduling_state & ThreadSchedMasks::LowMask);
+    }
+
+    bool IsRunning() const {
+        return is_running;
+    }
+
+    void SetIsRunning(bool value) {
+        is_running = value;
+    }
+
 private:
     explicit Thread(KernelCore& kernel);
     ~Thread() override;
 
-    void ChangeScheduler();
+    void SetSchedulingStatus(ThreadSchedStatus new_status);
+    void SetCurrentPriority(u32 new_priority);
+    ResultCode SetCoreAndAffinityMask(s32 new_core, u64 new_affinity_mask);
+
+    void AdjustSchedulingOnStatus(u32 old_flags);
+    void AdjustSchedulingOnPriority(u32 old_priority);
+    void AdjustSchedulingOnAffinity(u64 old_affinity_mask, s32 old_core);
 
     Core::ARM_Interface::ThreadContext context{};
 
@@ -453,6 +504,13 @@ private:
 
     ThreadActivity activity = ThreadActivity::Normal;
 
+    s32 ideal_core_override = -1;
+    u64 affinity_mask_override = 0x1;
+    u32 affinity_override_count = 0;
+
+    u32 scheduling_state = 0;
+    bool is_running = false;
+
     std::string name;
 };
 

src/core/hle/kernel/transfer_memory.cpp

@@ -23,6 +23,8 @@ SharedPtr<TransferMemory> TransferMemory::Create(KernelCore& kernel, VAddr base_
     transfer_memory->owner_permissions = permissions;
     transfer_memory->owner_process = kernel.CurrentProcess();
 
+    transfer_memory->MapMemory(base_address, size, permissions);
+
     return transfer_memory;
 }
 

src/core/hle/kernel/wait_object.cpp

@@ -6,6 +6,8 @@
 #include "common/assert.h"
 #include "common/common_types.h"
 #include "common/logging/log.h"
+#include "core/core.h"
+#include "core/core_cpu.h"
 #include "core/hle/kernel/object.h"
 #include "core/hle/kernel/process.h"
 #include "core/hle/kernel/thread.h"
@@ -48,17 +50,8 @@ SharedPtr<Thread> WaitObject::GetHighestPriorityReadyThread() const {
         if (ShouldWait(thread.get()))
             continue;
 
-        // A thread is ready to run if it's either in ThreadStatus::WaitSynch
-        // and the rest of the objects it is waiting on are ready.
-        bool ready_to_run = true;
-        if (thread_status == ThreadStatus::WaitSynch) {
-            ready_to_run = thread->AllWaitObjectsReady();
-        }
-
-        if (ready_to_run) {
-            candidate = thread.get();
-            candidate_priority = thread->GetPriority();
-        }
+        candidate = thread.get();
+        candidate_priority = thread->GetPriority();
     }
 
     return candidate;
@@ -95,6 +88,7 @@ void WaitObject::WakeupWaitingThread(SharedPtr<Thread> thread) {
     }
     if (resume) {
         thread->ResumeFromWait();
+        Core::System::GetInstance().PrepareReschedule(thread->GetProcessorID());
     }
 }
 

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

@@ -841,17 +841,16 @@ private:
     void PopInteractiveOutData(Kernel::HLERequestContext& ctx) {
         LOG_DEBUG(Service_AM, "called");
 
-        IPC::ResponseBuilder rb{ctx, 2, 0, 1};
-
         const auto storage = applet->GetBroker().PopInteractiveDataToGame();
         if (storage == nullptr) {
             LOG_ERROR(Service_AM,
                       "storage is a nullptr. There is no data in the current interactive channel");
-
+            IPC::ResponseBuilder rb{ctx, 2};
             rb.Push(ERR_NO_DATA_IN_CHANNEL);
             return;
         }
 
+        IPC::ResponseBuilder rb{ctx, 2, 0, 1};
         rb.Push(RESULT_SUCCESS);
         rb.PushIpcInterface<IStorage>(std::move(*storage));
     }

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

@@ -27,9 +27,9 @@ AppletDataBroker::AppletDataBroker(Kernel::KernelCore& kernel) {
     state_changed_event = Kernel::WritableEvent::CreateEventPair(
         kernel, Kernel::ResetType::Manual, "ILibraryAppletAccessor:StateChangedEvent");
     pop_out_data_event = Kernel::WritableEvent::CreateEventPair(
-        kernel, Kernel::ResetType::Manual, "ILibraryAppletAccessor:PopDataOutEvent");
+        kernel, Kernel::ResetType::Automatic, "ILibraryAppletAccessor:PopDataOutEvent");
     pop_interactive_out_data_event = Kernel::WritableEvent::CreateEventPair(
-        kernel, Kernel::ResetType::Manual, "ILibraryAppletAccessor:PopInteractiveDataOutEvent");
+        kernel, Kernel::ResetType::Automatic, "ILibraryAppletAccessor:PopInteractiveDataOutEvent");
 }
 
 AppletDataBroker::~AppletDataBroker() = default;

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

@@ -91,6 +91,7 @@ void SoftwareKeyboard::ExecuteInteractive() {
 
     if (status == INTERACTIVE_STATUS_OK) {
         complete = true;
+        broker.SignalStateChanged();
     } else {
         std::array<char16_t, SWKBD_OUTPUT_INTERACTIVE_BUFFER_SIZE / 2 - 2> string;
         std::memcpy(string.data(), data.data() + 4, string.size() * 2);

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

@@ -165,13 +165,14 @@ void Controller_NPad::InitNewlyAddedControler(std::size_t controller_idx) {
     controller.battery_level[0] = BATTERY_FULL;
     controller.battery_level[1] = BATTERY_FULL;
     controller.battery_level[2] = BATTERY_FULL;
+    styleset_changed_events[controller_idx].writable->Signal();
 }
 
 void Controller_NPad::OnInit() {
     auto& kernel = system.Kernel();
     for (std::size_t i = 0; i < styleset_changed_events.size(); i++) {
         styleset_changed_events[i] = Kernel::WritableEvent::CreateEventPair(
-            kernel, Kernel::ResetType::Automatic, fmt::format("npad:NpadStyleSetChanged_{}", i));
+            kernel, Kernel::ResetType::Manual, fmt::format("npad:NpadStyleSetChanged_{}", i));
     }
 
     if (!IsControllerActivated()) {
@@ -433,7 +434,6 @@ void Controller_NPad::SetSupportedNPadIdTypes(u8* data, std::size_t length) {
     supported_npad_id_types.clear();
     supported_npad_id_types.resize(length / sizeof(u32));
     std::memcpy(supported_npad_id_types.data(), data, length);
-    bool had_controller_update = false;
     for (std::size_t i = 0; i < connected_controllers.size(); i++) {
         auto& controller = connected_controllers[i];
         if (!controller.is_connected) {
@@ -452,10 +452,6 @@ void Controller_NPad::SetSupportedNPadIdTypes(u8* data, std::size_t length) {
                 controller.type = requested_controller;
                 InitNewlyAddedControler(i);
             }
-            had_controller_update = true;
-        }
-        if (had_controller_update) {
-            styleset_changed_events[i].writable->Signal();
         }
     }
 }
@@ -481,7 +477,6 @@ void Controller_NPad::SetNpadMode(u32 npad_id, NPadAssignments assignment_mode)
     const std::size_t npad_index = NPadIdToIndex(npad_id);
     ASSERT(npad_index < shared_memory_entries.size());
     if (shared_memory_entries[npad_index].pad_assignment != assignment_mode) {
-        styleset_changed_events[npad_index].writable->Signal();
         shared_memory_entries[npad_index].pad_assignment = assignment_mode;
     }
 }
@@ -507,7 +502,6 @@ Kernel::SharedPtr<Kernel::ReadableEvent> Controller_NPad::GetStyleSetChangedEven
     // TODO(ogniK): Figure out the best time to signal this event. This event seems that it should
     // be signalled at least once, and signaled after a new controller is connected?
     const auto& styleset_event = styleset_changed_events[NPadIdToIndex(npad_id)];
-    styleset_event.writable->Signal();
     return styleset_event.readable;
 }
 

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

@@ -201,13 +201,13 @@ Hid::Hid(Core::System& system) : ServiceFramework("hid"), system(system) {
         {120, &Hid::SetNpadJoyHoldType, "SetNpadJoyHoldType"},
         {121, &Hid::GetNpadJoyHoldType, "GetNpadJoyHoldType"},
         {122, &Hid::SetNpadJoyAssignmentModeSingleByDefault, "SetNpadJoyAssignmentModeSingleByDefault"},
-        {123, nullptr, "SetNpadJoyAssignmentModeSingleByDefault"},
+        {123, &Hid::SetNpadJoyAssignmentModeSingle, "SetNpadJoyAssignmentModeSingle"},
         {124, &Hid::SetNpadJoyAssignmentModeDual, "SetNpadJoyAssignmentModeDual"},
         {125, &Hid::MergeSingleJoyAsDualJoy, "MergeSingleJoyAsDualJoy"},
         {126, &Hid::StartLrAssignmentMode, "StartLrAssignmentMode"},
         {127, &Hid::StopLrAssignmentMode, "StopLrAssignmentMode"},
         {128, &Hid::SetNpadHandheldActivationMode, "SetNpadHandheldActivationMode"},
-        {129, nullptr, "GetNpadHandheldActivationMode"},
+        {129, &Hid::GetNpadHandheldActivationMode, "GetNpadHandheldActivationMode"},
         {130, &Hid::SwapNpadAssignment, "SwapNpadAssignment"},
         {131, nullptr, "IsUnintendedHomeButtonInputProtectionEnabled"},
         {132, nullptr, "EnableUnintendedHomeButtonInputProtection"},
@@ -544,10 +544,126 @@ void Hid::SetNpadJoyAssignmentModeSingleByDefault(Kernel::HLERequestContext& ctx
     LOG_WARNING(Service_HID, "(STUBBED) called, npad_id={}, applet_resource_user_id={}", npad_id,
                 applet_resource_user_id);
 
+    auto& controller = applet_resource->GetController<Controller_NPad>(HidController::NPad);
+    controller.SetNpadMode(npad_id, Controller_NPad::NPadAssignments::Single);
+
     IPC::ResponseBuilder rb{ctx, 2};
     rb.Push(RESULT_SUCCESS);
 }
 
