core: Implement multicore support.

src/core/arm/unicorn/arm_unicorn.cpp

@@ -52,7 +52,7 @@ static void InterruptHook(uc_engine* uc, u32 intNo, void* user_data) {
 static bool UnmappedMemoryHook(uc_engine* uc, uc_mem_type type, u64 addr, int size, u64 value,
                                void* user_data) {
     ARM_Interface::ThreadContext ctx{};
-    Core::CPU().SaveContext(ctx);
+    Core::CurrentArmInterface().SaveContext(ctx);
     ASSERT_MSG(false, "Attempted to read from unmapped memory: 0x{:X}, pc=0x{:X}, lr=0x{:X}", addr,
                ctx.pc, ctx.cpu_registers[30]);
     return {};

src/core/core.h

@@ -108,20 +108,26 @@ public:
 
     PerfStats::Results GetAndResetPerfStats();
 
-    /**
+    ARM_Interface& CurrentArmInterface() {
-     * Gets a reference to the emulated CPU.
+        return CurrentCpuCore().ArmInterface();
-     * @returns A reference to the emulated CPU.
+    }
-     */
+
-    ARM_Interface& CPU() {
+    ARM_Interface& ArmInterface(size_t core_index) {
-        return CurrentCpuCore().CPU();
+        ASSERT(core_index < NUM_CPU_CORES);
+        return cpu_cores[core_index]->ArmInterface();
     }
 
     Tegra::GPU& GPU() {
         return *gpu_core;
     }
 
-    Kernel::Scheduler& Scheduler() {
+    Kernel::Scheduler& CurrentScheduler() {
-        return CurrentCpuCore().Scheduler();
+        return *CurrentCpuCore().Scheduler();
+    }
+
+    const std::shared_ptr<Kernel::Scheduler>& Scheduler(size_t core_index) {
+        ASSERT(core_index < NUM_CPU_CORES);
+        return cpu_cores[core_index]->Scheduler();
     }
 
     Kernel::SharedPtr<Kernel::Process>& CurrentProcess() {
@@ -198,8 +204,8 @@ private:
     std::map<std::thread::id, std::shared_ptr<Cpu>> thread_to_cpu;
 };
 
-inline ARM_Interface& CPU() {
+inline ARM_Interface& CurrentArmInterface() {
-    return System::GetInstance().CPU();
+    return System::GetInstance().CurrentArmInterface();
 }
 
 inline TelemetrySession& Telemetry() {

src/core/core_cpu.cpp

@@ -33,7 +33,7 @@ Cpu::Cpu(std::shared_ptr<CpuBarrier> cpu_barrier, size_t core_index)
         arm_interface = std::make_shared<ARM_Unicorn>();
     }
 
-    scheduler = std::make_unique<Kernel::Scheduler>(arm_interface.get());
+    scheduler = std::make_shared<Kernel::Scheduler>(arm_interface.get());
 }
 
 void Cpu::RunLoop(bool tight_loop) {

src/core/core_cpu.h

@@ -51,12 +51,16 @@ public:
 
     void PrepareReschedule();
 
-    ARM_Interface& CPU() {
+    ARM_Interface& ArmInterface() {
         return *arm_interface;
     }
 
-    Kernel::Scheduler& Scheduler() {
+    const ARM_Interface& ArmInterface() const {
-        return *scheduler;
+        return *arm_interface;
+    }
+
+    const std::shared_ptr<Kernel::Scheduler>& Scheduler() const {
+        return scheduler;
     }
 
     bool IsMainCore() const {
@@ -68,7 +72,7 @@ private:
 
     std::shared_ptr<ARM_Interface> arm_interface;
     std::shared_ptr<CpuBarrier> cpu_barrier;
-    std::unique_ptr<Kernel::Scheduler> scheduler;
+    std::shared_ptr<Kernel::Scheduler> scheduler;
 
     bool reschedule_pending{};
     size_t core_index;

