Implement lazy flush-on-read for Buffers (SSBO/Copy) (#1790)

* Initial implementation of buffer flush (VERY WIP)

* Host shaders need to be rebuilt for the SSBO write flag.

* New approach with reserved regions and gl sync

* Fix a ton of buffer issues.

* Remove unused buffer unmapped behaviour

* Revert "Remove unused buffer unmapped behaviour"

This reverts commit f1700e52fb8760180ac5e0987a07d409d1e70ece.

* Delete modified ranges on unmap

Fixes potential crashes in Super Smash Bros, where a previously modified range could lie on either side of an unmap.

* Cache some more delegates.

* Dispose Sync on Close

* Also create host sync for GPFifo syncpoint increment.

* Copy buffer optimization, add docs

* Fix race condition with OpenGL Sync

* Enable read tracking on CommandBuffer, insert syncpoint on WaitForIdle

* Performance: Only flush individual pages of SSBO at a time

This avoids flushing large amounts of data when only a small amount is actually used.

* Signal Modified rather than flushing after clear

* Fix some docs and code style.

* Introduce a new test for tracking memory protection.

Sucessfully demonstrates that the bug causing write protection to be cleared by a read action has been fixed. (these tests fail on master)

* Address Comments

* Add host sync for SetReference

This ensures that any indirect draws will correctly flush any related buffer data written before them. Fixes some flashing and misplaced world geometry in MH rise.

* Make PageAlign static

* Re-enable read tracking, for reads.

Ryujinx.Cpu/MemoryManager.cs

@@ -131,7 +131,7 @@ namespace Ryujinx.Cpu
         /// <exception cref="InvalidMemoryRegionException">Throw for unhandled invalid or unmapped memory accesses</exception>
         public T Read<T>(ulong va) where T : unmanaged
         {
-            return MemoryMarshal.Cast<byte, T>(GetSpan(va, Unsafe.SizeOf<T>()))[0];
+            return MemoryMarshal.Cast<byte, T>(GetSpan(va, Unsafe.SizeOf<T>(), true))[0];
         }
 
         /// <summary>

Ryujinx.Cpu/Tracking/CpuMultiRegionHandle.cs

@@ -18,6 +18,7 @@ namespace Ryujinx.Cpu.Tracking
         public void QueryModified(Action<ulong, ulong> modifiedAction) => _impl.QueryModified(modifiedAction);
         public void QueryModified(ulong address, ulong size, Action<ulong, ulong> modifiedAction) => _impl.QueryModified(address, size, modifiedAction);
         public void QueryModified(ulong address, ulong size, Action<ulong, ulong> modifiedAction, int sequenceNumber) => _impl.QueryModified(address, size, modifiedAction, sequenceNumber);
+        public void RegisterAction(ulong address, ulong size, RegionSignal action) => _impl.RegisterAction(address, size, action);
         public void SignalWrite() => _impl.SignalWrite();
     }
 }

Ryujinx.Cpu/Tracking/CpuSmartMultiRegionHandle.cs

@@ -15,6 +15,7 @@ namespace Ryujinx.Cpu.Tracking
         }
 
         public void Dispose() => _impl.Dispose();
+        public void RegisterAction(RegionSignal action) => _impl.RegisterAction(action);
         public void QueryModified(Action<ulong, ulong> modifiedAction) => _impl.QueryModified(modifiedAction);
         public void QueryModified(ulong address, ulong size, Action<ulong, ulong> modifiedAction) => _impl.QueryModified(address, size, modifiedAction);
         public void QueryModified(ulong address, ulong size, Action<ulong, ulong> modifiedAction, int sequenceNumber) => _impl.QueryModified(address, size, modifiedAction, sequenceNumber);

Ryujinx.Graphics.GAL/IRenderer.cs

@@ -21,6 +21,8 @@ namespace Ryujinx.Graphics.GAL
         ISampler CreateSampler(SamplerCreateInfo info);
         ITexture CreateTexture(TextureCreateInfo info, float scale);
 
+        void CreateSync(ulong id);
+
         void DeleteBuffer(BufferHandle buffer);
 
         byte[] GetBufferData(BufferHandle buffer, int offset, int size);
@@ -39,6 +41,8 @@ namespace Ryujinx.Graphics.GAL
 
         void ResetCounter(CounterType type);
 