+void Hid::SetNpadJoyAssignmentModeSingle(Kernel::HLERequestContext& ctx) {
+    // TODO: Check the differences between this and SetNpadJoyAssignmentModeSingleByDefault
+    IPC::RequestParser rp{ctx};
+    const auto npad_id{rp.Pop<u32>()};
+    const auto applet_resource_user_id{rp.Pop<u64>()};
+    const auto npad_joy_device_type{rp.Pop<u64>()};
+
+    LOG_WARNING(Service_HID,
+                "(STUBBED) called, npad_id={}, applet_resource_user_id={}, npad_joy_device_type={}",
+                npad_id, applet_resource_user_id, npad_joy_device_type);
+
+    auto& controller = applet_resource->GetController<Controller_NPad>(HidController::NPad);
+    controller.SetNpadMode(npad_id, Controller_NPad::NPadAssignments::Single);
+
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(RESULT_SUCCESS);
+}
+
+void Hid::SetNpadJoyAssignmentModeDual(Kernel::HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    const auto npad_id{rp.Pop<u32>()};
+    const auto applet_resource_user_id{rp.Pop<u64>()};
+
+    LOG_DEBUG(Service_HID, "called, npad_id={}, applet_resource_user_id={}", npad_id,
+              applet_resource_user_id);
+
+    auto& controller = applet_resource->GetController<Controller_NPad>(HidController::NPad);
+    controller.SetNpadMode(npad_id, Controller_NPad::NPadAssignments::Dual);
+
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(RESULT_SUCCESS);
+}
+
+void Hid::MergeSingleJoyAsDualJoy(Kernel::HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    const auto unknown_1{rp.Pop<u32>()};
+    const auto unknown_2{rp.Pop<u32>()};
+    const auto applet_resource_user_id{rp.Pop<u64>()};
+
+    LOG_WARNING(Service_HID,
+                "(STUBBED) called, unknown_1={}, unknown_2={}, applet_resource_user_id={}",
+                unknown_1, unknown_2, applet_resource_user_id);
+
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(RESULT_SUCCESS);
+}
+
+void Hid::StartLrAssignmentMode(Kernel::HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    const auto applet_resource_user_id{rp.Pop<u64>()};
+
+    LOG_DEBUG(Service_HID, "called, applet_resource_user_id={}", applet_resource_user_id);
+    auto& controller = applet_resource->GetController<Controller_NPad>(HidController::NPad);
+    controller.StartLRAssignmentMode();
+
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(RESULT_SUCCESS);
+}
+
+void Hid::StopLrAssignmentMode(Kernel::HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    const auto applet_resource_user_id{rp.Pop<u64>()};
+
+    LOG_DEBUG(Service_HID, "called, applet_resource_user_id={}", applet_resource_user_id);
+    auto& controller = applet_resource->GetController<Controller_NPad>(HidController::NPad);
+    controller.StopLRAssignmentMode();
+
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(RESULT_SUCCESS);
+}
+
+void Hid::SetNpadHandheldActivationMode(Kernel::HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    const auto applet_resource_user_id{rp.Pop<u64>()};
+    const auto mode{rp.Pop<u64>()};
+
+    LOG_WARNING(Service_HID, "(STUBBED) called, applet_resource_user_id={}, mode={}",
+                applet_resource_user_id, mode);
+
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(RESULT_SUCCESS);
+}
+
+void Hid::GetNpadHandheldActivationMode(Kernel::HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    const auto applet_resource_user_id{rp.Pop<u64>()};
+
+    LOG_WARNING(Service_HID, "(STUBBED) called, applet_resource_user_id={}",
+                applet_resource_user_id);
+
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(RESULT_SUCCESS);
+}
+
+void Hid::SwapNpadAssignment(Kernel::HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    const auto npad_1{rp.Pop<u32>()};
+    const auto npad_2{rp.Pop<u32>()};
+    const auto applet_resource_user_id{rp.Pop<u64>()};
+
+    LOG_DEBUG(Service_HID, "called, applet_resource_user_id={}, npad_1={}, npad_2={}",
+              applet_resource_user_id, npad_1, npad_2);
+
+    auto& controller = applet_resource->GetController<Controller_NPad>(HidController::NPad);
+    IPC::ResponseBuilder rb{ctx, 2};
+    if (controller.SwapNpadAssignment(npad_1, npad_2)) {
+        rb.Push(RESULT_SUCCESS);
+    } else {
+        LOG_ERROR(Service_HID, "Npads are not connected!");
+        rb.Push(ERR_NPAD_NOT_CONNECTED);
+    }
+}
+
 void Hid::BeginPermitVibrationSession(Kernel::HLERequestContext& ctx) {
     IPC::RequestParser rp{ctx};
     const auto applet_resource_user_id{rp.Pop<u64>()};
@@ -622,47 +738,6 @@ void Hid::GetActualVibrationValue(Kernel::HLERequestContext& ctx) {
         applet_resource->GetController<Controller_NPad>(HidController::NPad).GetLastVibration());
 }
 
-void Hid::SetNpadJoyAssignmentModeDual(Kernel::HLERequestContext& ctx) {
-    IPC::RequestParser rp{ctx};
-    const auto npad_id{rp.Pop<u32>()};
-    const auto applet_resource_user_id{rp.Pop<u64>()};
-
-    LOG_DEBUG(Service_HID, "called, npad_id={}, applet_resource_user_id={}", npad_id,
-              applet_resource_user_id);
-
-    auto& controller = applet_resource->GetController<Controller_NPad>(HidController::NPad);
-    controller.SetNpadMode(npad_id, Controller_NPad::NPadAssignments::Dual);
-
-    IPC::ResponseBuilder rb{ctx, 2};
-    rb.Push(RESULT_SUCCESS);
-}
-
-void Hid::MergeSingleJoyAsDualJoy(Kernel::HLERequestContext& ctx) {
-    IPC::RequestParser rp{ctx};
-    const auto unknown_1{rp.Pop<u32>()};
-    const auto unknown_2{rp.Pop<u32>()};
-    const auto applet_resource_user_id{rp.Pop<u64>()};
-
-    LOG_WARNING(Service_HID,
-                "(STUBBED) called, unknown_1={}, unknown_2={}, applet_resource_user_id={}",
-                unknown_1, unknown_2, applet_resource_user_id);
-
-    IPC::ResponseBuilder rb{ctx, 2};
-    rb.Push(RESULT_SUCCESS);
-}
-
-void Hid::SetNpadHandheldActivationMode(Kernel::HLERequestContext& ctx) {
-    IPC::RequestParser rp{ctx};
-    const auto applet_resource_user_id{rp.Pop<u64>()};
-    const auto mode{rp.Pop<u64>()};
-
-    LOG_WARNING(Service_HID, "(STUBBED) called, applet_resource_user_id={}, mode={}",
-                applet_resource_user_id, mode);
-
-    IPC::ResponseBuilder rb{ctx, 2};
-    rb.Push(RESULT_SUCCESS);
-}
-
 void Hid::GetVibrationDeviceInfo(Kernel::HLERequestContext& ctx) {
     LOG_DEBUG(Service_HID, "called");
 
@@ -756,49 +831,6 @@ void Hid::SetPalmaBoostMode(Kernel::HLERequestContext& ctx) {
     rb.Push(RESULT_SUCCESS);
 }
 
-void Hid::StartLrAssignmentMode(Kernel::HLERequestContext& ctx) {
-    IPC::RequestParser rp{ctx};
-    const auto applet_resource_user_id{rp.Pop<u64>()};
-
-    LOG_DEBUG(Service_HID, "called, applet_resource_user_id={}", applet_resource_user_id);
-    auto& controller = applet_resource->GetController<Controller_NPad>(HidController::NPad);
-    controller.StartLRAssignmentMode();
-
-    IPC::ResponseBuilder rb{ctx, 2};
-    rb.Push(RESULT_SUCCESS);
-}
-
-void Hid::StopLrAssignmentMode(Kernel::HLERequestContext& ctx) {
-    IPC::RequestParser rp{ctx};
-    const auto applet_resource_user_id{rp.Pop<u64>()};
-
-    LOG_DEBUG(Service_HID, "called, applet_resource_user_id={}", applet_resource_user_id);
-    auto& controller = applet_resource->GetController<Controller_NPad>(HidController::NPad);
-    controller.StopLRAssignmentMode();
-
-    IPC::ResponseBuilder rb{ctx, 2};
-    rb.Push(RESULT_SUCCESS);
-}
-
-void Hid::SwapNpadAssignment(Kernel::HLERequestContext& ctx) {
-    IPC::RequestParser rp{ctx};
-    const auto npad_1{rp.Pop<u32>()};
-    const auto npad_2{rp.Pop<u32>()};
-    const auto applet_resource_user_id{rp.Pop<u64>()};
-
-    LOG_DEBUG(Service_HID, "called, applet_resource_user_id={}, npad_1={}, npad_2={}",
-              applet_resource_user_id, npad_1, npad_2);
-
-    auto& controller = applet_resource->GetController<Controller_NPad>(HidController::NPad);
-    IPC::ResponseBuilder rb{ctx, 2};
-    if (controller.SwapNpadAssignment(npad_1, npad_2)) {
-        rb.Push(RESULT_SUCCESS);
-    } else {
-        LOG_ERROR(Service_HID, "Npads are not connected!");
-        rb.Push(ERR_NPAD_NOT_CONNECTED);
-    }
-}
-
 class HidDbg final : public ServiceFramework<HidDbg> {
 public:
     explicit HidDbg() : ServiceFramework{"hid:dbg"} {

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

@@ -105,14 +105,19 @@ private:
     void SetNpadJoyHoldType(Kernel::HLERequestContext& ctx);
     void GetNpadJoyHoldType(Kernel::HLERequestContext& ctx);
     void SetNpadJoyAssignmentModeSingleByDefault(Kernel::HLERequestContext& ctx);
+    void SetNpadJoyAssignmentModeSingle(Kernel::HLERequestContext& ctx);
+    void SetNpadJoyAssignmentModeDual(Kernel::HLERequestContext& ctx);
+    void MergeSingleJoyAsDualJoy(Kernel::HLERequestContext& ctx);
+    void StartLrAssignmentMode(Kernel::HLERequestContext& ctx);
+    void StopLrAssignmentMode(Kernel::HLERequestContext& ctx);
+    void SetNpadHandheldActivationMode(Kernel::HLERequestContext& ctx);
+    void GetNpadHandheldActivationMode(Kernel::HLERequestContext& ctx);
+    void SwapNpadAssignment(Kernel::HLERequestContext& ctx);
     void BeginPermitVibrationSession(Kernel::HLERequestContext& ctx);
     void EndPermitVibrationSession(Kernel::HLERequestContext& ctx);
     void SendVibrationValue(Kernel::HLERequestContext& ctx);
     void SendVibrationValues(Kernel::HLERequestContext& ctx);
     void GetActualVibrationValue(Kernel::HLERequestContext& ctx);
-    void SetNpadJoyAssignmentModeDual(Kernel::HLERequestContext& ctx);
-    void MergeSingleJoyAsDualJoy(Kernel::HLERequestContext& ctx);
-    void SetNpadHandheldActivationMode(Kernel::HLERequestContext& ctx);
     void GetVibrationDeviceInfo(Kernel::HLERequestContext& ctx);
     void CreateActiveVibrationDeviceList(Kernel::HLERequestContext& ctx);
     void PermitVibration(Kernel::HLERequestContext& ctx);
@@ -122,9 +127,6 @@ private:
     void StopSixAxisSensor(Kernel::HLERequestContext& ctx);
     void SetIsPalmaAllConnectable(Kernel::HLERequestContext& ctx);
     void SetPalmaBoostMode(Kernel::HLERequestContext& ctx);
-    void StartLrAssignmentMode(Kernel::HLERequestContext& ctx);
-    void StopLrAssignmentMode(Kernel::HLERequestContext& ctx);
-    void SwapNpadAssignment(Kernel::HLERequestContext& ctx);
 
     std::shared_ptr<IAppletResource> applet_resource;
     Core::System& system;

src/core/hle/service/lbl/lbl.cpp

@@ -10,6 +10,8 @@
 #include "core/hle/service/lbl/lbl.h"
 #include "core/hle/service/service.h"
 #include "core/hle/service/sm/sm.h"
+#include "core/settings.h"
+#include "video_core/renderer_base.h"
 
 namespace Service::LBL {
 
@@ -18,21 +20,21 @@ public:
     explicit LBL() : ServiceFramework{"lbl"} {
         // clang-format off
         static const FunctionInfo functions[] = {
-            {0, nullptr, "SaveCurrentSetting"},
-            {1, nullptr, "LoadCurrentSetting"},
-            {2, nullptr, "SetCurrentBrightnessSetting"},
-            {3, nullptr, "GetCurrentBrightnessSetting"},
-            {4, nullptr, "ApplyCurrentBrightnessSettingToBacklight"},
-            {5, nullptr, "GetBrightnessSettingAppliedToBacklight"},
-            {6, nullptr, "SwitchBacklightOn"},
-            {7, nullptr, "SwitchBacklightOff"},
-            {8, nullptr, "GetBacklightSwitchStatus"},
-            {9, nullptr, "EnableDimming"},
-            {10, nullptr, "DisableDimming"},
-            {11, nullptr, "IsDimmingEnabled"},
-            {12, nullptr, "EnableAutoBrightnessControl"},
-            {13, nullptr, "DisableAutoBrightnessControl"},
-            {14, nullptr, "IsAutoBrightnessControlEnabled"},
+            {0, &LBL::SaveCurrentSetting, "SaveCurrentSetting"},
+            {1, &LBL::LoadCurrentSetting, "LoadCurrentSetting"},
+            {2, &LBL::SetCurrentBrightnessSetting, "SetCurrentBrightnessSetting"},
+            {3, &LBL::GetCurrentBrightnessSetting, "GetCurrentBrightnessSetting"},
+            {4, &LBL::ApplyCurrentBrightnessSettingToBacklight, "ApplyCurrentBrightnessSettingToBacklight"},
+            {5, &LBL::GetBrightnessSettingAppliedToBacklight, "GetBrightnessSettingAppliedToBacklight"},
+            {6, &LBL::SwitchBacklightOn, "SwitchBacklightOn"},
+            {7, &LBL::SwitchBacklightOff, "SwitchBacklightOff"},
+            {8, &LBL::GetBacklightSwitchStatus, "GetBacklightSwitchStatus"},
+            {9, &LBL::EnableDimming, "EnableDimming"},
+            {10, &LBL::DisableDimming, "DisableDimming"},
+            {11, &LBL::IsDimmingEnabled, "IsDimmingEnabled"},
+            {12, &LBL::EnableAutoBrightnessControl, "EnableAutoBrightnessControl"},
+            {13, &LBL::DisableAutoBrightnessControl, "DisableAutoBrightnessControl"},
+            {14, &LBL::IsAutoBrightnessControlEnabled, "IsAutoBrightnessControlEnabled"},
             {15, nullptr, "SetAmbientLightSensorValue"},
             {16, nullptr, "GetAmbientLightSensorValue"},
             {17, nullptr, "SetBrightnessReflectionDelayLevel"},
@@ -42,8 +44,8 @@ public:
             {21, nullptr, "SetCurrentAmbientLightSensorMapping"},
             {22, nullptr, "GetCurrentAmbientLightSensorMapping"},
             {23, nullptr, "IsAmbientLightSensorAvailable"},
-            {24, nullptr, "SetCurrentBrightnessSettingForVrMode"},
-            {25, nullptr, "GetCurrentBrightnessSettingForVrMode"},
+            {24, &LBL::SetCurrentBrightnessSettingForVrMode, "SetCurrentBrightnessSettingForVrMode"},
+            {25, &LBL::GetCurrentBrightnessSettingForVrMode, "GetCurrentBrightnessSettingForVrMode"},
             {26, &LBL::EnableVrMode, "EnableVrMode"},
             {27, &LBL::DisableVrMode, "DisableVrMode"},
             {28, &LBL::IsVrModeEnabled, "IsVrModeEnabled"},
@@ -53,13 +55,209 @@ public:
         RegisterHandlers(functions);
     }
 
