VideoCommon/Fifo: Refactor to class, move to Core::System.

Source/Core/Core/Core.cpp

@@ -277,7 +277,9 @@ void Stop()  // - Hammertime!
   // Dump left over jobs
   HostDispatchJobs();
 
-  Fifo::EmulatorState(false);
+  auto& system = Core::System::GetInstance();
+
+  system.GetFifo().EmulatorState(false);
 
   INFO_LOG_FMT(CONSOLE, "Stop [Main Thread]\t\t---- Shutting down ----");
 
@@ -285,7 +287,7 @@ void Stop()  // - Hammertime!
   INFO_LOG_FMT(CONSOLE, "{}", StopMessage(true, "Stop CPU"));
   CPU::Stop();
 
-  if (Core::System::GetInstance().IsDualCoreMode())
+  if (system.IsDualCoreMode())
   {
     // Video_EnterLoop() should now exit so that EmuThread()
     // will continue concurrently with the rest of the commands
@@ -597,7 +599,7 @@ static void EmuThread(std::unique_ptr<BootParameters> boot, WindowSystemInfo wsi
     wiifs_guard.Dismiss();
 
   // This adds the SyncGPU handler to CoreTiming, so now CoreTiming::Advance might block.
-  Fifo::Prepare();
+  system.GetFifo().Prepare();
 
   // Setup our core
   if (Config::Get(Config::MAIN_CPU_CORE) != PowerPC::CPUCore::Interpreter)
@@ -622,7 +624,7 @@ static void EmuThread(std::unique_ptr<BootParameters> boot, WindowSystemInfo wsi
     s_cpu_thread = std::thread(cpuThreadFunc, savestate_path, delete_savestate);
 
     // become the GPU thread
-    Fifo::RunGpuLoop();
+    system.GetFifo().RunGpuLoop();
 
     // We have now exited the Video Loop
     INFO_LOG_FMT(CONSOLE, "{}", StopMessage(false, "Video Loop Ended"));
@@ -766,7 +768,8 @@ static bool PauseAndLock(bool do_lock, bool unpause_on_unlock)
 
   // video has to come after CPU, because CPU thread can wait for video thread
   // (s_efbAccessRequested).
-  Fifo::PauseAndLock(do_lock, false);
+  auto& system = Core::System::GetInstance();
+  system.GetFifo().PauseAndLock(do_lock, false);
 
   ResetRumble();
 
@@ -1029,7 +1032,10 @@ void UpdateWantDeterminism(bool initial)
       const auto ios = IOS::HLE::GetIOS();
       if (ios)
         ios->UpdateWantDeterminism(new_want_determinism);
-      Fifo::UpdateWantDeterminism(new_want_determinism);
+
+      auto& system = Core::System::GetInstance();
+      system.GetFifo().UpdateWantDeterminism(new_want_determinism);
+
       // We need to clear the cache because some parts of the JIT depend on want_determinism,
       // e.g. use of FMA.
       JitInterface::ClearCache();

Source/Core/Core/CoreTiming.cpp

@@ -354,7 +354,8 @@ void CoreTimingManager::Idle()
     // When the FIFO is processing data we must not advance because in this way
     // the VI will be desynchronized. So, We are waiting until the FIFO finish and
     // while we process only the events required by the FIFO.
-    Fifo::FlushGpu();
+    auto& system = Core::System::GetInstance();
+    system.GetFifo().FlushGpu();
   }
 
   PowerPC::UpdatePerformanceMonitor(PowerPC::ppcState.downcount, 0, 0);

Source/Core/Core/HW/CPU.cpp

@@ -191,7 +191,8 @@ void Run()
 static void RunAdjacentSystems(bool running)
 {
   // NOTE: We're assuming these will not try to call Break or EnableStepping.
-  Fifo::EmulatorState(running);
+  auto& system = Core::System::GetInstance();
+  system.GetFifo().EmulatorState(running);
   // Core is responsible for shutting down the sound stream.
   if (s_state != State::PowerDown)
     AudioCommon::SetSoundStreamRunning(Core::System::GetInstance(), running);

Source/Core/Core/HW/SystemTimers.cpp

@@ -173,7 +173,7 @@ void PatchEngineCallback(Core::System& system, u64 userdata, s64 cycles_late)
 void ThrottleCallback(Core::System& system, u64 deadline, s64 cyclesLate)
 {
   // Allow the GPU thread to sleep. Setting this flag here limits the wakeups to 1 kHz.
-  Fifo::GpuMaySleep();
+  system.GetFifo().GpuMaySleep();
 
   const u64 time = Common::Timer::NowUs();
 