src/core/gdbstub/gdbstub.cpp

@@ -598,11 +598,11 @@ static void ReadRegister() {
     }
 
     if (id <= SP_REGISTER) {
-        LongToGdbHex(reply, Core::CPU().GetReg(static_cast<int>(id)));
+        LongToGdbHex(reply, Core::CurrentArmInterface().GetReg(static_cast<int>(id)));
     } else if (id == PC_REGISTER) {
-        LongToGdbHex(reply, Core::CPU().GetPC());
+        LongToGdbHex(reply, Core::CurrentArmInterface().GetPC());
     } else if (id == CPSR_REGISTER) {
-        IntToGdbHex(reply, Core::CPU().GetCPSR());
+        IntToGdbHex(reply, Core::CurrentArmInterface().GetCPSR());
     } else {
         return SendReply("E01");
     }
@@ -618,16 +618,16 @@ static void ReadRegisters() {
     u8* bufptr = buffer;
 
     for (int reg = 0; reg <= SP_REGISTER; reg++) {
-        LongToGdbHex(bufptr + reg * 16, Core::CPU().GetReg(reg));
+        LongToGdbHex(bufptr + reg * 16, Core::CurrentArmInterface().GetReg(reg));
     }
 
     bufptr += (32 * 16);
 
-    LongToGdbHex(bufptr, Core::CPU().GetPC());
+    LongToGdbHex(bufptr, Core::CurrentArmInterface().GetPC());
 
     bufptr += 16;
 
-    IntToGdbHex(bufptr, Core::CPU().GetCPSR());
+    IntToGdbHex(bufptr, Core::CurrentArmInterface().GetCPSR());
 
     bufptr += 8;
 
@@ -646,11 +646,11 @@ static void WriteRegister() {
     }
 
     if (id <= SP_REGISTER) {
-        Core::CPU().SetReg(id, GdbHexToLong(buffer_ptr));
+        Core::CurrentArmInterface().SetReg(id, GdbHexToLong(buffer_ptr));
     } else if (id == PC_REGISTER) {
-        Core::CPU().SetPC(GdbHexToLong(buffer_ptr));
+        Core::CurrentArmInterface().SetPC(GdbHexToLong(buffer_ptr));
     } else if (id == CPSR_REGISTER) {
-        Core::CPU().SetCPSR(GdbHexToInt(buffer_ptr));
+        Core::CurrentArmInterface().SetCPSR(GdbHexToInt(buffer_ptr));
     } else {
         return SendReply("E01");
     }
@@ -667,11 +667,11 @@ static void WriteRegisters() {
 
     for (int i = 0, reg = 0; reg <= CPSR_REGISTER; i++, reg++) {
         if (reg <= SP_REGISTER) {
-            Core::CPU().SetReg(reg, GdbHexToLong(buffer_ptr + i * 16));
+            Core::CurrentArmInterface().SetReg(reg, GdbHexToLong(buffer_ptr + i * 16));
         } else if (reg == PC_REGISTER) {
-            Core::CPU().SetPC(GdbHexToLong(buffer_ptr + i * 16));
+            Core::CurrentArmInterface().SetPC(GdbHexToLong(buffer_ptr + i * 16));
         } else if (reg == CPSR_REGISTER) {
-            Core::CPU().SetCPSR(GdbHexToInt(buffer_ptr + i * 16));
+            Core::CurrentArmInterface().SetCPSR(GdbHexToInt(buffer_ptr + i * 16));
         } else {
             UNIMPLEMENTED();
         }

src/core/hle/kernel/svc.cpp

@@ -485,22 +485,28 @@ static void ExitProcess() {
 
     Core::CurrentProcess()->status = ProcessStatus::Exited;
 
-    // Stop all the process threads that are currently waiting for objects.
+    auto stop_threads = [](const std::vector<SharedPtr<Thread>>& thread_list) {
-    auto& thread_list = Core::System::GetInstance().Scheduler().GetThreadList();
+        for (auto& thread : thread_list) {
-    for (auto& thread : thread_list) {
+            if (thread->owner_process != Core::CurrentProcess())
-        if (thread->owner_process != Core::CurrentProcess())
+                continue;
-            continue;
 
-        if (thread == GetCurrentThread())
+            if (thread == GetCurrentThread())
-            continue;
+                continue;
 