+    void LoadFromSettings() {
+        current_brightness = Settings::values.backlight_brightness;
+        current_vr_mode_brightness = Settings::values.backlight_brightness;
+
+        if (auto_brightness_enabled) {
+            return;
+        }
+
+        if (vr_mode_enabled) {
+            Renderer().SetCurrentBrightness(current_vr_mode_brightness);
+        } else {
+            Renderer().SetCurrentBrightness(current_brightness);
+        }
+    }
+
 private:
+    f32 GetAutoBrightnessValue() const {
+        return 0.5f;
+    }
+
+    VideoCore::RendererBase& Renderer() {
+        return Core::System::GetInstance().Renderer();
+    }
+
+    void SaveCurrentSetting(Kernel::HLERequestContext& ctx) {
+        LOG_DEBUG(Service_LBL, "called");
+
+        Settings::values.backlight_brightness = current_brightness;
+
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(RESULT_SUCCESS);
+    }
+
+    void LoadCurrentSetting(Kernel::HLERequestContext& ctx) {
+        LOG_DEBUG(Service_LBL, "called");
+
+        LoadFromSettings();
+
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(RESULT_SUCCESS);
+    }
+
+    void SetCurrentBrightnessSetting(Kernel::HLERequestContext& ctx) {
+        IPC::RequestParser rp{ctx};
+        const auto value = rp.PopRaw<f32>();
+
+        LOG_DEBUG(Service_LBL, "called, value={:.3f}", value);
+
+        current_brightness = std::clamp(value, 0.0f, 1.0f);
+
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(RESULT_SUCCESS);
+    }
+
+    void GetCurrentBrightnessSetting(Kernel::HLERequestContext& ctx) {
+        LOG_DEBUG(Service_LBL, "called");
+
+        IPC::ResponseBuilder rb{ctx, 3};
+        rb.Push(RESULT_SUCCESS);
+        rb.Push(current_brightness);
+    }
+
+    void ApplyCurrentBrightnessSettingToBacklight(Kernel::HLERequestContext& ctx) {
+        LOG_DEBUG(Service_LBL, "called");
+
+        if (!auto_brightness_enabled) {
+            Renderer().SetCurrentBrightness(vr_mode_enabled ? current_vr_mode_brightness
+                                                            : current_brightness);
+        }
+
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(RESULT_SUCCESS);
+    }
+
+    void GetBrightnessSettingAppliedToBacklight(Kernel::HLERequestContext& ctx) {
+        LOG_DEBUG(Service_LBL, "called");
+
+        IPC::ResponseBuilder rb{ctx, 3};
+        rb.Push(RESULT_SUCCESS);
+        rb.Push(Renderer().GetCurrentResultantBrightness());
+    }
+
+    void SwitchBacklightOn(Kernel::HLERequestContext& ctx) {
+        IPC::RequestParser rp{ctx};
+        const auto fade_time = rp.PopRaw<u64>();
+
+        LOG_DEBUG(Service_LBL, "called, fade_time={:016X}", fade_time);
+
+        Renderer().SetBacklightStatus(true, fade_time);
+
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(RESULT_SUCCESS);
+    }
+
+    void SwitchBacklightOff(Kernel::HLERequestContext& ctx) {
+        IPC::RequestParser rp{ctx};
+        const auto fade_time = rp.PopRaw<u64>();
+
+        LOG_DEBUG(Service_LBL, "called, fade_time={:016X}", fade_time);
+
+        Renderer().SetBacklightStatus(false, fade_time);
+
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(RESULT_SUCCESS);
+    }
+
+    void GetBacklightSwitchStatus(Kernel::HLERequestContext& ctx) {
+        LOG_DEBUG(Service_LBL, "called");
+
+        IPC::ResponseBuilder rb{ctx, 3};
+        rb.Push(RESULT_SUCCESS);
+        rb.Push<u8>(Renderer().GetBacklightStatus());
+    }
+
+    void EnableDimming(Kernel::HLERequestContext& ctx) {
+        LOG_DEBUG(Service_LBL, "called");
+
+        dimming_enabled = true;
+
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(RESULT_SUCCESS);
+    }
+
+    void DisableDimming(Kernel::HLERequestContext& ctx) {
+        LOG_DEBUG(Service_LBL, "callled");
+
+        dimming_enabled = false;
+
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(RESULT_SUCCESS);
+    }
+
+    void IsDimmingEnabled(Kernel::HLERequestContext& ctx) {
+        LOG_DEBUG(Service_LBL, "called");
+
+        IPC::ResponseBuilder rb{ctx, 3};
+        rb.Push(RESULT_SUCCESS);
+        rb.Push<u8>(dimming_enabled);
+    }
+
+    void EnableAutoBrightnessControl(Kernel::HLERequestContext& ctx) {
+        LOG_DEBUG(Service_LBL, "called");
+
+        auto_brightness_enabled = true;
+        Renderer().SetCurrentBrightness(GetAutoBrightnessValue());
+
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(RESULT_SUCCESS);
+    }
+
+    void DisableAutoBrightnessControl(Kernel::HLERequestContext& ctx) {
+        LOG_DEBUG(Service_LBL, "called");
+
+        auto_brightness_enabled = false;
+        Renderer().SetCurrentBrightness(current_brightness);
+
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(RESULT_SUCCESS);
+    }
+
+    void IsAutoBrightnessControlEnabled(Kernel::HLERequestContext& ctx) {
+        LOG_DEBUG(Service_LBL, "called");
+
+        IPC::ResponseBuilder rb{ctx, 3};
+        rb.Push(RESULT_SUCCESS);
+        rb.Push<u8>(auto_brightness_enabled);
+    }
+
+    void SetCurrentBrightnessSettingForVrMode(Kernel::HLERequestContext& ctx) {
+        IPC::RequestParser rp{ctx};
+        const auto value = rp.PopRaw<f32>();
+
+        LOG_DEBUG(Service_LBL, "called, value={:.3f}", value);
+
+        current_vr_mode_brightness = std::clamp(value, 0.0f, 1.0f);
+
+        if (vr_mode_enabled && !auto_brightness_enabled) {
+            Renderer().SetCurrentBrightness(value);
+        }
+
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(RESULT_SUCCESS);
+    }
+
+    void GetCurrentBrightnessSettingForVrMode(Kernel::HLERequestContext& ctx) {
+        LOG_DEBUG(Service_LBL, "called");
+
+        IPC::ResponseBuilder rb{ctx, 3};
+        rb.Push(RESULT_SUCCESS);
+        rb.Push(current_vr_mode_brightness);
+    }
+
     void EnableVrMode(Kernel::HLERequestContext& ctx) {
         LOG_DEBUG(Service_LBL, "called");
 
         IPC::ResponseBuilder rb{ctx, 2};
         rb.Push(RESULT_SUCCESS);
 
+        if (!vr_mode_enabled && !auto_brightness_enabled &&
+            current_brightness != current_vr_mode_brightness) {
+            Renderer().SetCurrentBrightness(current_vr_mode_brightness);
+        }
+
         vr_mode_enabled = true;
     }
 
@@ -69,6 +267,11 @@ private:
         IPC::ResponseBuilder rb{ctx, 2};
         rb.Push(RESULT_SUCCESS);
 
+        if (vr_mode_enabled && !auto_brightness_enabled &&
+            current_brightness != current_vr_mode_brightness) {
+            Renderer().SetCurrentBrightness(current_brightness);
+        }
+
         vr_mode_enabled = false;
     }
 
@@ -80,9 +283,27 @@ private:
         rb.Push(vr_mode_enabled);
     }
 
+    bool auto_brightness_enabled = false;
+    bool dimming_enabled = true;
+
+    f32 current_brightness = GetAutoBrightnessValue();
+    f32 current_vr_mode_brightness = GetAutoBrightnessValue();
+
     bool vr_mode_enabled = false;
 };
 
+void RequestLoadCurrentSetting(SM::ServiceManager& sm) {
+    if (&sm == nullptr) {
+        return;
+    }
+
+    const auto lbl = sm.GetService<LBL>("lbl");
+
+    if (lbl) {
+        lbl->LoadFromSettings();
+    }
+}
+
 void InstallInterfaces(SM::ServiceManager& sm) {
     std::make_shared<LBL>()->InstallAsService(sm);
 }

src/core/hle/service/lbl/lbl.h

@@ -10,6 +10,9 @@ class ServiceManager;
 
 namespace Service::LBL {
 
+// Requests the LBL service passed to load brightness values from Settings
+void RequestLoadCurrentSetting(SM::ServiceManager& sm);
+
 void InstallInterfaces(SM::ServiceManager& sm);
 
 } // namespace Service::LBL

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

@@ -22,6 +22,18 @@ u32 nvhost_nvdec::ioctl(Ioctl command, const std::vector<u8>& input, const std::
     switch (static_cast<IoctlCommand>(command.raw)) {
     case IoctlCommand::IocSetNVMAPfdCommand:
         return SetNVMAPfd(input, output);
+    case IoctlCommand::IocSubmit:
+        return Submit(input, output);
+    case IoctlCommand::IocGetSyncpoint:
+        return GetSyncpoint(input, output);
+    case IoctlCommand::IocGetWaitbase:
+        return GetWaitbase(input, output);
+    case IoctlCommand::IocMapBuffer:
+        return MapBuffer(input, output);
+    case IoctlCommand::IocMapBufferEx:
+        return MapBufferEx(input, output);
+    case IoctlCommand::IocUnmapBufferEx:
+        return UnmapBufferEx(input, output);
     }
 