Source/Core/Core/System.cpp

@@ -19,6 +19,7 @@
 #include "Core/HW/Sram.h"
 #include "Core/HW/VideoInterface.h"
 #include "VideoCommon/CommandProcessor.h"
+#include "VideoCommon/Fifo.h"
 
 namespace Core
 {
@@ -35,6 +36,7 @@ struct System::Impl
   DVDInterface::DVDInterfaceState m_dvd_interface_state;
   DVDThread::DVDThreadState m_dvd_thread_state;
   ExpansionInterface::ExpansionInterfaceState m_expansion_interface_state;
+  Fifo::FifoManager m_fifo;
   Memory::MemoryManager m_memory;
   MemoryInterface::MemoryInterfaceState m_memory_interface_state;
   SerialInterface::SerialInterfaceState m_serial_interface_state;
@@ -120,6 +122,11 @@ ExpansionInterface::ExpansionInterfaceState& System::GetExpansionInterfaceState(
   return m_impl->m_expansion_interface_state;
 }
 
+Fifo::FifoManager& System::GetFifo() const
+{
+  return m_impl->m_fifo;
+}
+
 Memory::MemoryManager& System::GetMemory() const
 {
   return m_impl->m_memory;

Source/Core/Core/System.h

@@ -36,6 +36,10 @@ namespace ExpansionInterface
 {
 class ExpansionInterfaceState;
 };
+namespace Fifo
+{
+class FifoManager;
+}
 namespace Memory
 {
 class MemoryManager;
@@ -94,6 +98,7 @@ public:
   DVDInterface::DVDInterfaceState& GetDVDInterfaceState() const;
   DVDThread::DVDThreadState& GetDVDThreadState() const;
   ExpansionInterface::ExpansionInterfaceState& GetExpansionInterfaceState() const;
+  Fifo::FifoManager& GetFifo() const;
   Memory::MemoryManager& GetMemory() const;
   MemoryInterface::MemoryInterfaceState& GetMemoryInterfaceState() const;
   SerialInterface::SerialInterfaceState& GetSerialInterfaceState() const;

Source/Core/DolphinQt/Host.cpp

@@ -114,9 +114,10 @@ static void RunWithGPUThreadInactive(std::function<void()> f)
     // the CPU and GPU threads are the same thread, and we already checked for the GPU thread.)
 
     const bool was_running = Core::GetState() == Core::State::Running;
-    Fifo::PauseAndLock(true, was_running);
+    auto& fifo = Core::System::GetInstance().GetFifo();
+    fifo.PauseAndLock(true, was_running);
     f();
-    Fifo::PauseAndLock(false, was_running);
+    fifo.PauseAndLock(false, was_running);
   }
   else
   {

Source/Core/VideoCommon/AsyncRequests.cpp

@@ -5,6 +5,7 @@
 
 #include <mutex>
 
+#include "Core/System.h"
 #include "VideoCommon/Fifo.h"
 #include "VideoCommon/RenderBase.h"
 #include "VideoCommon/Statistics.h"
@@ -90,7 +91,8 @@ void AsyncRequests::PushEvent(const AsyncRequests::Event& event, bool blocking)
 
   m_queue.push(event);
 
-  Fifo::RunGpu();
+  auto& system = Core::System::GetInstance();
+  system.GetFifo().RunGpu();
   if (blocking)
   {
     m_cond.wait(lock, [this] { return m_queue.empty(); });
@@ -159,7 +161,7 @@ void AsyncRequests::HandleEvent(const AsyncRequests::Event& e)
     break;
 
   case Event::FIFO_RESET:
-    Fifo::ResetVideoBuffer();
+    Core::System::GetInstance().GetFifo().ResetVideoBuffer();
     break;
 
   case Event::PERF_QUERY:

Source/Core/VideoCommon/BPStructs.cpp

@@ -179,14 +179,17 @@ static void BPWritten(const BPCmd& bp, int cycles_into_future)
     switch (bp.newvalue & 0xFF)
     {
     case 0x02:
+    {
       INCSTAT(g_stats.this_frame.num_draw_done);
       g_texture_cache->FlushEFBCopies();
       g_framebuffer_manager->InvalidatePeekCache(false);
       g_framebuffer_manager->RefreshPeekCache();
-      if (!Fifo::UseDeterministicGPUThread())
+      auto& system = Core::System::GetInstance();
+      if (!system.GetFifo().UseDeterministicGPUThread())
         PixelEngine::SetFinish(cycles_into_future);  // may generate interrupt
       DEBUG_LOG_FMT(VIDEO, "GXSetDrawDone SetPEFinish (value: {:#04X})", bp.newvalue & 0xFFFF);
       return;
+    }
 
     default:
       WARN_LOG_FMT(VIDEO, "GXSetDrawDone ??? (value {:#04X})", bp.newvalue & 0xFFFF);
@@ -194,23 +197,29 @@ static void BPWritten(const BPCmd& bp, int cycles_into_future)
     }
     return;
   case BPMEM_PE_TOKEN_ID:  // Pixel Engine Token ID
+  {
     INCSTAT(g_stats.this_frame.num_token);
     g_texture_cache->FlushEFBCopies();
     g_framebuffer_manager->InvalidatePeekCache(false);
     g_framebuffer_manager->RefreshPeekCache();
-    if (!Fifo::UseDeterministicGPUThread())
+    auto& system = Core::System::GetInstance();
+    if (!system.GetFifo().UseDeterministicGPUThread())
       PixelEngine::SetToken(static_cast<u16>(bp.newvalue & 0xFFFF), false, cycles_into_future);
     DEBUG_LOG_FMT(VIDEO, "SetPEToken {:#06X}", bp.newvalue & 0xFFFF);
     return;
+  }
   case BPMEM_PE_TOKEN_INT_ID:  // Pixel Engine Interrupt Token ID
+  {
     INCSTAT(g_stats.this_frame.num_token_int);
     g_texture_cache->FlushEFBCopies();
     g_framebuffer_manager->InvalidatePeekCache(false);
     g_framebuffer_manager->RefreshPeekCache();
-    if (!Fifo::UseDeterministicGPUThread())
+    auto& system = Core::System::GetInstance();
+    if (!system.GetFifo().UseDeterministicGPUThread())
       PixelEngine::SetToken(static_cast<u16>(bp.newvalue & 0xFFFF), true, cycles_into_future);
     DEBUG_LOG_FMT(VIDEO, "SetPEToken + INT {:#06X}", bp.newvalue & 0xFFFF);
     return;
+  }
 
   // ------------------------
   // EFB copy command. This copies a rectangle from the EFB to either RAM in a texture format or to

Source/Core/VideoCommon/CommandProcessor.cpp

@@ -223,8 +223,8 @@ void CommandProcessorManager::RegisterMMIO(Core::System& system, MMIO::Mapping*
 
   mmio->Register(base | STATUS_REGISTER, MMIO::ComplexRead<u16>([](Core::System& system, u32) {
                    auto& cp = system.GetCommandProcessor();
-                   Fifo::SyncGPUForRegisterAccess();
-                   cp.SetCpStatusRegister();
+                   system.GetFifo().SyncGPUForRegisterAccess();
+                   cp.SetCpStatusRegister(system);
                    return cp.m_cp_status_reg.Hex;
                  }),
                  MMIO::InvalidWrite<u16>());
@@ -234,8 +234,8 @@ void CommandProcessorManager::RegisterMMIO(Core::System& system, MMIO::Mapping*
                    auto& cp = system.GetCommandProcessor();
                    UCPCtrlReg tmp(val);
                    cp.m_cp_ctrl_reg.Hex = tmp.Hex;
-                   cp.SetCpControlRegister();
-                   Fifo::RunGpu();
+                   cp.SetCpControlRegister(system);
+                   system.GetFifo().RunGpu();
                  }));
 
   mmio->Register(base | CLEAR_REGISTER, MMIO::DirectRead<u16>(&m_cp_clear_reg.Hex),
@@ -244,7 +244,7 @@ void CommandProcessorManager::RegisterMMIO(Core::System& system, MMIO::Mapping*
                    UCPClearReg tmp(val);
                    cp.m_cp_clear_reg.Hex = tmp.Hex;
                    cp.SetCpClearRegister();
-                   Fifo::RunGpu();
+                   system.GetFifo().RunGpu();
                  }));
 