+        void WaitSync(ulong id);
+
         void Initialize(GraphicsDebugLevel logLevel);
     }
 }

Ryujinx.Graphics.Gpu/Engine/Compute.cs

@@ -97,7 +97,7 @@ namespace Ryujinx.Graphics.Gpu.Engine
 
                 SbDescriptor sbDescriptor = _context.PhysicalMemory.Read<SbDescriptor>(sbDescAddress);
 
-                BufferManager.SetComputeStorageBuffer(sb.Slot, sbDescriptor.PackAddress(), (uint)sbDescriptor.Size);
+                BufferManager.SetComputeStorageBuffer(sb.Slot, sbDescriptor.PackAddress(), (uint)sbDescriptor.Size, sb.Flags);
             }
 
             BufferManager.SetComputeStorageBufferBindings(info.SBuffers);

Ryujinx.Graphics.Gpu/Engine/GPFifo/GPFifoClass.cs

@@ -39,6 +39,7 @@ namespace Ryujinx.Graphics.Gpu.Engine.GPFifo
                 { nameof(GPFifoClassState.Semaphored), new RwCallback(Semaphored, null) },
                 { nameof(GPFifoClassState.Syncpointb), new RwCallback(Syncpointb, null) },
                 { nameof(GPFifoClassState.WaitForIdle), new RwCallback(WaitForIdle, null) },
+                { nameof(GPFifoClassState.SetReference), new RwCallback(SetReference, null) },
                 { nameof(GPFifoClassState.LoadMmeInstructionRam), new RwCallback(LoadMmeInstructionRam, null) },
                 { nameof(GPFifoClassState.LoadMmeStartAddressRam), new RwCallback(LoadMmeStartAddressRam, null) },
                 { nameof(GPFifoClassState.SetMmeShadowRamControl), new RwCallback(SetMmeShadowRamControl, null) }
@@ -136,6 +137,7 @@ namespace Ryujinx.Graphics.Gpu.Engine.GPFifo
             }
             else if (operation == SyncpointbOperation.Incr)
             {
+                _context.CreateHostSyncIfNeeded();
                 _context.Synchronization.IncrementSyncpoint(syncpointId);
             }
 
@@ -150,6 +152,17 @@ namespace Ryujinx.Graphics.Gpu.Engine.GPFifo
         {
             _context.Methods.PerformDeferredDraws();
             _context.Renderer.Pipeline.Barrier();
+
+            _context.CreateHostSyncIfNeeded();
+        }
+
+        /// <summary>
+        /// Used as an indirect data barrier on NVN. When used, access to previously written data must be coherent.
+        /// </summary>
+        /// <param name="argument">Method call argument</param>
+        public void SetReference(int argument)
+        {
+            _context.CreateHostSyncIfNeeded();
         }
 
         /// <summary>

Ryujinx.Graphics.Gpu/Engine/GPFifo/GPFifoDevice.cs

@@ -52,7 +52,7 @@ namespace Ryujinx.Graphics.Gpu.Engine.GPFifo
             {
                 if (Words == null)
                 {
-                    Words = MemoryMarshal.Cast<byte, int>(context.MemoryManager.GetSpan(EntryAddress, (int)EntryCount * 4)).ToArray();
+                    Words = MemoryMarshal.Cast<byte, int>(context.MemoryManager.GetSpan(EntryAddress, (int)EntryCount * 4, true)).ToArray();
                 }
             }
         }

Ryujinx.Graphics.Gpu/Engine/MethodIncrementSyncpoint.cs

@@ -13,6 +13,7 @@ namespace Ryujinx.Graphics.Gpu.Engine
         {
             uint syncpointId = (uint)(argument) & 0xFFFF;
 
+            _context.CreateHostSyncIfNeeded();
             _context.Renderer.UpdateCounters(); // Poll the query counters, the game may want an updated result.
             _context.Synchronization.IncrementSyncpoint(syncpointId);
         }

Ryujinx.Graphics.Gpu/Engine/Methods.cs

@@ -61,6 +61,7 @@ namespace Ryujinx.Graphics.Gpu.Engine
 
             context.MemoryManager.MemoryUnmapped += _counterCache.MemoryUnmappedHandler;
             context.MemoryManager.MemoryUnmapped += TextureManager.MemoryUnmappedHandler;
+            context.MemoryManager.MemoryUnmapped += BufferManager.MemoryUnmappedHandler;
         }
 