     UNIMPLEMENTED_MSG("Unimplemented ioctl");
@@ -30,11 +42,67 @@ u32 nvhost_nvdec::ioctl(Ioctl command, const std::vector<u8>& input, const std::
 
 u32 nvhost_nvdec::SetNVMAPfd(const std::vector<u8>& input, std::vector<u8>& output) {
     IoctlSetNvmapFD params{};
-    std::memcpy(&params, input.data(), input.size());
+    std::memcpy(&params, input.data(), sizeof(IoctlSetNvmapFD));
     LOG_DEBUG(Service_NVDRV, "called, fd={}", params.nvmap_fd);
 
     nvmap_fd = params.nvmap_fd;
     return 0;
 }
 
+u32 nvhost_nvdec::Submit(const std::vector<u8>& input, std::vector<u8>& output) {
+    IoctlSubmit params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlSubmit));
+    LOG_WARNING(Service_NVDRV, "(STUBBED) called");
+    std::memcpy(output.data(), &params, sizeof(IoctlSubmit));
+    return 0;
+}
+
+u32 nvhost_nvdec::GetSyncpoint(const std::vector<u8>& input, std::vector<u8>& output) {
+    IoctlGetSyncpoint params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlGetSyncpoint));
+    LOG_INFO(Service_NVDRV, "called, unknown=0x{:X}", params.unknown);
+    params.value = 0; // Seems to be hard coded at 0
+    std::memcpy(output.data(), &params, sizeof(IoctlGetSyncpoint));
+    return 0;
+}
+
+u32 nvhost_nvdec::GetWaitbase(const std::vector<u8>& input, std::vector<u8>& output) {
+    IoctlGetWaitbase params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlGetWaitbase));
+    LOG_INFO(Service_NVDRV, "called, unknown=0x{:X}", params.unknown);
+    params.value = 0; // Seems to be hard coded at 0
+    std::memcpy(output.data(), &params, sizeof(IoctlGetWaitbase));
+    return 0;
+}
+
+u32 nvhost_nvdec::MapBuffer(const std::vector<u8>& input, std::vector<u8>& output) {
+    IoctlMapBuffer params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlMapBuffer));
+    LOG_WARNING(Service_NVDRV, "(STUBBED) called with address={:08X}{:08X}", params.address_2,
+                params.address_1);
+    params.address_1 = 0;
+    params.address_2 = 0;
+    std::memcpy(output.data(), &params, sizeof(IoctlMapBuffer));
+    return 0;
+}
+
+u32 nvhost_nvdec::MapBufferEx(const std::vector<u8>& input, std::vector<u8>& output) {
+    IoctlMapBufferEx params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlMapBufferEx));
+    LOG_WARNING(Service_NVDRV, "(STUBBED) called with address={:08X}{:08X}", params.address_2,
+                params.address_1);
+    params.address_1 = 0;
+    params.address_2 = 0;
+    std::memcpy(output.data(), &params, sizeof(IoctlMapBufferEx));
+    return 0;
+}
+
+u32 nvhost_nvdec::UnmapBufferEx(const std::vector<u8>& input, std::vector<u8>& output) {
+    IoctlUnmapBufferEx params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlUnmapBufferEx));
+    LOG_WARNING(Service_NVDRV, "(STUBBED) called");
+    std::memcpy(output.data(), &params, sizeof(IoctlUnmapBufferEx));
+    return 0;
+}
+
 } // namespace Service::Nvidia::Devices

src/core/hle/service/nvdrv/devices/nvhost_nvdec.h

@@ -23,16 +23,66 @@ public:
 private:
     enum class IoctlCommand : u32_le {
         IocSetNVMAPfdCommand = 0x40044801,
+        IocSubmit = 0xC0400001,
+        IocGetSyncpoint = 0xC0080002,
+        IocGetWaitbase = 0xC0080003,
+        IocMapBuffer = 0xC01C0009,
+        IocMapBufferEx = 0xC0A40009,
+        IocUnmapBufferEx = 0xC0A4000A,
     };
 
     struct IoctlSetNvmapFD {
         u32_le nvmap_fd;
     };
-    static_assert(sizeof(IoctlSetNvmapFD) == 4, "IoctlSetNvmapFD is incorrect size");
+    static_assert(sizeof(IoctlSetNvmapFD) == 0x4, "IoctlSetNvmapFD is incorrect size");
+
+    struct IoctlSubmit {
+        INSERT_PADDING_BYTES(0x40); // TODO(DarkLordZach): RE this structure
+    };
+    static_assert(sizeof(IoctlSubmit) == 0x40, "IoctlSubmit has incorrect size");
+
+    struct IoctlGetSyncpoint {
+        u32 unknown; // seems to be ignored? Nintendo added this
+        u32 value;
+    };
+    static_assert(sizeof(IoctlGetSyncpoint) == 0x08, "IoctlGetSyncpoint has incorrect size");
+
+    struct IoctlGetWaitbase {
+        u32 unknown; // seems to be ignored? Nintendo added this
+        u32 value;
+    };
+    static_assert(sizeof(IoctlGetWaitbase) == 0x08, "IoctlGetWaitbase has incorrect size");
+
+    struct IoctlMapBuffer {
+        u32 unknown;
+        u32 address_1;
+        u32 address_2;
+        INSERT_PADDING_BYTES(0x10); // TODO(DarkLordZach): RE this structure
+    };
+    static_assert(sizeof(IoctlMapBuffer) == 0x1C, "IoctlMapBuffer is incorrect size");
+
+    struct IoctlMapBufferEx {
+        u32 unknown;
+        u32 address_1;
+        u32 address_2;
+        INSERT_PADDING_BYTES(0x98); // TODO(DarkLordZach): RE this structure
+    };
+    static_assert(sizeof(IoctlMapBufferEx) == 0xA4, "IoctlMapBufferEx has incorrect size");
+
+    struct IoctlUnmapBufferEx {
+        INSERT_PADDING_BYTES(0xA4); // TODO(DarkLordZach): RE this structure
+    };
+    static_assert(sizeof(IoctlUnmapBufferEx) == 0xA4, "IoctlUnmapBufferEx has incorrect size");
 
     u32_le nvmap_fd{};
 
     u32 SetNVMAPfd(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 Submit(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 GetSyncpoint(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 GetWaitbase(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 MapBuffer(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 MapBufferEx(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 UnmapBufferEx(const std::vector<u8>& input, std::vector<u8>& output);
 };
 
 } // namespace Service::Nvidia::Devices

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

@@ -22,6 +22,18 @@ u32 nvhost_vic::ioctl(Ioctl command, const std::vector<u8>& input, const std::ve
     switch (static_cast<IoctlCommand>(command.raw)) {
     case IoctlCommand::IocSetNVMAPfdCommand:
         return SetNVMAPfd(input, output);
+    case IoctlCommand::IocSubmit:
+        return Submit(input, output);
+    case IoctlCommand::IocGetSyncpoint:
+        return GetSyncpoint(input, output);
+    case IoctlCommand::IocGetWaitbase:
+        return GetWaitbase(input, output);
+    case IoctlCommand::IocMapBuffer:
+        return MapBuffer(input, output);
+    case IoctlCommand::IocMapBufferEx:
+        return MapBuffer(input, output);
+    case IoctlCommand::IocUnmapBufferEx:
+        return UnmapBufferEx(input, output);
     }
 
     UNIMPLEMENTED_MSG("Unimplemented ioctl");
@@ -30,11 +42,67 @@ u32 nvhost_vic::ioctl(Ioctl command, const std::vector<u8>& input, const std::ve
 
 u32 nvhost_vic::SetNVMAPfd(const std::vector<u8>& input, std::vector<u8>& output) {
     IoctlSetNvmapFD params{};
-    std::memcpy(&params, input.data(), input.size());
+    std::memcpy(&params, input.data(), sizeof(IoctlSetNvmapFD));
     LOG_DEBUG(Service_NVDRV, "called, fd={}", params.nvmap_fd);
 
     nvmap_fd = params.nvmap_fd;
     return 0;
 }
 
+u32 nvhost_vic::Submit(const std::vector<u8>& input, std::vector<u8>& output) {
+    IoctlSubmit params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlSubmit));
+    LOG_WARNING(Service_NVDRV, "(STUBBED) called");
+    std::memcpy(output.data(), &params, sizeof(IoctlSubmit));
+    return 0;
+}
+
+u32 nvhost_vic::GetSyncpoint(const std::vector<u8>& input, std::vector<u8>& output) {
+    IoctlGetSyncpoint params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlGetSyncpoint));
+    LOG_INFO(Service_NVDRV, "called, unknown=0x{:X}", params.unknown);
+    params.value = 0; // Seems to be hard coded at 0
+    std::memcpy(output.data(), &params, sizeof(IoctlGetSyncpoint));
+    return 0;
+}
+
+u32 nvhost_vic::GetWaitbase(const std::vector<u8>& input, std::vector<u8>& output) {
+    IoctlGetWaitbase params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlGetWaitbase));
+    LOG_INFO(Service_NVDRV, "called, unknown=0x{:X}", params.unknown);
+    params.value = 0; // Seems to be hard coded at 0
+    std::memcpy(output.data(), &params, sizeof(IoctlGetWaitbase));
+    return 0;
+}
+
+u32 nvhost_vic::MapBuffer(const std::vector<u8>& input, std::vector<u8>& output) {
+    IoctlMapBuffer params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlMapBuffer));
+    LOG_WARNING(Service_NVDRV, "(STUBBED) called with address={:08X}{:08X}", params.address_2,
+                params.address_1);
+    params.address_1 = 0;
+    params.address_2 = 0;
+    std::memcpy(output.data(), &params, sizeof(IoctlMapBuffer));
+    return 0;
+}
+
+u32 nvhost_vic::MapBufferEx(const std::vector<u8>& input, std::vector<u8>& output) {
+    IoctlMapBufferEx params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlMapBufferEx));
+    LOG_WARNING(Service_NVDRV, "(STUBBED) called with address={:08X}{:08X}", params.address_2,
+                params.address_1);
+    params.address_1 = 0;
+    params.address_2 = 0;
+    std::memcpy(output.data(), &params, sizeof(IoctlMapBufferEx));
+    return 0;
+}
+
+u32 nvhost_vic::UnmapBufferEx(const std::vector<u8>& input, std::vector<u8>& output) {
+    IoctlUnmapBufferEx params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlUnmapBufferEx));
+    LOG_WARNING(Service_NVDRV, "(STUBBED) called");
+    std::memcpy(output.data(), &params, sizeof(IoctlUnmapBufferEx));
+    return 0;
+}
+
 } // namespace Service::Nvidia::Devices

src/core/hle/service/nvdrv/devices/nvhost_vic.h

@@ -23,6 +23,12 @@ public:
 private:
     enum class IoctlCommand : u32_le {
         IocSetNVMAPfdCommand = 0x40044801,
+        IocSubmit = 0xC0400001,
+        IocGetSyncpoint = 0xC0080002,
+        IocGetWaitbase = 0xC0080003,
+        IocMapBuffer = 0xC01C0009,
+        IocMapBufferEx = 0xC03C0009,
+        IocUnmapBufferEx = 0xC03C000A,
     };
 
     struct IoctlSetNvmapFD {
@@ -30,9 +36,53 @@ private:
     };
     static_assert(sizeof(IoctlSetNvmapFD) == 4, "IoctlSetNvmapFD is incorrect size");
 
+    struct IoctlSubmit {
+        INSERT_PADDING_BYTES(0x40); // TODO(DarkLordZach): RE this structure
+    };
+    static_assert(sizeof(IoctlSubmit) == 0x40, "IoctlSubmit is incorrect size");
+
+    struct IoctlGetSyncpoint {
+        u32 unknown; // seems to be ignored? Nintendo added this
+        u32 value;
+    };
+    static_assert(sizeof(IoctlGetSyncpoint) == 0x8, "IoctlGetSyncpoint is incorrect size");
+
+    struct IoctlGetWaitbase {
+        u32 unknown; // seems to be ignored? Nintendo added this
+        u32 value;
+    };
+    static_assert(sizeof(IoctlGetWaitbase) == 0x8, "IoctlGetWaitbase is incorrect size");
+
+    struct IoctlMapBuffer {
+        u32 unknown;
+        u32 address_1;
+        u32 address_2;
+        INSERT_PADDING_BYTES(0x10); // TODO(DarkLordZach): RE this structure
+    };
+    static_assert(sizeof(IoctlMapBuffer) == 0x1C, "IoctlMapBuffer is incorrect size");
+
+    struct IoctlMapBufferEx {
+        u32 unknown;
+        u32 address_1;
+        u32 address_2;
+        INSERT_PADDING_BYTES(0x30); // TODO(DarkLordZach): RE this structure
+    };
+    static_assert(sizeof(IoctlMapBufferEx) == 0x3C, "IoctlMapBufferEx is incorrect size");
+
+    struct IoctlUnmapBufferEx {
+        INSERT_PADDING_BYTES(0x3C); // TODO(DarkLordZach): RE this structure
+    };
+    static_assert(sizeof(IoctlUnmapBufferEx) == 0x3C, "IoctlUnmapBufferEx is incorrect size");
+
     u32_le nvmap_fd{};
 
     u32 SetNVMAPfd(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 Submit(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 GetSyncpoint(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 GetWaitbase(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 MapBuffer(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 MapBufferEx(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 UnmapBufferEx(const std::vector<u8>& input, std::vector<u8>& output);
 };
 
 } // namespace Service::Nvidia::Devices

src/core/settings.cpp

@@ -6,6 +6,8 @@
 #include "core/core.h"
 #include "core/gdbstub/gdbstub.h"
 #include "core/hle/service/hid/hid.h"
+#include "core/hle/service/lbl/lbl.h"
+#include "core/hle/service/sm/sm.h"
 #include "core/settings.h"
 #include "video_core/renderer_base.h"
 