   mmio->Register(base | PERF_SELECT, MMIO::InvalidRead<u16>(), MMIO::Nop<u16>());
@@ -284,7 +284,7 @@ void CommandProcessorManager::RegisterMMIO(Core::System& system, MMIO::Mapping*
   {
     fifo_rw_distance_hi_r = MMIO::ComplexRead<u16>([](Core::System& system, u32) {
       const auto& fifo = system.GetCommandProcessor().GetFifo();
-      Fifo::SyncGPUForRegisterAccess();
+      system.GetFifo().SyncGPUForRegisterAccess();
       if (fifo.CPWritePointer.load(std::memory_order_relaxed) >=
           fifo.SafeCPReadPointer.load(std::memory_order_relaxed))
       {
@@ -306,16 +306,16 @@ void CommandProcessorManager::RegisterMMIO(Core::System& system, MMIO::Mapping*
   {
     fifo_rw_distance_hi_r = MMIO::ComplexRead<u16>([](Core::System& system, u32) {
       const auto& fifo = system.GetCommandProcessor().GetFifo();
-      Fifo::SyncGPUForRegisterAccess();
+      system.GetFifo().SyncGPUForRegisterAccess();
       return fifo.CPReadWriteDistance.load(std::memory_order_relaxed) >> 16;
     });
   }
   mmio->Register(base | FIFO_RW_DISTANCE_HI, fifo_rw_distance_hi_r,
                  MMIO::ComplexWrite<u16>([WMASK_HI_RESTRICT](Core::System& system, u32, u16 val) {
                    auto& fifo = system.GetCommandProcessor().GetFifo();
-                   Fifo::SyncGPUForRegisterAccess();
+                   system.GetFifo().SyncGPUForRegisterAccess();
                    WriteHigh(fifo.CPReadWriteDistance, val & WMASK_HI_RESTRICT);
-                   Fifo::RunGpu();
+                   system.GetFifo().RunGpu();
                  }));
 
   mmio->Register(
@@ -330,12 +330,12 @@ void CommandProcessorManager::RegisterMMIO(Core::System& system, MMIO::Mapping*
   {
     fifo_read_hi_r = MMIO::ComplexRead<u16>([](Core::System& system, u32) {
       auto& fifo = system.GetCommandProcessor().GetFifo();
-      Fifo::SyncGPUForRegisterAccess();
+      system.GetFifo().SyncGPUForRegisterAccess();
       return fifo.SafeCPReadPointer.load(std::memory_order_relaxed) >> 16;
     });
     fifo_read_hi_w = MMIO::ComplexWrite<u16>([WMASK_HI_RESTRICT](Core::System& sys, u32, u16 val) {
       auto& fifo = sys.GetCommandProcessor().GetFifo();
-      Fifo::SyncGPUForRegisterAccess();
+      sys.GetFifo().SyncGPUForRegisterAccess();
       WriteHigh(fifo.CPReadPointer, val & WMASK_HI_RESTRICT);
       fifo.SafeCPReadPointer.store(fifo.CPReadPointer.load(std::memory_order_relaxed),
                                    std::memory_order_relaxed);
@@ -345,12 +345,12 @@ void CommandProcessorManager::RegisterMMIO(Core::System& system, MMIO::Mapping*
   {
     fifo_read_hi_r = MMIO::ComplexRead<u16>([](Core::System& system, u32) {
       const auto& fifo = system.GetCommandProcessor().GetFifo();
-      Fifo::SyncGPUForRegisterAccess();
+      system.GetFifo().SyncGPUForRegisterAccess();
       return fifo.CPReadPointer.load(std::memory_order_relaxed) >> 16;
     });
     fifo_read_hi_w = MMIO::ComplexWrite<u16>([WMASK_HI_RESTRICT](Core::System& sys, u32, u16 val) {
       auto& fifo = sys.GetCommandProcessor().GetFifo();
-      Fifo::SyncGPUForRegisterAccess();
+      sys.GetFifo().SyncGPUForRegisterAccess();
       WriteHigh(fifo.CPReadPointer, val & WMASK_HI_RESTRICT);
     });
   }
@@ -366,7 +366,7 @@ void CommandProcessorManager::GatherPipeBursted(Core::System& system)
   // if we aren't linked, we don't care about gather pipe data
   if (!m_cp_ctrl_reg.GPLinkEnable)
   {
-    if (IsOnThread(system) && !Fifo::UseDeterministicGPUThread())
+    if (IsOnThread(system) && !system.GetFifo().UseDeterministicGPUThread())
     {
       // In multibuffer mode is not allowed write in the same FIFO attached to the GPU.
       // Fix Pokemon XD in DC mode.
@@ -374,10 +374,10 @@ void CommandProcessorManager::GatherPipeBursted(Core::System& system)
           (ProcessorInterface::Fifo_CPUBase == fifo.CPBase.load(std::memory_order_relaxed)) &&
           fifo.CPReadWriteDistance.load(std::memory_order_relaxed) > 0)
       {
-        Fifo::FlushGpu();
+        system.GetFifo().FlushGpu();
       }
     }
-    Fifo::RunGpu();
+    system.GetFifo().RunGpu();
     return;
   }
 