-        // TODO(Subv): When are the other running/ready threads terminated?
+            // TODO(Subv): When are the other running/ready threads terminated?
-        ASSERT_MSG(thread->status == THREADSTATUS_WAIT_SYNCH_ANY ||
+            ASSERT_MSG(thread->status == THREADSTATUS_WAIT_SYNCH_ANY ||
-                       thread->status == THREADSTATUS_WAIT_SYNCH_ALL,
+                           thread->status == THREADSTATUS_WAIT_SYNCH_ALL,
-                   "Exiting processes with non-waiting threads is currently unimplemented");
+                       "Exiting processes with non-waiting threads is currently unimplemented");
 
-        thread->Stop();
+            thread->Stop();
-    }
+        }
+    };
+
+    auto& system = Core::System::GetInstance();
+    stop_threads(system.Scheduler(0)->GetThreadList());
+    stop_threads(system.Scheduler(1)->GetThreadList());
+    stop_threads(system.Scheduler(2)->GetThreadList());
+    stop_threads(system.Scheduler(3)->GetThreadList());
 
     // Kill the current thread
     GetCurrentThread()->Stop();
@@ -530,14 +536,9 @@ static ResultCode CreateThread(Handle* out_handle, VAddr entry_point, u64 arg, V
 
     switch (processor_id) {
     case THREADPROCESSORID_0:
-        break;
     case THREADPROCESSORID_1:
     case THREADPROCESSORID_2:
     case THREADPROCESSORID_3:
-        // TODO(bunnei): Implement support for other processor IDs
-        NGLOG_ERROR(Kernel_SVC,
-                    "Newly created thread must run in another thread ({}), unimplemented.",
-                    processor_id);
         break;
     default:
         ASSERT_MSG(false, "Unsupported thread processor ID: {}", processor_id);
@@ -576,7 +577,7 @@ static ResultCode StartThread(Handle thread_handle) {
 
 /// Called when a thread exits
 static void ExitThread() {
-    NGLOG_TRACE(Kernel_SVC, "called, pc=0x{:08X}", Core::CPU().GetPC());
+    NGLOG_TRACE(Kernel_SVC, "called, pc=0x{:08X}", Core::CurrentArmInterface().GetPC());
 
     ExitCurrentThread();
     Core::System::GetInstance().PrepareReschedule();
@@ -588,7 +589,7 @@ static void SleepThread(s64 nanoseconds) {
 
     // Don't attempt to yield execution if there are no available threads to run,
     // this way we avoid a useless reschedule to the idle thread.
-    if (nanoseconds == 0 && !Core::System::GetInstance().Scheduler().HaveReadyThreads())
+    if (nanoseconds == 0 && !Core::System::GetInstance().CurrentScheduler().HaveReadyThreads())
         return;
 
     // Sleep current thread and check for next thread to schedule
@@ -634,7 +635,7 @@ static ResultCode SignalProcessWideKey(VAddr condition_variable_addr, s32 target
                 condition_variable_addr, target);
 
     u32 processed = 0;
-    auto& thread_list = Core::System::GetInstance().Scheduler().GetThreadList();
+    auto& thread_list = Core::System::GetInstance().CurrentScheduler().GetThreadList();
 
     for (auto& thread : thread_list) {
         if (thread->condvar_wait_address != condition_variable_addr)

src/core/hle/kernel/svc_wrap.h

@@ -13,14 +13,14 @@
 
 namespace Kernel {
 
-#define PARAM(n) Core::CPU().GetReg(n)
+#define PARAM(n) Core::CurrentArmInterface().GetReg(n)
 
 /**
  * HLE a function return from the current ARM userland process
  * @param res Result to return
  */
 static inline void FuncReturn(u64 res) {
-    Core::CPU().SetReg(0, res);
+    Core::CurrentArmInterface().SetReg(0, res);
 }
 
 ////////////////////////////////////////////////////////////////////////////////////////////////////
@@ -45,7 +45,7 @@ template <ResultCode func(u32*, u32)>
 void SvcWrap() {
     u32 param_1 = 0;
     u32 retval = func(&param_1, (u32)PARAM(1)).raw;
-    Core::CPU().SetReg(1, param_1);
+    Core::CurrentArmInterface().SetReg(1, param_1);
     FuncReturn(retval);
 }
 