         /// <summary>
@@ -333,7 +334,7 @@ namespace Ryujinx.Graphics.Gpu.Engine
 
                     SbDescriptor sbDescriptor = _context.PhysicalMemory.Read<SbDescriptor>(sbDescAddress);
 
-                    BufferManager.SetGraphicsStorageBuffer(stage, sb.Slot, sbDescriptor.PackAddress(), (uint)sbDescriptor.Size);
+                    BufferManager.SetGraphicsStorageBuffer(stage, sb.Slot, sbDescriptor.PackAddress(), (uint)sbDescriptor.Size, sb.Flags);
                 }
             }
         }

Ryujinx.Graphics.Gpu/GpuContext.cs

@@ -4,6 +4,7 @@ using Ryujinx.Graphics.Gpu.Engine.GPFifo;
 using Ryujinx.Graphics.Gpu.Memory;
 using Ryujinx.Graphics.Gpu.Synchronization;
 using System;
+using System.Collections.Generic;
 using System.Threading;
 
 namespace Ryujinx.Graphics.Gpu
@@ -59,6 +60,18 @@ namespace Ryujinx.Graphics.Gpu
         /// </summary>
         internal int SequenceNumber { get; private set; }
 
+        /// <summary>
+        /// Internal sync number, used to denote points at which host synchronization can be requested.
+        /// </summary>
+        internal ulong SyncNumber { get; private set; }
+
+        /// <summary>
+        /// Actions to be performed when a CPU waiting sync point is triggered.
+        /// If there are more than 0 items when this happens, a host sync object will be generated for the given <see cref="SyncNumber"/>,
+        /// and the SyncNumber will be incremented.
+        /// </summary>
+        internal List<Action> SyncActions { get; }
+
         private readonly Lazy<Capabilities> _caps;
 
         /// <summary>
@@ -87,6 +100,8 @@ namespace Ryujinx.Graphics.Gpu
             _caps = new Lazy<Capabilities>(Renderer.GetCapabilities);
 
             HostInitalized = new ManualResetEvent(false);
+
+            SyncActions = new List<Action>();
         }
 
         /// <summary>
@@ -118,6 +133,37 @@ namespace Ryujinx.Graphics.Gpu
             PhysicalMemory = new PhysicalMemory(cpuMemory);
         }
 
+        /// <summary>
+        /// Registers an action to be performed the next time a syncpoint is incremented.
+        /// This will also ensure a host sync object is created, and <see cref="SyncNumber"/> is incremented.
+        /// </summary>
+        /// <param name="action">The action to be performed on sync object creation</param>
+        public void RegisterSyncAction(Action action)
+        {
+            SyncActions.Add(action);
+        }
+
+        /// <summary>
+        /// Creates a host sync object if there are any pending sync actions. The actions will then be called.
+        /// If no actions are present, a host sync object is not created.
+        /// </summary>
+        public void CreateHostSyncIfNeeded()
+        {
+            if (SyncActions.Count > 0)
+            {
+                Renderer.CreateSync(SyncNumber);
+
+                SyncNumber++;
+
+                foreach (Action action in SyncActions)
+                {
+                    action();
+                }
+
+                SyncActions.Clear();
+            }
+        }
+
         /// <summary>
         /// Disposes all GPU resources currently cached.
         /// It's an error to push any GPU commands after disposal.

Ryujinx.Graphics.Gpu/Memory/Buffer.cs

@@ -1,6 +1,7 @@
 using Ryujinx.Cpu.Tracking;
 using Ryujinx.Graphics.GAL;
 using Ryujinx.Memory.Range;
+using Ryujinx.Memory.Tracking;
 using System;
 
 namespace Ryujinx.Graphics.Gpu.Memory
@@ -34,12 +35,28 @@ namespace Ryujinx.Graphics.Gpu.Memory
         /// </summary>
         public ulong EndAddress => Address + Size;
 