@@ -405,7 +405,7 @@ void CommandProcessorManager::GatherPipeBursted(Core::System& system)
 
   fifo.CPReadWriteDistance.fetch_add(GPFifo::GATHER_PIPE_SIZE, std::memory_order_seq_cst);
 
-  Fifo::RunGpu();
+  system.GetFifo().RunGpu();
 
   ASSERT_MSG(COMMANDPROCESSOR,
              fifo.CPReadWriteDistance.load(std::memory_order_relaxed) <=
@@ -442,12 +442,12 @@ void CommandProcessorManager::UpdateInterrupts(Core::System& system, u64 userdat
   }
   system.GetCoreTiming().ForceExceptionCheck(0);
   m_interrupt_waiting.Clear();
-  Fifo::RunGpu();
+  system.GetFifo().RunGpu();
 }
 
 void CommandProcessorManager::UpdateInterruptsFromVideoBackend(Core::System& system, u64 userdata)
 {
-  if (!Fifo::UseDeterministicGPUThread())
+  if (!system.GetFifo().UseDeterministicGPUThread())
   {
     system.GetCoreTiming().ScheduleEvent(0, m_event_type_update_interrupts, userdata,
                                          CoreTiming::FromThread::NON_CPU);
@@ -573,7 +573,7 @@ void CommandProcessorManager::SetCPStatusFromCPU(Core::System& system)
   }
 }
 
-void CommandProcessorManager::SetCpStatusRegister()
+void CommandProcessorManager::SetCpStatusRegister(Core::System& system)
 {
   const auto& fifo = m_fifo;
 
@@ -583,7 +583,7 @@ void CommandProcessorManager::SetCpStatusRegister()
                              (fifo.CPReadPointer.load(std::memory_order_relaxed) ==
                               fifo.CPWritePointer.load(std::memory_order_relaxed));
   m_cp_status_reg.CommandIdle = !fifo.CPReadWriteDistance.load(std::memory_order_relaxed) ||
-                                Fifo::AtBreakpoint() ||
+                                system.GetFifo().AtBreakpoint() ||
                                 !fifo.bFF_GPReadEnable.load(std::memory_order_relaxed);
   m_cp_status_reg.UnderflowLoWatermark = fifo.bFF_LoWatermark.load(std::memory_order_relaxed);
   m_cp_status_reg.OverflowHiWatermark = fifo.bFF_HiWatermark.load(std::memory_order_relaxed);
@@ -597,7 +597,7 @@ void CommandProcessorManager::SetCpStatusRegister()
                 m_cp_status_reg.UnderflowLoWatermark ? "ON" : "OFF");
 }
 
-void CommandProcessorManager::SetCpControlRegister()
+void CommandProcessorManager::SetCpControlRegister(Core::System& system)
 {
   auto& fifo = m_fifo;
 
@@ -610,7 +610,7 @@ void CommandProcessorManager::SetCpControlRegister()
   if (fifo.bFF_GPReadEnable.load(std::memory_order_relaxed) && !m_cp_ctrl_reg.GPReadEnable)
   {
     fifo.bFF_GPReadEnable.store(m_cp_ctrl_reg.GPReadEnable, std::memory_order_relaxed);
-    Fifo::FlushGpu();
+    system.GetFifo().FlushGpu();
   }
   else
   {

Source/Core/VideoCommon/CommandProcessor.h

@@ -174,8 +174,8 @@ public:
   bool IsInterruptWaiting() const;
 
   void SetCpClearRegister();
-  void SetCpControlRegister();
-  void SetCpStatusRegister();
+  void SetCpControlRegister(Core::System& system);
+  void SetCpStatusRegister(Core::System& system);
 
   void HandleUnknownOpcode(u8 cmd_byte, const u8* buffer, bool preprocess);
 