@@ -53,7 +53,7 @@ template <ResultCode func(u32*, u64)>
 void SvcWrap() {
     u32 param_1 = 0;
     u32 retval = func(&param_1, PARAM(1)).raw;
-    Core::CPU().SetReg(1, param_1);
+    Core::CurrentArmInterface().SetReg(1, param_1);
     FuncReturn(retval);
 }
 
@@ -66,7 +66,7 @@ template <ResultCode func(u64*, u64)>
 void SvcWrap() {
     u64 param_1 = 0;
     u32 retval = func(&param_1, PARAM(1)).raw;
-    Core::CPU().SetReg(1, param_1);
+    Core::CurrentArmInterface().SetReg(1, param_1);
     FuncReturn(retval);
 }
 
@@ -85,8 +85,8 @@ void SvcWrap() {
     u32 param_1 = 0;
     u64 param_2 = 0;
     ResultCode retval = func((u32)(PARAM(2) & 0xFFFFFFFF), &param_1, &param_2);
-    Core::CPU().SetReg(1, param_1);
+    Core::CurrentArmInterface().SetReg(1, param_1);
-    Core::CPU().SetReg(2, param_2);
+    Core::CurrentArmInterface().SetReg(2, param_2);
     FuncReturn(retval.raw);
 }
 
@@ -120,7 +120,7 @@ template <ResultCode func(u32*, u64, u64, s64)>
 void SvcWrap() {
     u32 param_1 = 0;
     ResultCode retval = func(&param_1, PARAM(1), (u32)(PARAM(2) & 0xFFFFFFFF), (s64)PARAM(3));
-    Core::CPU().SetReg(1, param_1);
+    Core::CurrentArmInterface().SetReg(1, param_1);
     FuncReturn(retval.raw);
 }
 
@@ -133,7 +133,7 @@ template <ResultCode func(u64*, u64, u64, u64)>
 void SvcWrap() {
     u64 param_1 = 0;
     u32 retval = func(&param_1, PARAM(1), PARAM(2), PARAM(3)).raw;
-    Core::CPU().SetReg(1, param_1);
+    Core::CurrentArmInterface().SetReg(1, param_1);
     FuncReturn(retval);
 }
 
@@ -143,7 +143,7 @@ void SvcWrap() {
     u32 retval =
         func(&param_1, PARAM(1), PARAM(2), PARAM(3), (u32)PARAM(4), (s32)(PARAM(5) & 0xFFFFFFFF))
             .raw;
-    Core::CPU().SetReg(1, param_1);
+    Core::CurrentArmInterface().SetReg(1, param_1);
     FuncReturn(retval);
 }
 
@@ -166,7 +166,7 @@ template <ResultCode func(u32*, u64, u64, u32)>
 void SvcWrap() {
     u32 param_1 = 0;
     u32 retval = func(&param_1, PARAM(1), PARAM(2), (u32)(PARAM(3) & 0xFFFFFFFF)).raw;
-    Core::CPU().SetReg(1, param_1);
+    Core::CurrentArmInterface().SetReg(1, param_1);
     FuncReturn(retval);
 }
 
@@ -175,7 +175,7 @@ void SvcWrap() {
     u32 param_1 = 0;
     u32 retval =
         func(&param_1, PARAM(1), (u32)(PARAM(2) & 0xFFFFFFFF), (u32)(PARAM(3) & 0xFFFFFFFF)).raw;
-    Core::CPU().SetReg(1, param_1);
+    Core::CurrentArmInterface().SetReg(1, param_1);
     FuncReturn(retval);
 }
 

src/core/hle/kernel/thread.cpp

@@ -64,7 +64,7 @@ void Thread::Stop() {
     // Clean up thread from ready queue
     // This is only needed when the thread is termintated forcefully (SVC TerminateProcess)
     if (status == THREADSTATUS_READY) {
-        Core::System::GetInstance().Scheduler().UnscheduleThread(this, current_priority);
+        scheduler->UnscheduleThread(this, current_priority);
     }
 
     status = THREADSTATUS_DEAD;
@@ -92,7 +92,7 @@ void WaitCurrentThread_Sleep() {
 void ExitCurrentThread() {
     Thread* thread = GetCurrentThread();
     thread->Stop();
-    Core::System::GetInstance().Scheduler().RemoveThread(thread);
+    Core::System::GetInstance().CurrentScheduler().RemoveThread(thread);
 }
 