@@ -70,6 +72,7 @@ void Apply() {
     auto& system_instance = Core::System::GetInstance();
     if (system_instance.IsPoweredOn()) {
         system_instance.Renderer().RefreshBaseSettings();
+        Service::LBL::RequestLoadCurrentSetting(system_instance.ServiceManager());
     }
 
     Service::HID::ReloadInputDevices();

src/core/settings.h

@@ -428,6 +428,8 @@ struct Values {
     float bg_green;
     float bg_blue;
 
+    float backlight_brightness = 0.5f;
+
     std::string log_filter;
 
     bool use_dev_keys;

src/video_core/engines/shader_bytecode.h

@@ -544,7 +544,7 @@ enum class VoteOperation : u64 {
     Eq = 2,  // allThreadsEqualNV
 };
 
-enum class ImageAtomicSize : u64 {
+enum class ImageAtomicOperationType : u64 {
     U32 = 0,
     S32 = 1,
     U64 = 2,
@@ -1432,11 +1432,11 @@ union Instruction {
             ASSERT(mode == SurfaceDataMode::D_BA);
             return store_data_layout;
         }
-    } sust;
+    } suldst;
 
     union {
         BitField<28, 1, u64> is_ba;
-        BitField<51, 3, ImageAtomicSize> size;
+        BitField<51, 3, ImageAtomicOperationType> operation_type;
         BitField<33, 3, ImageType> image_type;
         BitField<29, 4, ImageAtomicOperation> operation;
         BitField<49, 2, OutOfBoundsStore> out_of_bounds_store;
@@ -1595,6 +1595,7 @@ public:
         TMML_B, // Texture Mip Map Level
         TMML,   // Texture Mip Map Level
         SUST,   // Surface Store
+        SULD,   // Surface Load
         SUATOM, // Surface Atomic Operation
         EXIT,
         NOP,
@@ -1884,6 +1885,7 @@ private:
             INST("110111110110----", Id::TMML_B, Type::Texture, "TMML_B"),
             INST("1101111101011---", Id::TMML, Type::Texture, "TMML"),
             INST("11101011001-----", Id::SUST, Type::Image, "SUST"),
+            INST("11101011000-----", Id::SULD, Type::Image, "SULD"),
             INST("1110101000------", Id::SUATOM, Type::Image, "SUATOM_D"),
             INST("0101000010110---", Id::NOP, Type::Trivial, "NOP"),
             INST("11100000--------", Id::IPA, Type::Trivial, "IPA"),

src/video_core/renderer_base.cpp

@@ -40,4 +40,35 @@ void RendererBase::RequestScreenshot(void* data, std::function<void()> callback,
     renderer_settings.screenshot_requested = true;
 }
 
+f32 RendererBase::GetCurrentResultantBrightness() const {
+    return renderer_settings.current_brightness / 2.0f;
+}
+
+void RendererBase::SetBacklightStatus(bool enabled, u64 fade_transition_time) {
+    if (fade_transition_time == 0) {
+        // Needed to ensure the renderer recognizes that a change must occur.
+        fade_transition_time = 1;
+    }
+
+    if (enabled && renderer_settings.current_brightness == 0) {
+        renderer_settings.current_brightness = current_brightness_backup;
+        renderer_settings.backlight_fade_time = fade_transition_time;
+    } else if (!enabled && renderer_settings.current_brightness != 0) {
+        current_brightness_backup = renderer_settings.current_brightness;
+        renderer_settings.current_brightness = 0;
+        renderer_settings.backlight_fade_time = fade_transition_time;
+    }
+}
+
+bool RendererBase::GetBacklightStatus() const {
+    return renderer_settings.current_brightness != 0;
+}
+
+void RendererBase::SetCurrentBrightness(f32 value) {
+    if (value != renderer_settings.current_brightness) {
+        renderer_settings.current_brightness = value * 2.0f;
+        renderer_settings.backlight_fade_time = 1;
+    }
+}
+
 } // namespace VideoCore

src/video_core/renderer_base.h

@@ -28,6 +28,10 @@ struct RendererSettings {
     void* screenshot_bits;
     std::function<void()> screenshot_complete_callback;
     Layout::FramebufferLayout screenshot_framebuffer_layout;
+
+    // Backlight & Brightness
+    std::atomic<f32> current_brightness{1.f};
+    std::atomic<u64> backlight_fade_time{0};
 };
 
 class RendererBase : NonCopyable {
@@ -86,6 +90,17 @@ public:
     void RequestScreenshot(void* data, std::function<void()> callback,
                            const Layout::FramebufferLayout& layout);
 
+    // Gets the current brightness, even if it has been changed from the set value. Most of the time
+    // for yuzu this will simply match what was returned, but implementations are free to change the
+    // value in settings.
+    f32 GetCurrentResultantBrightness() const;
+
+    void SetBacklightStatus(bool enabled, u64 fade_transition_time);
+
+    bool GetBacklightStatus() const;
+
+    void SetCurrentBrightness(f32 value);
+
 protected:
     Core::Frontend::EmuWindow& render_window; ///< Reference to the render window handle.
     std::unique_ptr<RasterizerInterface> rasterizer;
@@ -97,6 +112,9 @@ protected:
 private:
     /// Updates the framebuffer layout of the contained render window handle.
     void UpdateCurrentFramebufferLayout();
+
+    // Value of brightness before backlight switch used to preserve value.
+    f32 current_brightness_backup;
 };
 
 } // namespace VideoCore

src/video_core/renderer_opengl/gl_device.cpp

@@ -2,8 +2,10 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <algorithm>
 #include <array>
 #include <cstddef>
+#include <vector>
 #include <glad/glad.h>
 
 #include "common/logging/log.h"
@@ -30,9 +32,27 @@ bool TestProgram(const GLchar* glsl) {
     return link_status == GL_TRUE;
 }
 
+std::vector<std::string_view> GetExtensions() {
+    GLint num_extensions;
+    glGetIntegerv(GL_NUM_EXTENSIONS, &num_extensions);
+    std::vector<std::string_view> extensions;
+    extensions.reserve(num_extensions);
+    for (GLint index = 0; index < num_extensions; ++index) {
+        extensions.push_back(
+            reinterpret_cast<const char*>(glGetStringi(GL_EXTENSIONS, static_cast<GLuint>(index))));
+    }
+    return extensions;
+}
+
+bool HasExtension(const std::vector<std::string_view>& images, std::string_view extension) {
+    return std::find(images.begin(), images.end(), extension) != images.end();
+}
+
 } // Anonymous namespace
 
 Device::Device() {
+    const std::vector extensions = GetExtensions();
+
     uniform_buffer_alignment = GetInteger<std::size_t>(GL_UNIFORM_BUFFER_OFFSET_ALIGNMENT);
     shader_storage_alignment = GetInteger<std::size_t>(GL_SHADER_STORAGE_BUFFER_OFFSET_ALIGNMENT);
     max_vertex_attributes = GetInteger<u32>(GL_MAX_VERTEX_ATTRIBS);
@@ -40,6 +60,7 @@ Device::Device() {
     has_warp_intrinsics = GLAD_GL_NV_gpu_shader5 && GLAD_GL_NV_shader_thread_group &&
                           GLAD_GL_NV_shader_thread_shuffle;
     has_vertex_viewport_layer = GLAD_GL_ARB_shader_viewport_layer_array;
+    has_image_load_formatted = HasExtension(extensions, "GL_EXT_shader_image_load_formatted");
     has_variable_aoffi = TestVariableAoffi();
     has_component_indexing_bug = TestComponentIndexingBug();
     has_precise_bug = TestPreciseBug();
@@ -55,6 +76,7 @@ Device::Device(std::nullptr_t) {
     max_varyings = 15;
     has_warp_intrinsics = true;
     has_vertex_viewport_layer = true;
+    has_image_load_formatted = true;
     has_variable_aoffi = true;
     has_component_indexing_bug = false;
     has_precise_bug = false;

src/video_core/renderer_opengl/gl_device.h

@@ -38,6 +38,10 @@ public:
         return has_vertex_viewport_layer;
     }
 
+    bool HasImageLoadFormatted() const {
+        return has_image_load_formatted;
+    }
+
     bool HasVariableAoffi() const {
         return has_variable_aoffi;
     }
@@ -61,6 +65,7 @@ private:
     u32 max_varyings{};
     bool has_warp_intrinsics{};
     bool has_vertex_viewport_layer{};
+    bool has_image_load_formatted{};
     bool has_variable_aoffi{};
     bool has_component_indexing_bug{};
     bool has_precise_bug{};

src/video_core/renderer_opengl/gl_shader_cache.cpp

@@ -211,14 +211,14 @@ CachedProgram SpecializeShader(const std::string& code, const GLShader::ShaderEn
     const auto primitive_mode{variant.primitive_mode};
     const auto texture_buffer_usage{variant.texture_buffer_usage};
 
-    std::string source = "#version 430 core\n"
-                         "#extension GL_ARB_separate_shader_objects : enable\n"
-                         "#extension GL_NV_gpu_shader5 : enable\n"
-                         "#extension GL_NV_shader_thread_group : enable\n"
-                         "#extension GL_NV_shader_thread_shuffle : enable\n";
-    if (entries.shader_viewport_layer_array) {
-        source += "#extension GL_ARB_shader_viewport_layer_array : enable\n";
-    }
+    std::string source = R"(#version 430 core
+#extension GL_ARB_separate_shader_objects : enable
+#extension GL_ARB_shader_viewport_layer_array : enable
+#extension GL_EXT_shader_image_load_formatted : enable
+#extension GL_NV_gpu_shader5 : enable
+#extension GL_NV_shader_thread_group : enable
+#extension GL_NV_shader_thread_shuffle : enable
+)";
     if (program_type == ProgramType::Compute) {
         source += "#extension GL_ARB_compute_variable_group_size : require\n";
     }

src/video_core/renderer_opengl/gl_shader_decompiler.cpp

@@ -19,6 +19,7 @@
 #include "video_core/renderer_opengl/gl_device.h"
 #include "video_core/renderer_opengl/gl_rasterizer.h"
 #include "video_core/renderer_opengl/gl_shader_decompiler.h"
+#include "video_core/shader/node.h"
 #include "video_core/shader/shader_ir.h"
 
 namespace OpenGL::GLShader {
@@ -241,6 +242,26 @@ constexpr const char* GetTypeString(Type type) {
     }
 }
 
+constexpr const char* GetImageTypeDeclaration(Tegra::Shader::ImageType image_type) {
+    switch (image_type) {
+    case Tegra::Shader::ImageType::Texture1D:
+        return "1D";
+    case Tegra::Shader::ImageType::TextureBuffer:
+        return "Buffer";
+    case Tegra::Shader::ImageType::Texture1DArray:
+        return "1DArray";
+    case Tegra::Shader::ImageType::Texture2D:
+        return "2D";
+    case Tegra::Shader::ImageType::Texture2DArray:
+        return "2DArray";
+    case Tegra::Shader::ImageType::Texture3D:
+        return "3D";
+    default:
+        UNREACHABLE();
+        return "1D";
+    }
+}
+
 /// Generates code to use for a swizzle operation.
 constexpr const char* GetSwizzle(u32 element) {
     constexpr std::array swizzle = {".x", ".y", ".z", ".w"};
@@ -398,8 +419,6 @@ public:
                                                        usage.is_read, usage.is_written);
         }
         entries.clip_distances = ir.GetClipDistances();
-        entries.shader_viewport_layer_array =
-            IsVertexShader(stage) && (ir.UsesLayer() || ir.UsesViewportIndex());
         entries.shader_length = ir.GetLength();
         return entries;
     }
@@ -722,42 +741,6 @@ private:
     void DeclareImages() {
         const auto& images{ir.GetImages()};
         for (const auto& [offset, image] : images) {
-            const char* image_type = [&] {
-                switch (image.GetType()) {
-                case Tegra::Shader::ImageType::Texture1D:
-                    return "image1D";
-                case Tegra::Shader::ImageType::TextureBuffer:
-                    return "imageBuffer";
-                case Tegra::Shader::ImageType::Texture1DArray:
-                    return "image1DArray";
-                case Tegra::Shader::ImageType::Texture2D:
-                    return "image2D";
-                case Tegra::Shader::ImageType::Texture2DArray:
-                    return "image2DArray";
-                case Tegra::Shader::ImageType::Texture3D:
-                    return "image3D";
-                default:
-                    UNREACHABLE();
-                    return "image1D";
-                }
-            }();
-
-            const auto [type_prefix, format] = [&]() -> std::pair<const char*, const char*> {
-                if (!image.IsSizeKnown()) {
-                    return {"", ""};
-                }
-                switch (image.GetSize()) {
-                case Tegra::Shader::ImageAtomicSize::U32:
-                    return {"u", "r32ui, "};
-                case Tegra::Shader::ImageAtomicSize::S32:
-                    return {"i", "r32i, "};
-                default:
-                    UNIMPLEMENTED_MSG("Unimplemented atomic size={}",
-                                      static_cast<u32>(image.GetSize()));
-                    return {"", ""};
-                }
-            }();
-
             std::string qualifier = "coherent volatile";
             if (image.IsRead() && !image.IsWritten()) {
                 qualifier += " readonly";
@@ -765,9 +748,10 @@ private:
                 qualifier += " writeonly";
             }
 