Source/Core/VideoCommon/Fifo.h

@@ -3,21 +3,28 @@
 
 #pragma once
 
+#include <atomic>
 #include <cstddef>
+#include <optional>
+
+#include "Common/BlockingLoop.h"
 #include "Common/CommonTypes.h"
+#include "Common/Event.h"
+#include "Common/Flag.h"
 
 class PointerWrap;
 
+namespace Core
+{
+class System;
+}
+namespace CoreTiming
+{
+struct EventType;
+}
+
 namespace Fifo
 {
-void Init();
-void Shutdown();
-void Prepare();  // Must be called from the CPU thread.
-void DoState(PointerWrap& f);
-void PauseAndLock(bool doLock, bool unpauseOnUnlock);
-void UpdateWantDeterminism(bool want);
-bool UseDeterministicGPUThread();
-
 // Used for diagnostics.
 enum class SyncGPUReason
 {
@@ -29,23 +36,95 @@ enum class SyncGPUReason
   Swap,
   AuxSpace,
 };
-// In deterministic GPU thread mode this waits for the GPU to be done with pending work.
-void SyncGPU(SyncGPUReason reason, bool may_move_read_ptr = true);
 
-// In single core mode, this runs the GPU for a single slice.
-// In dual core mode, this synchronizes with the GPU thread.
-void SyncGPUForRegisterAccess();
+class FifoManager final
+{
+public:
+  FifoManager();
+  FifoManager(const FifoManager& other) = delete;
+  FifoManager(FifoManager&& other) = delete;
+  FifoManager& operator=(const FifoManager& other) = delete;
+  FifoManager& operator=(FifoManager&& other) = delete;
+  ~FifoManager();
 
-void PushFifoAuxBuffer(const void* ptr, size_t size);
-void* PopFifoAuxBuffer(size_t size);
+  void Init();
+  void Shutdown();
+  void Prepare();  // Must be called from the CPU thread.
+  void DoState(PointerWrap& f);
+  void PauseAndLock(bool doLock, bool unpauseOnUnlock);
+  void UpdateWantDeterminism(bool want);
+  bool UseDeterministicGPUThread() const { return m_use_deterministic_gpu_thread; }
 
-void FlushGpu();
-void RunGpu();
-void GpuMaySleep();
-void RunGpuLoop();
-void ExitGpuLoop();
-void EmulatorState(bool running);
-bool AtBreakpoint();
-void ResetVideoBuffer();
+  // In deterministic GPU thread mode this waits for the GPU to be done with pending work.
+  void SyncGPU(SyncGPUReason reason, bool may_move_read_ptr = true);
 
+  // In single core mode, this runs the GPU for a single slice.
+  // In dual core mode, this synchronizes with the GPU thread.
+  void SyncGPUForRegisterAccess();
+
+  void PushFifoAuxBuffer(const void* ptr, size_t size);
+  void* PopFifoAuxBuffer(size_t size);
+
+  void FlushGpu();
+  void RunGpu();
+  void GpuMaySleep();
+  void RunGpuLoop();
+  void ExitGpuLoop();
+  void EmulatorState(bool running);
+  bool AtBreakpoint() const;
+  void ResetVideoBuffer();
+
+private:
+  void RefreshConfig();
+  void ReadDataFromFifo(u32 readPtr);
+  void ReadDataFromFifoOnCPU(u32 readPtr);
+  int RunGpuOnCpu(int ticks);
+  int WaitForGpuThread(int ticks);
+  static void SyncGPUCallback(Core::System& system, u64 ticks, s64 cyclesLate);
+
+  static constexpr u32 FIFO_SIZE = 2 * 1024 * 1024;
+
+  Common::BlockingLoop m_gpu_mainloop;
+
+  Common::Flag m_emu_running_state;
+
+  // Most of this array is unlikely to be faulted in...
+  u8 m_fifo_aux_data[FIFO_SIZE]{};
+  u8* m_fifo_aux_write_ptr = nullptr;
+  u8* m_fifo_aux_read_ptr = nullptr;
+
+  // This could be in SConfig, but it depends on multiple settings
+  // and can change at runtime.
+  bool m_use_deterministic_gpu_thread = false;
+
+  CoreTiming::EventType* m_event_sync_gpu = nullptr;
+
+  // STATE_TO_SAVE
+  u8* m_video_buffer = nullptr;
+  u8* m_video_buffer_read_ptr = nullptr;
+  std::atomic<u8*> m_video_buffer_write_ptr = nullptr;
+  std::atomic<u8*> m_video_buffer_seen_ptr = nullptr;
+  u8* m_video_buffer_pp_read_ptr = nullptr;
+  // The read_ptr is always owned by the GPU thread.  In normal mode, so is the
+  // write_ptr, despite it being atomic.  In deterministic GPU thread mode,
+  // things get a bit more complicated:
+  // - The seen_ptr is written by the GPU thread, and points to what it's already
+  // processed as much of as possible - in the case of a partial command which
+  // caused it to stop, not the same as the read ptr.  It's written by the GPU,
+  // under the lock, and updating the cond.
+  // - The write_ptr is written by the CPU thread after it copies data from the
+  // FIFO.  Maybe someday it will be under the lock.  For now, because RunGpuLoop
+  // polls, it's just atomic.
+  // - The pp_read_ptr is the CPU preprocessing version of the read_ptr.
+
+  std::atomic<int> m_sync_ticks = 0;
+  bool m_syncing_suspended = false;
+  Common::Event m_sync_wakeup_event;
+
+  std::optional<size_t> m_config_callback_id = std::nullopt;
+  bool m_config_sync_gpu = false;
+  int m_config_sync_gpu_max_distance = 0;
+  int m_config_sync_gpu_min_distance = 0;
+  float m_config_sync_gpu_overclock = 0.0f;
+};
 }  // namespace Fifo