 /**
@@ -188,7 +188,7 @@ void Thread::ResumeFromWait() {
     wakeup_callback = nullptr;
 
     status = THREADSTATUS_READY;
-    Core::System::GetInstance().Scheduler().ScheduleThread(this, current_priority);
+    scheduler->ScheduleThread(this, current_priority);
     Core::System::GetInstance().PrepareReschedule();
 }
 
@@ -259,8 +259,6 @@ ResultVal<SharedPtr<Thread>> Thread::Create(std::string name, VAddr entry_point,
 
     SharedPtr<Thread> thread(new Thread);
 
-    Core::System::GetInstance().Scheduler().AddThread(thread, priority);
-
     thread->thread_id = NewThreadId();
     thread->status = THREADSTATUS_DORMANT;
     thread->entry_point = entry_point;
@@ -275,6 +273,8 @@ ResultVal<SharedPtr<Thread>> Thread::Create(std::string name, VAddr entry_point,
     thread->name = std::move(name);
     thread->callback_handle = wakeup_callback_handle_table.Create(thread).Unwrap();
     thread->owner_process = owner_process;
+    thread->scheduler = Core::System().GetInstance().Scheduler(static_cast<size_t>(processor_id));
+    thread->scheduler->AddThread(thread, priority);
 
     // Find the next available TLS index, and mark it as used
     auto& tls_slots = owner_process->tls_slots;
@@ -337,7 +337,7 @@ void Thread::SetPriority(u32 priority) {
 }
 
 void Thread::BoostPriority(u32 priority) {
-    Core::System::GetInstance().Scheduler().SetThreadPriority(this, priority);
+    scheduler->SetThreadPriority(this, priority);
     current_priority = priority;
 }
 
@@ -406,7 +406,7 @@ void Thread::UpdatePriority() {
     if (new_priority == current_priority)
         return;
 
-    Core::System::GetInstance().Scheduler().SetThreadPriority(this, new_priority);
+    scheduler->SetThreadPriority(this, new_priority);
 
     current_priority = new_priority;
 
@@ -421,7 +421,7 @@ void Thread::UpdatePriority() {
  * Gets the current thread
  */
 Thread* GetCurrentThread() {
-    return Core::System::GetInstance().Scheduler().GetCurrentThread();
+    return Core::System::GetInstance().CurrentScheduler().GetCurrentThread();
 }
 
 void ThreadingInit() {

src/core/hle/kernel/thread.h

@@ -4,6 +4,7 @@
 
 #pragma once
 
+#include <memory>
 #include <string>
 #include <unordered_map>
 #include <vector>
@@ -56,6 +57,7 @@ enum class ThreadWakeupReason {
 namespace Kernel {
 
 class Process;
+class Scheduler;
 
 class Thread final : public WaitObject {
 public:
@@ -240,6 +242,8 @@ public:
     // available. In case of a timeout, the object will be nullptr.
     std::function<WakeupCallback> wakeup_callback;
 
+    std::shared_ptr<Scheduler> scheduler;
+
 private:
     Thread();
     ~Thread() override;

src/core/hle/kernel/vm_manager.cpp

@@ -104,8 +104,15 @@ ResultVal<VMManager::VMAHandle> VMManager::MapMemoryBlock(VAddr target,
     VirtualMemoryArea& final_vma = vma_handle->second;
     ASSERT(final_vma.size == size);
 
-    Core::CPU().MapBackingMemory(target, size, block->data() + offset,
+    auto& system = Core::System::GetInstance();
-                                 VMAPermission::ReadWriteExecute);
+    system.ArmInterface(0).MapBackingMemory(target, size, block->data() + offset,
+                                            VMAPermission::ReadWriteExecute);
+    system.ArmInterface(1).MapBackingMemory(target, size, block->data() + offset,
+                                            VMAPermission::ReadWriteExecute);
+    system.ArmInterface(2).MapBackingMemory(target, size, block->data() + offset,
+                                            VMAPermission::ReadWriteExecute);
+    system.ArmInterface(3).MapBackingMemory(target, size, block->data() + offset,
+                                            VMAPermission::ReadWriteExecute);
 
     final_vma.type = VMAType::AllocatedMemoryBlock;
     final_vma.permissions = VMAPermission::ReadWrite;
@@ -126,7 +133,11 @@ ResultVal<VMManager::VMAHandle> VMManager::MapBackingMemory(VAddr target, u8* me
     VirtualMemoryArea& final_vma = vma_handle->second;
     ASSERT(final_vma.size == size);
 