-            code.AddLine("layout (binding = IMAGE_BINDING_{}) {} uniform "
-                         "{} {};",
-                         image.GetIndex(), qualifier, image_type, GetImage(image));
+            const char* format = image.IsAtomic() ? "r32ui, " : "";
+            const char* type_declaration = GetImageTypeDeclaration(image.GetType());
+            code.AddLine("layout ({}binding = IMAGE_BINDING_{}) {} uniform uimage{} {};", format,
+                         image.GetIndex(), qualifier, type_declaration, GetImage(image));
         }
         if (!images.empty()) {
             code.AddNewLine();
@@ -1234,28 +1218,13 @@ private:
     }
 
     std::string BuildImageValues(Operation operation) {
+        constexpr std::array constructors{"uint", "uvec2", "uvec3", "uvec4"};
         const auto meta{std::get<MetaImage>(operation.GetMeta())};
-        const auto [constructors, type] = [&]() -> std::pair<std::array<const char*, 4>, Type> {
-            constexpr std::array float_constructors{"float", "vec2", "vec3", "vec4"};
-            if (!meta.image.IsSizeKnown()) {
-                return {float_constructors, Type::Float};
-            }
-            switch (meta.image.GetSize()) {
-            case Tegra::Shader::ImageAtomicSize::U32:
-                return {{"uint", "uvec2", "uvec3", "uvec4"}, Type::Uint};
-            case Tegra::Shader::ImageAtomicSize::S32:
-                return {{"int", "ivec2", "ivec3", "ivec4"}, Type::Uint};
-            default:
-                UNIMPLEMENTED_MSG("Unimplemented image size={}",
-                                  static_cast<u32>(meta.image.GetSize()));
-                return {float_constructors, Type::Float};
-            }
-        }();
 
         const std::size_t values_count{meta.values.size()};
         std::string expr = fmt::format("{}(", constructors.at(values_count - 1));
         for (std::size_t i = 0; i < values_count; ++i) {
-            expr += Visit(meta.values.at(i)).As(type);
+            expr += Visit(meta.values.at(i)).AsUint();
             if (i + 1 < values_count) {
                 expr += ", ";
             }
@@ -1264,29 +1233,6 @@ private:
         return expr;
     }
 
-    Expression AtomicImage(Operation operation, const char* opname) {
-        constexpr std::array constructors{"int(", "ivec2(", "ivec3(", "ivec4("};
-        const auto meta{std::get<MetaImage>(operation.GetMeta())};
-        ASSERT(meta.values.size() == 1);
-        ASSERT(meta.image.IsSizeKnown());
-
-        const auto type = [&]() {
-            switch (const auto size = meta.image.GetSize()) {
-            case Tegra::Shader::ImageAtomicSize::U32:
-                return Type::Uint;
-            case Tegra::Shader::ImageAtomicSize::S32:
-                return Type::Int;
-            default:
-                UNIMPLEMENTED_MSG("Unimplemented image size={}", static_cast<u32>(size));
-                return Type::Uint;
-            }
-        }();
-
-        return {fmt::format("{}({}, {}, {})", opname, GetImage(meta.image),
-                            BuildIntegerCoordinates(operation), Visit(meta.values[0]).As(type)),
-                type};
-    }
-
     Expression Assign(Operation operation) {
         const Node& dest = operation[0];
         const Node& src = operation[1];
@@ -1545,6 +1491,8 @@ private:
         case Tegra::Shader::HalfType::H1_H1:
             return {fmt::format("vec2({}[1])", operand.AsHalfFloat()), Type::HalfFloat};
         }
+        UNREACHABLE();
+        return {"0", Type::Int};
     }
 
     Expression HMergeF32(Operation operation) {
@@ -1809,6 +1757,19 @@ private:
         return {tmp, Type::Float};
     }
 
+    Expression ImageLoad(Operation operation) {
+        if (!device.HasImageLoadFormatted()) {
+            LOG_ERROR(Render_OpenGL,
+                      "Device lacks GL_EXT_shader_image_load_formatted, stubbing image load");
+            return {"0", Type::Int};
+        }
+
+        const auto meta{std::get<MetaImage>(operation.GetMeta())};
+        return {fmt::format("imageLoad({}, {}){}", GetImage(meta.image),
+                            BuildIntegerCoordinates(operation), GetSwizzle(meta.element)),
+                Type::Uint};
+    }
+
     Expression ImageStore(Operation operation) {
         const auto meta{std::get<MetaImage>(operation.GetMeta())};
         code.AddLine("imageStore({}, {}, {});", GetImage(meta.image),
@@ -1816,31 +1777,14 @@ private:
         return {};
     }
 
-    Expression AtomicImageAdd(Operation operation) {
-        return AtomicImage(operation, "imageAtomicAdd");
-    }
+    template <const std::string_view& opname>
+    Expression AtomicImage(Operation operation) {
+        const auto meta{std::get<MetaImage>(operation.GetMeta())};
+        ASSERT(meta.values.size() == 1);
 
-    Expression AtomicImageMin(Operation operation) {
-        return AtomicImage(operation, "imageAtomicMin");
-    }
-
-    Expression AtomicImageMax(Operation operation) {
-        return AtomicImage(operation, "imageAtomicMax");
-    }
-    Expression AtomicImageAnd(Operation operation) {
-        return AtomicImage(operation, "imageAtomicAnd");
-    }
-
-    Expression AtomicImageOr(Operation operation) {
-        return AtomicImage(operation, "imageAtomicOr");
-    }
-
-    Expression AtomicImageXor(Operation operation) {
-        return AtomicImage(operation, "imageAtomicXor");
-    }
-
-    Expression AtomicImageExchange(Operation operation) {
-        return AtomicImage(operation, "imageAtomicExchange");
+        return {fmt::format("imageAtomic{}({}, {}, {})", opname, GetImage(meta.image),
+                            BuildIntegerCoordinates(operation), Visit(meta.values[0]).AsUint()),
+                Type::Uint};
     }
 
     Expression Branch(Operation operation) {
@@ -2035,6 +1979,12 @@ private:
         Func() = delete;
         ~Func() = delete;
 
+        static constexpr std::string_view Add = "Add";
+        static constexpr std::string_view And = "And";
+        static constexpr std::string_view Or = "Or";
+        static constexpr std::string_view Xor = "Xor";
+        static constexpr std::string_view Exchange = "Exchange";
+
         static constexpr std::string_view ShuffleIndexed = "shuffleNV";
         static constexpr std::string_view ShuffleUp = "shuffleUpNV";
         static constexpr std::string_view ShuffleDown = "shuffleDownNV";
@@ -2172,14 +2122,14 @@ private:
         &GLSLDecompiler::TextureQueryLod,
         &GLSLDecompiler::TexelFetch,
 
+        &GLSLDecompiler::ImageLoad,
         &GLSLDecompiler::ImageStore,
-        &GLSLDecompiler::AtomicImageAdd,
-        &GLSLDecompiler::AtomicImageMin,
-        &GLSLDecompiler::AtomicImageMax,
-        &GLSLDecompiler::AtomicImageAnd,
-        &GLSLDecompiler::AtomicImageOr,
-        &GLSLDecompiler::AtomicImageXor,
-        &GLSLDecompiler::AtomicImageExchange,
+
+        &GLSLDecompiler::AtomicImage<Func::Add>,
+        &GLSLDecompiler::AtomicImage<Func::And>,
+        &GLSLDecompiler::AtomicImage<Func::Or>,
+        &GLSLDecompiler::AtomicImage<Func::Xor>,
+        &GLSLDecompiler::AtomicImage<Func::Exchange>,
 
         &GLSLDecompiler::Branch,
         &GLSLDecompiler::BranchIndirect,

src/video_core/renderer_opengl/gl_shader_decompiler.h

@@ -90,7 +90,6 @@ struct ShaderEntries {
     std::vector<ImageEntry> images;
     std::vector<GlobalMemoryEntry> global_memory_entries;
     std::array<bool, Maxwell::NumClipDistances> clip_distances{};
-    bool shader_viewport_layer_array{};
     std::size_t shader_length{};
 };
 

src/video_core/renderer_opengl/gl_shader_disk_cache.cpp

@@ -343,20 +343,17 @@ std::optional<ShaderDiskCacheDecompiled> ShaderDiskCacheOpenGL::LoadDecompiledEn
         u8 is_bindless{};
         u8 is_written{};
         u8 is_read{};
-        u8 is_size_known{};
-        u32 size{};
+        u8 is_atomic{};
         if (!LoadObjectFromPrecompiled(offset) || !LoadObjectFromPrecompiled(index) ||
             !LoadObjectFromPrecompiled(type) || !LoadObjectFromPrecompiled(is_bindless) ||
             !LoadObjectFromPrecompiled(is_written) || !LoadObjectFromPrecompiled(is_read) ||
-            !LoadObjectFromPrecompiled(is_size_known) || !LoadObjectFromPrecompiled(size)) {
+            !LoadObjectFromPrecompiled(is_atomic)) {
             return {};
         }
         entry.entries.images.emplace_back(
             static_cast<std::size_t>(offset), static_cast<std::size_t>(index),
             static_cast<Tegra::Shader::ImageType>(type), is_bindless != 0, is_written != 0,
-            is_read != 0,
-            is_size_known ? std::make_optional(static_cast<Tegra::Shader::ImageAtomicSize>(size))
-                          : std::nullopt);
+            is_read != 0, is_atomic != 0);
     }
 
     u32 global_memory_count{};
@@ -382,12 +379,6 @@ std::optional<ShaderDiskCacheDecompiled> ShaderDiskCacheOpenGL::LoadDecompiledEn
         }
     }
 
-    bool shader_viewport_layer_array{};
-    if (!LoadObjectFromPrecompiled(shader_viewport_layer_array)) {
-        return {};
-    }
-    entry.entries.shader_viewport_layer_array = shader_viewport_layer_array;
-
     u64 shader_length{};
     if (!LoadObjectFromPrecompiled(shader_length)) {
         return {};
@@ -435,14 +426,13 @@ bool ShaderDiskCacheOpenGL::SaveDecompiledFile(u64 unique_identifier, const std:
         return false;
     }
     for (const auto& image : entries.images) {
-        const u32 size = image.IsSizeKnown() ? static_cast<u32>(image.GetSize()) : 0U;
         if (!SaveObjectToPrecompiled(static_cast<u64>(image.GetOffset())) ||
             !SaveObjectToPrecompiled(static_cast<u64>(image.GetIndex())) ||
             !SaveObjectToPrecompiled(static_cast<u32>(image.GetType())) ||
             !SaveObjectToPrecompiled(static_cast<u8>(image.IsBindless() ? 1 : 0)) ||
             !SaveObjectToPrecompiled(static_cast<u8>(image.IsWritten() ? 1 : 0)) ||
             !SaveObjectToPrecompiled(static_cast<u8>(image.IsRead() ? 1 : 0)) ||
-            !SaveObjectToPrecompiled(image.IsSizeKnown()) || !SaveObjectToPrecompiled(size)) {
+            !SaveObjectToPrecompiled(static_cast<u8>(image.IsAtomic() ? 1 : 0))) {
             return false;
         }
     }
@@ -464,10 +454,6 @@ bool ShaderDiskCacheOpenGL::SaveDecompiledFile(u64 unique_identifier, const std:
         }
     }
 
-    if (!SaveObjectToPrecompiled(entries.shader_viewport_layer_array)) {
-        return false;
-    }
-
     if (!SaveObjectToPrecompiled(static_cast<u64>(entries.shader_length))) {
         return false;
     }

src/video_core/renderer_opengl/renderer_opengl.cpp

@@ -54,11 +54,13 @@ in vec2 frag_tex_coord;
 out vec4 color;
 
 uniform sampler2D color_texture;
+uniform vec4 backlight;
 
 void main() {
     // Swap RGBA -> ABGR so we don't have to do this on the CPU. This needs to change if we have to
     // support more framebuffer pixel formats.
-    color = texture(color_texture, frag_tex_coord);
+    // Also multiply the color by the backlight multiplier supplied.
+    color = texture(color_texture, frag_tex_coord) * backlight;
 }
 )";
 