Source/Core/VideoCommon/OpcodeDecoding.cpp

@@ -151,13 +151,14 @@ public:
     {
       m_in_display_list = true;
 
+      auto& system = Core::System::GetInstance();
+
       if constexpr (is_preprocess)
       {
-        auto& system = Core::System::GetInstance();
         auto& memory = system.GetMemory();
         const u8* const start_address = memory.GetPointer(address);
 
-        Fifo::PushFifoAuxBuffer(start_address, size);
+        system.GetFifo().PushFifoAuxBuffer(start_address, size);
 
         if (start_address != nullptr)
         {
@@ -168,13 +169,13 @@ public:
       {
         const u8* start_address;
 
-        if (Fifo::UseDeterministicGPUThread())
+        auto& fifo = system.GetFifo();
+        if (fifo.UseDeterministicGPUThread())
         {
-          start_address = static_cast<u8*>(Fifo::PopFifoAuxBuffer(size));
+          start_address = static_cast<u8*>(fifo.PopFifoAuxBuffer(size));
         }
         else
         {
-          auto& system = Core::System::GetInstance();
           auto& memory = system.GetMemory();
           start_address = memory.GetPointer(address);
         }

Source/Core/VideoCommon/PixelEngine.cpp

@@ -333,7 +333,8 @@ static void RaiseEvent(int cycles_into_future)
 
   CoreTiming::FromThread from = CoreTiming::FromThread::NON_CPU;
   s64 cycles = 0;  // we don't care about timings for dual core mode.
-  if (!Core::System::GetInstance().IsDualCoreMode() || Fifo::UseDeterministicGPUThread())
+  auto& system = Core::System::GetInstance();
+  if (!system.IsDualCoreMode() || system.GetFifo().UseDeterministicGPUThread())
   {
     from = CoreTiming::FromThread::CPU;
 

Source/Core/VideoCommon/VideoBackendBase.cpp

@@ -83,7 +83,8 @@ std::string VideoBackendBase::BadShaderFilename(const char* shader_stage, int co
 
 void VideoBackendBase::Video_ExitLoop()
 {
-  Fifo::ExitGpuLoop();
+  auto& system = Core::System::GetInstance();
+  system.GetFifo().ExitGpuLoop();
 }
 
 // Run from the CPU thread (from VideoInterface.cpp)
@@ -92,7 +93,8 @@ void VideoBackendBase::Video_OutputXFB(u32 xfb_addr, u32 fb_width, u32 fb_stride
 {
   if (m_initialized && g_renderer && !g_ActiveConfig.bImmediateXFB)
   {
-    Fifo::SyncGPU(Fifo::SyncGPUReason::Swap);
+    auto& system = Core::System::GetInstance();
+    system.GetFifo().SyncGPU(Fifo::SyncGPUReason::Swap);
 