-    Core::CPU().MapBackingMemory(target, size, memory, VMAPermission::ReadWriteExecute);
+    auto& system = Core::System::GetInstance();
+    system.ArmInterface(0).MapBackingMemory(target, size, memory, VMAPermission::ReadWriteExecute);
+    system.ArmInterface(1).MapBackingMemory(target, size, memory, VMAPermission::ReadWriteExecute);
+    system.ArmInterface(2).MapBackingMemory(target, size, memory, VMAPermission::ReadWriteExecute);
+    system.ArmInterface(3).MapBackingMemory(target, size, memory, VMAPermission::ReadWriteExecute);
 
     final_vma.type = VMAType::BackingMemory;
     final_vma.permissions = VMAPermission::ReadWrite;
@@ -184,7 +195,11 @@ ResultCode VMManager::UnmapRange(VAddr target, u64 size) {
 
     ASSERT(FindVMA(target)->second.size >= size);
 
-    Core::CPU().UnmapMemory(target, size);
+    auto& system = Core::System::GetInstance();
+    system.ArmInterface(0).UnmapMemory(target, size);
+    system.ArmInterface(1).UnmapMemory(target, size);
+    system.ArmInterface(2).UnmapMemory(target, size);
+    system.ArmInterface(3).UnmapMemory(target, size);
 
     return RESULT_SUCCESS;
 }

src/core/memory.cpp

@@ -28,8 +28,13 @@ static PageTable* current_page_table = nullptr;
 
 void SetCurrentPageTable(PageTable* page_table) {
     current_page_table = page_table;
-    if (Core::System::GetInstance().IsPoweredOn()) {
+
-        Core::CPU().PageTableChanged();
+    auto& system = Core::System::GetInstance();
+    if (system.IsPoweredOn()) {
+        system.ArmInterface(0).PageTableChanged();
+        system.ArmInterface(1).PageTableChanged();
+        system.ArmInterface(2).PageTableChanged();
+        system.ArmInterface(3).PageTableChanged();
     }
 }
 

src/yuzu/debugger/registers.cpp

@@ -63,7 +63,7 @@ void RegistersWidget::OnDebugModeEntered() {
 
     for (int i = 0; i < core_registers->childCount(); ++i)
         core_registers->child(i)->setText(
-            1, QString("0x%1").arg(Core::CPU().GetReg(i), 8, 16, QLatin1Char('0')));
+            1, QString("0x%1").arg(Core::CurrentArmInterface().GetReg(i), 8, 16, QLatin1Char('0')));
 
     UpdateCPSRValues();
 }
@@ -122,7 +122,7 @@ void RegistersWidget::CreateCPSRChildren() {
 }
 
 void RegistersWidget::UpdateCPSRValues() {
-    const u32 cpsr_val = Core::CPU().GetCPSR();
+    const u32 cpsr_val = Core::CurrentArmInterface().GetCPSR();
 
     cpsr->setText(1, QString("0x%1").arg(cpsr_val, 8, 16, QLatin1Char('0')));
     cpsr->child(0)->setText(

src/yuzu/debugger/wait_tree.cpp

@@ -51,7 +51,7 @@ std::size_t WaitTreeItem::Row() const {
 }
 
 std::vector<std::unique_ptr<WaitTreeThread>> WaitTreeItem::MakeThreadItemList() {
-    const auto& threads = Core::System::GetInstance().Scheduler().GetThreadList();
+    const auto& threads = Core::System::GetInstance().Scheduler(0)->GetThreadList();
     std::vector<std::unique_ptr<WaitTreeThread>> item_list;
     item_list.reserve(threads.size());
     for (std::size_t i = 0; i < threads.size(); ++i) {