@@ -123,8 +125,13 @@ void RendererOpenGL::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
         // Load the framebuffer from memory, draw it to the screen, and swap buffers
         LoadFBToScreenInfo(*framebuffer);
 
-        if (renderer_settings.screenshot_requested)
+        if (renderer_settings.screenshot_requested) {
             CaptureScreenshot();
+        }
+
+        if (renderer_settings.backlight_fade_time > 0) {
+            UpdateBacklight();
+        }
 
         DrawScreen(render_window.GetFramebufferLayout());
 
@@ -210,9 +217,13 @@ void RendererOpenGL::InitOpenGLObjects() {
     state.Apply();
     uniform_modelview_matrix = glGetUniformLocation(shader.handle, "modelview_matrix");
     uniform_color_texture = glGetUniformLocation(shader.handle, "color_texture");
+    uniform_backlight = glGetUniformLocation(shader.handle, "backlight");
     attrib_position = glGetAttribLocation(shader.handle, "vert_position");
     attrib_tex_coord = glGetAttribLocation(shader.handle, "vert_tex_coord");
 
+    // Initialize backlight
+    glUniform4f(uniform_backlight, 1.f, 1.f, 1.f, 1.f);
+
     // Generate VBO handle for drawing
     vertex_buffer.Create();
 
@@ -421,6 +432,29 @@ void RendererOpenGL::CaptureScreenshot() {
     renderer_settings.screenshot_requested = false;
 }
 
+void RendererOpenGL::UpdateBacklight() {
+    constexpr u64 PER_FRAME_FADE_TIME = 1000000000.0f / 60;
+
+    const auto fade_time = renderer_settings.backlight_fade_time.load(std::memory_order_relaxed);
+    auto value = renderer_settings.current_brightness.load(std::memory_order_relaxed);
+    if (fade_time <= PER_FRAME_FADE_TIME) {
+        glUniform4f(uniform_backlight, value, value, value, value);
+        renderer_settings.backlight_fade_time = 0;
+        fade_time_max = 0;
+    } else {
+        if (fade_time_max == 0) {
+            fade_time_max = fade_time;
+            value_max = value;
+        }
+
+        value += (value_max - value) * PER_FRAME_FADE_TIME / fade_time_max;
+
+        glUniform4f(uniform_backlight, value, value, value, value);
+        renderer_settings.backlight_fade_time -= PER_FRAME_FADE_TIME;
+        renderer_settings.current_brightness = value;
+    }
+}
+
 static const char* GetSource(GLenum source) {
 #define RET(s)                                                                                     \
     case GL_DEBUG_SOURCE_##s:                                                                      \

src/video_core/renderer_opengl/renderer_opengl.h

@@ -70,6 +70,7 @@ private:
     void UpdateFramerate();
 
     void CaptureScreenshot();
+    void UpdateBacklight();
 
     // Loads framebuffer from emulated memory into the display information structure
     void LoadFBToScreenInfo(const Tegra::FramebufferConfig& framebuffer);
@@ -97,6 +98,7 @@ private:
     // Shader uniform location indices
     GLuint uniform_modelview_matrix;
     GLuint uniform_color_texture;
+    GLuint uniform_backlight;
 
     // Shader attribute input indices
     GLuint attrib_position;
@@ -105,6 +107,10 @@ private:
     /// Used for transforming the framebuffer orientation
     Tegra::FramebufferConfig::TransformFlags framebuffer_transform_flags;
     Common::Rectangle<int> framebuffer_crop_rect;
+
+    // Used for backlight transitions
+    u64 fade_time_max = 0;
+    f32 value_max = 0;
 };
 
 } // namespace OpenGL

src/video_core/renderer_vulkan/vk_shader_decompiler.cpp

@@ -19,6 +19,7 @@
 #include "video_core/engines/shader_header.h"
 #include "video_core/renderer_vulkan/vk_device.h"
 #include "video_core/renderer_vulkan/vk_shader_decompiler.h"
+#include "video_core/shader/node.h"
 #include "video_core/shader/shader_ir.h"
 
 namespace Vulkan::VKShader {
@@ -939,6 +940,11 @@ private:
         return {};
     }
 
+    Id ImageLoad(Operation operation) {
+        UNIMPLEMENTED();
+        return {};
+    }
+
     Id ImageStore(Operation operation) {
         UNIMPLEMENTED();
         return {};
@@ -949,16 +955,6 @@ private:
         return {};
     }
 
-    Id AtomicImageMin(Operation operation) {
-        UNIMPLEMENTED();
-        return {};
-    }
-
-    Id AtomicImageMax(Operation operation) {
-        UNIMPLEMENTED();
-        return {};
-    }
-
     Id AtomicImageAnd(Operation operation) {
         UNIMPLEMENTED();
         return {};
@@ -1440,10 +1436,9 @@ private:
         &SPIRVDecompiler::TextureQueryLod,
         &SPIRVDecompiler::TexelFetch,
 
+        &SPIRVDecompiler::ImageLoad,
         &SPIRVDecompiler::ImageStore,
         &SPIRVDecompiler::AtomicImageAdd,
-        &SPIRVDecompiler::AtomicImageMin,
-        &SPIRVDecompiler::AtomicImageMax,
         &SPIRVDecompiler::AtomicImageAnd,
         &SPIRVDecompiler::AtomicImageOr,
         &SPIRVDecompiler::AtomicImageXor,

src/video_core/shader/decode/image.cpp

@@ -41,11 +41,46 @@ u32 ShaderIR::DecodeImage(NodeBlock& bb, u32 pc) {
     const Instruction instr = {program_code[pc]};
     const auto opcode = OpCode::Decode(instr);
 
+    const auto GetCoordinates = [this, instr](Tegra::Shader::ImageType image_type) {
+        std::vector<Node> coords;
+        const std::size_t num_coords{GetImageTypeNumCoordinates(image_type)};
+        coords.reserve(num_coords);
+        for (std::size_t i = 0; i < num_coords; ++i) {
+            coords.push_back(GetRegister(instr.gpr8.Value() + i));
+        }
+        return coords;
+    };
+
     switch (opcode->get().GetId()) {
+    case OpCode::Id::SULD: {
+        UNIMPLEMENTED_IF(instr.suldst.mode != Tegra::Shader::SurfaceDataMode::P);
+        UNIMPLEMENTED_IF(instr.suldst.out_of_bounds_store !=
+                         Tegra::Shader::OutOfBoundsStore::Ignore);
+
+        const auto type{instr.suldst.image_type};
+        auto& image{instr.suldst.is_immediate ? GetImage(instr.image, type)
+                                              : GetBindlessImage(instr.gpr39, type)};
+        image.MarkRead();
+
+        u32 indexer = 0;
+        for (u32 element = 0; element < 4; ++element) {
+            if (!instr.suldst.IsComponentEnabled(element)) {
+                continue;
+            }
+            MetaImage meta{image, {}, element};
+            Node value = Operation(OperationCode::ImageLoad, meta, GetCoordinates(type));
+            SetTemporary(bb, indexer++, std::move(value));
+        }
+        for (u32 i = 0; i < indexer; ++i) {
+            SetRegister(bb, instr.gpr0.Value() + i, GetTemporary(i));
+        }
+        break;
+    }
     case OpCode::Id::SUST: {
-        UNIMPLEMENTED_IF(instr.sust.mode != Tegra::Shader::SurfaceDataMode::P);
-        UNIMPLEMENTED_IF(instr.sust.out_of_bounds_store != Tegra::Shader::OutOfBoundsStore::Ignore);
-        UNIMPLEMENTED_IF(instr.sust.component_mask_selector != 0xf); // Ensure we have an RGBA store
+        UNIMPLEMENTED_IF(instr.suldst.mode != Tegra::Shader::SurfaceDataMode::P);
+        UNIMPLEMENTED_IF(instr.suldst.out_of_bounds_store !=
+                         Tegra::Shader::OutOfBoundsStore::Ignore);
+        UNIMPLEMENTED_IF(instr.suldst.component_mask_selector != 0xf); // Ensure we have RGBA
 
         std::vector<Node> values;
         constexpr std::size_t hardcoded_size{4};
@@ -53,58 +88,51 @@ u32 ShaderIR::DecodeImage(NodeBlock& bb, u32 pc) {
             values.push_back(GetRegister(instr.gpr0.Value() + i));
         }
 
-        std::vector<Node> coords;
-        const std::size_t num_coords{GetImageTypeNumCoordinates(instr.sust.image_type)};
-        for (std::size_t i = 0; i < num_coords; ++i) {
-            coords.push_back(GetRegister(instr.gpr8.Value() + i));
-        }
-
-        const auto type{instr.sust.image_type};
-        auto& image{instr.sust.is_immediate ? GetImage(instr.image, type)
-                                            : GetBindlessImage(instr.gpr39, type)};
+        const auto type{instr.suldst.image_type};
+        auto& image{instr.suldst.is_immediate ? GetImage(instr.image, type)
+                                              : GetBindlessImage(instr.gpr39, type)};
         image.MarkWrite();
 
-        MetaImage meta{image, values};
-        bb.push_back(Operation(OperationCode::ImageStore, meta, std::move(coords)));
+        MetaImage meta{image, std::move(values)};
+        bb.push_back(Operation(OperationCode::ImageStore, meta, GetCoordinates(type)));
         break;
     }
     case OpCode::Id::SUATOM: {
         UNIMPLEMENTED_IF(instr.suatom_d.is_ba != 0);
 
-        Node value = GetRegister(instr.gpr0);
-
-        std::vector<Node> coords;
-        const std::size_t num_coords{GetImageTypeNumCoordinates(instr.sust.image_type)};
-        for (std::size_t i = 0; i < num_coords; ++i) {
-            coords.push_back(GetRegister(instr.gpr8.Value() + i));
-        }
-
         const OperationCode operation_code = [instr] {
-            switch (instr.suatom_d.operation) {
-            case Tegra::Shader::ImageAtomicOperation::Add:
-                return OperationCode::AtomicImageAdd;
-            case Tegra::Shader::ImageAtomicOperation::Min:
-                return OperationCode::AtomicImageMin;
-            case Tegra::Shader::ImageAtomicOperation::Max:
-                return OperationCode::AtomicImageMax;
-            case Tegra::Shader::ImageAtomicOperation::And:
-                return OperationCode::AtomicImageAnd;
-            case Tegra::Shader::ImageAtomicOperation::Or:
-                return OperationCode::AtomicImageOr;
-            case Tegra::Shader::ImageAtomicOperation::Xor:
-                return OperationCode::AtomicImageXor;
-            case Tegra::Shader::ImageAtomicOperation::Exch:
-                return OperationCode::AtomicImageExchange;
+            switch (instr.suatom_d.operation_type) {
+            case Tegra::Shader::ImageAtomicOperationType::S32:
+            case Tegra::Shader::ImageAtomicOperationType::U32:
+                switch (instr.suatom_d.operation) {
+                case Tegra::Shader::ImageAtomicOperation::Add:
+                    return OperationCode::AtomicImageAdd;
+                case Tegra::Shader::ImageAtomicOperation::And:
+                    return OperationCode::AtomicImageAnd;
+                case Tegra::Shader::ImageAtomicOperation::Or:
+                    return OperationCode::AtomicImageOr;
+                case Tegra::Shader::ImageAtomicOperation::Xor:
+                    return OperationCode::AtomicImageXor;
+                case Tegra::Shader::ImageAtomicOperation::Exch:
+                    return OperationCode::AtomicImageExchange;
+                }
             default:
-                UNIMPLEMENTED_MSG("Unimplemented operation={}",
-                                  static_cast<u32>(instr.suatom_d.operation.Value()));
-                return OperationCode::AtomicImageAdd;
+                break;
             }
+            UNIMPLEMENTED_MSG("Unimplemented operation={} type={}",
+                              static_cast<u64>(instr.suatom_d.operation.Value()),
+                              static_cast<u64>(instr.suatom_d.operation_type.Value()));
+            return OperationCode::AtomicImageAdd;
         }();
 