     AsyncRequests::Event e;
     e.time = ticks;
@@ -147,7 +149,8 @@ u32 VideoBackendBase::Video_GetQueryResult(PerfQueryType type)
     return 0;
   }
 
-  Fifo::SyncGPU(Fifo::SyncGPUReason::PerfQuery);
+  auto& system = Core::System::GetInstance();
+  system.GetFifo().SyncGPU(Fifo::SyncGPUReason::PerfQuery);
 
   AsyncRequests::Event e;
   e.time = 0;
@@ -185,7 +188,8 @@ u16 VideoBackendBase::Video_GetBoundingBox(int index)
     warn_once = false;
   }
 
-  Fifo::SyncGPU(Fifo::SyncGPUReason::BBox);
+  auto& system = Core::System::GetInstance();
+  system.GetFifo().SyncGPU(Fifo::SyncGPUReason::BBox);
 
   AsyncRequests::Event e;
   u16 result;
@@ -291,7 +295,8 @@ void VideoBackendBase::PopulateBackendInfoFromUI()
 
 void VideoBackendBase::DoState(PointerWrap& p)
 {
-  if (!Core::System::GetInstance().IsDualCoreMode())
+  auto& system = Core::System::GetInstance();
+  if (!system.IsDualCoreMode())
   {
     VideoCommon_DoState(p);
     return;
@@ -304,7 +309,7 @@ void VideoBackendBase::DoState(PointerWrap& p)
 
   // Let the GPU thread sleep after loading the state, so we're not spinning if paused after loading
   // a state. The next GP burst will wake it up again.
-  Fifo::GpuMaySleep();
+  system.GetFifo().GpuMaySleep();
 }
 
 void VideoBackendBase::InitializeShared()
@@ -319,7 +324,7 @@ void VideoBackendBase::InitializeShared()
   auto& system = Core::System::GetInstance();
   auto& command_processor = system.GetCommandProcessor();
   command_processor.Init(system);
-  Fifo::Init();
+  system.GetFifo().Init();
   PixelEngine::Init();
   BPInit();
   VertexLoaderManager::Init();
@@ -336,6 +341,7 @@ void VideoBackendBase::ShutdownShared()
 {
   m_initialized = false;
 
+  auto& system = Core::System::GetInstance();
   VertexLoaderManager::Clear();
-  Fifo::Shutdown();
+  system.GetFifo().Shutdown();
 }

Source/Core/VideoCommon/VideoState.cpp

@@ -60,10 +60,10 @@ void VideoCommon_DoState(PointerWrap& p)
   p.DoMarker("TMEM");
 
   // FIFO
-  Fifo::DoState(p);
+  auto& system = Core::System::GetInstance();
+  system.GetFifo().DoState(p);
   p.DoMarker("Fifo");
 
-  auto& system = Core::System::GetInstance();
   auto& command_processor = system.GetCommandProcessor();
   command_processor.DoState(p);
   p.DoMarker("CommandProcessor");

Source/Core/VideoCommon/XFStructs.cpp

@@ -257,13 +257,14 @@ void LoadIndexedXF(CPArray array, u32 index, u16 address, u8 size)
 
   u32* currData = (u32*)(&xfmem) + address;
   u32* newData;
-  if (Fifo::UseDeterministicGPUThread())
+  auto& system = Core::System::GetInstance();
+  auto& fifo = system.GetFifo();
+  if (fifo.UseDeterministicGPUThread())
   {
-    newData = (u32*)Fifo::PopFifoAuxBuffer(size * sizeof(u32));
+    newData = (u32*)fifo.PopFifoAuxBuffer(size * sizeof(u32));
   }
   else
   {
-    auto& system = Core::System::GetInstance();
     auto& memory = system.GetMemory();
     newData = (u32*)memory.GetPointer(g_main_cp_state.array_bases[array] +
                                       g_main_cp_state.array_strides[array] * index);
@@ -293,7 +294,7 @@ void PreprocessIndexedXF(CPArray array, u32 index, u16 address, u8 size)
                                          g_preprocess_cp_state.array_strides[array] * index);
 
   const size_t buf_size = size * sizeof(u32);
-  Fifo::PushFifoAuxBuffer(new_data, buf_size);
+  system.GetFifo().PushFifoAuxBuffer(new_data, buf_size);
 }
 
 std::pair<std::string, std::string> GetXFRegInfo(u32 address, u32 value)