-        const auto& image{GetImage(instr.image, instr.suatom_d.image_type, instr.suatom_d.size)};
+        Node value = GetRegister(instr.gpr0);
+
+        const auto type = instr.suatom_d.image_type;
+        auto& image = GetImage(instr.image, type);
+        image.MarkAtomic();
+
         MetaImage meta{image, {std::move(value)}};
-        SetRegister(bb, instr.gpr0, Operation(operation_code, meta, std::move(coords)));
+        SetRegister(bb, instr.gpr0, Operation(operation_code, meta, GetCoordinates(type)));
         break;
     }
     default:
@@ -114,35 +142,32 @@ u32 ShaderIR::DecodeImage(NodeBlock& bb, u32 pc) {
     return pc;
 }
 
-Image& ShaderIR::GetImage(Tegra::Shader::Image image, Tegra::Shader::ImageType type,
-                          std::optional<Tegra::Shader::ImageAtomicSize> size) {
+Image& ShaderIR::GetImage(Tegra::Shader::Image image, Tegra::Shader::ImageType type) {
     const auto offset{static_cast<std::size_t>(image.index.Value())};
-    if (const auto image = TryUseExistingImage(offset, type, size)) {
+    if (const auto image = TryUseExistingImage(offset, type)) {
         return *image;
     }
 
     const std::size_t next_index{used_images.size()};
-    return used_images.emplace(offset, Image{offset, next_index, type, size}).first->second;
+    return used_images.emplace(offset, Image{offset, next_index, type}).first->second;
 }
 
-Image& ShaderIR::GetBindlessImage(Tegra::Shader::Register reg, Tegra::Shader::ImageType type,
-                                  std::optional<Tegra::Shader::ImageAtomicSize> size) {
+Image& ShaderIR::GetBindlessImage(Tegra::Shader::Register reg, Tegra::Shader::ImageType type) {
     const Node image_register{GetRegister(reg)};
     const auto [base_image, cbuf_index, cbuf_offset]{
         TrackCbuf(image_register, global_code, static_cast<s64>(global_code.size()))};
     const auto cbuf_key{(static_cast<u64>(cbuf_index) << 32) | static_cast<u64>(cbuf_offset)};
 
-    if (const auto image = TryUseExistingImage(cbuf_key, type, size)) {
+    if (const auto image = TryUseExistingImage(cbuf_key, type)) {
         return *image;
     }
 
     const std::size_t next_index{used_images.size()};
-    return used_images.emplace(cbuf_key, Image{cbuf_index, cbuf_offset, next_index, type, size})
+    return used_images.emplace(cbuf_key, Image{cbuf_index, cbuf_offset, next_index, type})
         .first->second;
 }
 
-Image* ShaderIR::TryUseExistingImage(u64 offset, Tegra::Shader::ImageType type,
-                                     std::optional<Tegra::Shader::ImageAtomicSize> size) {
+Image* ShaderIR::TryUseExistingImage(u64 offset, Tegra::Shader::ImageType type) {
     auto it = used_images.find(offset);
     if (it == used_images.end()) {
         return nullptr;
@@ -150,14 +175,6 @@ Image* ShaderIR::TryUseExistingImage(u64 offset, Tegra::Shader::ImageType type,
     auto& image = it->second;
     ASSERT(image.GetType() == type);
 
-    if (size) {
-        // We know the size, if it's known it has to be the same as before, otherwise we can set it.
-        if (image.IsSizeKnown()) {
-            ASSERT(image.GetSize() == size);
-        } else {
-            image.SetSize(*size);
-        }
-    }
     return &image;
 }
 

src/video_core/shader/node.h

@@ -149,10 +149,10 @@ enum class OperationCode {
     TextureQueryLod,        /// (MetaTexture, float[N] coords) -> float4
     TexelFetch,             /// (MetaTexture, int[N], int) -> float4
 
-    ImageStore,          /// (MetaImage, int[N] values) -> void
+    ImageLoad,  /// (MetaImage, int[N] coords) -> void
+    ImageStore, /// (MetaImage, int[N] coords) -> void
+
     AtomicImageAdd,      /// (MetaImage, int[N] coords) -> void
-    AtomicImageMin,      /// (MetaImage, int[N] coords) -> void
-    AtomicImageMax,      /// (MetaImage, int[N] coords) -> void
     AtomicImageAnd,      /// (MetaImage, int[N] coords) -> void
     AtomicImageOr,       /// (MetaImage, int[N] coords) -> void
     AtomicImageXor,      /// (MetaImage, int[N] coords) -> void
@@ -294,21 +294,18 @@ private:
 
 class Image final {
 public:
-    constexpr explicit Image(std::size_t offset, std::size_t index, Tegra::Shader::ImageType type,
-                             std::optional<Tegra::Shader::ImageAtomicSize> size)
-        : offset{offset}, index{index}, type{type}, is_bindless{false}, size{size} {}
+    constexpr explicit Image(std::size_t offset, std::size_t index, Tegra::Shader::ImageType type)
+        : offset{offset}, index{index}, type{type}, is_bindless{false} {}
 
     constexpr explicit Image(u32 cbuf_index, u32 cbuf_offset, std::size_t index,
-                             Tegra::Shader::ImageType type,
-                             std::optional<Tegra::Shader::ImageAtomicSize> size)
+                             Tegra::Shader::ImageType type)
         : offset{(static_cast<u64>(cbuf_index) << 32) | cbuf_offset}, index{index}, type{type},
-          is_bindless{true}, size{size} {}
+          is_bindless{true} {}
 
     constexpr explicit Image(std::size_t offset, std::size_t index, Tegra::Shader::ImageType type,
-                             bool is_bindless, bool is_written, bool is_read,
-                             std::optional<Tegra::Shader::ImageAtomicSize> size)
+                             bool is_bindless, bool is_written, bool is_read, bool is_atomic)
         : offset{offset}, index{index}, type{type}, is_bindless{is_bindless},
-          is_written{is_written}, is_read{is_read}, size{size} {}
+          is_written{is_written}, is_read{is_read}, is_atomic{is_atomic} {}
 
     void MarkWrite() {
         is_written = true;
@@ -318,8 +315,10 @@ public:
         is_read = true;
     }
 
-    void SetSize(Tegra::Shader::ImageAtomicSize size_) {
-        size = size_;
+    void MarkAtomic() {
+        MarkWrite();
+        MarkRead();
+        is_atomic = true;
     }
 
     constexpr std::size_t GetOffset() const {
@@ -346,21 +345,17 @@ public:
         return is_read;
     }
 
+    constexpr bool IsAtomic() const {
+        return is_atomic;
+    }
+
     constexpr std::pair<u32, u32> GetBindlessCBuf() const {
         return {static_cast<u32>(offset >> 32), static_cast<u32>(offset)};
     }
 
-    constexpr bool IsSizeKnown() const {
-        return size.has_value();
-    }
-
-    constexpr Tegra::Shader::ImageAtomicSize GetSize() const {
-        return size.value();
-    }
-
     constexpr bool operator<(const Image& rhs) const {
-        return std::tie(offset, index, type, size, is_bindless) <
-               std::tie(rhs.offset, rhs.index, rhs.type, rhs.size, rhs.is_bindless);
+        return std::tie(offset, index, type, is_bindless) <
+               std::tie(rhs.offset, rhs.index, rhs.type, rhs.is_bindless);
     }
 
 private:
@@ -370,7 +365,7 @@ private:
     bool is_bindless{};
     bool is_written{};
     bool is_read{};
-    std::optional<Tegra::Shader::ImageAtomicSize> size{};
+    bool is_atomic{};
 };
 
 struct GlobalMemoryBase {
@@ -402,6 +397,7 @@ struct MetaTexture {
 struct MetaImage {
     const Image& image;
     std::vector<Node> values;
+    u32 element{};
 };
 
 /// Parameters that modify an operation but are not part of any particular operand

src/video_core/shader/shader_ir.h

@@ -284,16 +284,13 @@ private:
                                       bool is_shadow);
 
     /// Accesses an image.
-    Image& GetImage(Tegra::Shader::Image image, Tegra::Shader::ImageType type,
-                    std::optional<Tegra::Shader::ImageAtomicSize> size = {});
+    Image& GetImage(Tegra::Shader::Image image, Tegra::Shader::ImageType type);
 
     /// Access a bindless image sampler.
-    Image& GetBindlessImage(Tegra::Shader::Register reg, Tegra::Shader::ImageType type,
-                            std::optional<Tegra::Shader::ImageAtomicSize> size = {});
+    Image& GetBindlessImage(Tegra::Shader::Register reg, Tegra::Shader::ImageType type);
 
     /// Tries to access an existing image, updating it's state as needed
-    Image* TryUseExistingImage(u64 offset, Tegra::Shader::ImageType type,
-                               std::optional<Tegra::Shader::ImageAtomicSize> size);
+    Image* TryUseExistingImage(u64 offset, Tegra::Shader::ImageType type);
 
     /// Extracts a sequence of bits from a node
     Node BitfieldExtract(Node value, u32 offset, u32 bits);

src/yuzu/configuration/configure_graphics.cpp

@@ -62,6 +62,8 @@ ConfigureGraphics::ConfigureGraphics(QWidget* parent)
         }
         UpdateBackgroundColorButton(new_bg_color);
     });
+    connect(ui->brightness_reset, &QPushButton::pressed, this,
+            [this] { ui->brightness_slider->setValue(100); });
 }
 
 ConfigureGraphics::~ConfigureGraphics() = default;
@@ -80,6 +82,7 @@ void ConfigureGraphics::SetConfiguration() {
     ui->force_30fps_mode->setChecked(Settings::values.force_30fps_mode);
     UpdateBackgroundColorButton(QColor::fromRgbF(Settings::values.bg_red, Settings::values.bg_green,
                                                  Settings::values.bg_blue));
+    ui->brightness_slider->setValue(Settings::values.backlight_brightness * 100 + 50);
 }
 
 void ConfigureGraphics::ApplyConfiguration() {
@@ -93,6 +96,7 @@ void ConfigureGraphics::ApplyConfiguration() {
     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());
+    Settings::values.backlight_brightness = (ui->brightness_slider->value() - 50.0f) / 100.0f;
 }
 
 void ConfigureGraphics::changeEvent(QEvent* event) {

src/yuzu/configuration/configure_graphics.ui

@@ -111,6 +111,68 @@
           </item>
          </layout>
         </item>
+        <item>
+         <layout class="QHBoxLayout" name="horizontalLayout_3">
+          <item>
+           <widget class="QLabel" name="label_2">
+            <property name="text">
+             <string>Brightness</string>
+            </property>
+           </widget>
+          </item>
+          <item>
+           <spacer name="horizontalSpacer">
+            <property name="orientation">
+             <enum>Qt::Horizontal</enum>
+            </property>
+            <property name="sizeHint" stdset="0">
+             <size>
+              <width>40</width>
+              <height>20</height>
+             </size>
+            </property>
+           </spacer>
+          </item>
+          <item>
+           <widget class="QSlider" name="brightness_slider">
+            <property name="minimum">
+             <number>50</number>
+            </property>
+            <property name="maximum">
+             <number>150</number>
+            </property>
+            <property name="singleStep">
+             <number>10</number>
+            </property>
+            <property name="pageStep">
+             <number>20</number>
+            </property>
+            <property name="value">
+             <number>100</number>
+            </property>
+            <property name="orientation">
+             <enum>Qt::Horizontal</enum>
+            </property>
+            <property name="tickPosition">
+             <enum>QSlider::NoTicks</enum>
+            </property>
+           </widget>
+          </item>
+          <item>
+           <widget class="QPushButton" name="brightness_reset">
+            <property name="sizePolicy">
+             <sizepolicy hsizetype="Fixed" vsizetype="Fixed">
+              <horstretch>0</horstretch>
+              <verstretch>0</verstretch>
+             </sizepolicy>
+            </property>
+            <property name="text">
+             <string>Reset</string>
+            </property>
+           </widget>
+          </item>
+         </layout>
+        </item>
        </layout>
       </widget>
      </item>

src/yuzu/debugger/wait_tree.cpp

@@ -66,10 +66,7 @@ std::vector<std::unique_ptr<WaitTreeThread>> WaitTreeItem::MakeThreadItemList()
     };
 
     const auto& system = Core::System::GetInstance();
-    add_threads(system.Scheduler(0).GetThreadList());
-    add_threads(system.Scheduler(1).GetThreadList());
-    add_threads(system.Scheduler(2).GetThreadList());
-    add_threads(system.Scheduler(3).GetThreadList());
+    add_threads(system.GlobalScheduler().GetThreadList());
 
     return item_list;
 }