+        /// <summary>
+        /// Ranges of the buffer that have been modified on the GPU.
+        /// Ranges defined here cannot be updated from CPU until a CPU waiting sync point is reached.
+        /// Then, write tracking will signal, wait for GPU sync (generated at the syncpoint) and flush these regions.
+        /// </summary>
+        /// <remarks>
+        /// This is null until at least one modification occurs.
+        /// </remarks>
+        private BufferModifiedRangeList _modifiedRanges = null;
+
         private CpuMultiRegionHandle _memoryTrackingGranular;
+
         private CpuRegionHandle _memoryTracking;
+
+        private readonly RegionSignal _externalFlushDelegate;
+        private readonly Action<ulong, ulong> _loadDelegate;
         private readonly Action<ulong, ulong> _modifiedDelegate;
+
         private int _sequenceNumber;
 
         private bool _useGranular;
+        private bool _syncActionRegistered;
 
         /// <summary>
         /// Creates a new instance of the buffer.
@@ -66,6 +83,8 @@ namespace Ryujinx.Graphics.Gpu.Memory
                 _memoryTracking = context.PhysicalMemory.BeginTracking(address, size);
             }
 
+            _externalFlushDelegate = new RegionSignal(ExternalFlush);
+            _loadDelegate = new Action<ulong, ulong>(LoadRegion);
             _modifiedDelegate = new Action<ulong, ulong>(RegionModified);
         }
 
@@ -116,12 +135,131 @@ namespace Ryujinx.Graphics.Gpu.Memory
                 if (_memoryTracking.Dirty && _context.SequenceNumber != _sequenceNumber)
                 {
                     _memoryTracking.Reprotect();
+
+                    if (_modifiedRanges != null)
+                    {
+                        _modifiedRanges.ExcludeModifiedRegions(Address, Size, _loadDelegate);
+                    }
+                    else
+                    {
                         _context.Renderer.SetBufferData(Handle, 0, _context.PhysicalMemory.GetSpan(Address, (int)Size));
+                    }
+                    
                     _sequenceNumber = _context.SequenceNumber;
                 }
             }
         }
 
+        /// <summary>
+        /// Ensure that the modified range list exists.
+        /// </summary>
+        private void EnsureRangeList()
+        {
+            if (_modifiedRanges == null)
+            {
+                _modifiedRanges = new BufferModifiedRangeList(_context);
+            }
+        }
+
+        /// <summary>
+        /// Signal that the given region of the buffer has been modified.
+        /// </summary>
+        /// <param name="address">The start address of the modified region</param>
+        /// <param name="size">The size of the modified region</param>
+        public void SignalModified(ulong address, ulong size)
+        {
+            EnsureRangeList();
+
+            _modifiedRanges.SignalModified(address, size);
+
+            if (!_syncActionRegistered)
+            {
+                _context.RegisterSyncAction(SyncAction);
+                _syncActionRegistered = true;
+            }
+        }
+
+        /// <summary>
+        /// Indicate that mofifications in a given region of this buffer have been overwritten.
+        /// </summary>
+        /// <param name="address">The start address of the region</param>
+        /// <param name="size">The size of the region</param>
+        public void ClearModified(ulong address, ulong size)
+        {
+            if (_modifiedRanges != null)
+            {
+                _modifiedRanges.Clear(address, size);
+            }
+        }
+
+        /// <summary>
+        /// Action to be performed when a syncpoint is reached after modification.
+        /// This will register read/write tracking to flush the buffer from GPU when its memory is used.
+        /// </summary>
+        private void SyncAction()
+        {
+            _syncActionRegistered = false;
+
+            if (_useGranular)
+            {
+                _modifiedRanges.GetRanges(Address, Size, (address, size) =>
+                {
+                    _memoryTrackingGranular.RegisterAction(address, size, _externalFlushDelegate);
+                    SynchronizeMemory(address, size);
+                });
+            }
+            else
+            {
+                _memoryTracking.RegisterAction(_externalFlushDelegate);
+                SynchronizeMemory(Address, Size);
+            }
+        }
+
+        /// <summary>
+        /// Inherit modified ranges from another buffer.
+        /// </summary>
+        /// <param name="from">The buffer to inherit from</param>
+        public void InheritModifiedRanges(Buffer from)
+        {
+            if (from._modifiedRanges != null)
+            {
+                if (from._syncActionRegistered && !_syncActionRegistered)
+                {
+                    _context.RegisterSyncAction(SyncAction);
+                    _syncActionRegistered = true;
+                }
+
+                EnsureRangeList();
+                _modifiedRanges.InheritRanges(from._modifiedRanges, (ulong address, ulong size) =>
+                {
+                    if (_useGranular)
+                    {
+                        _memoryTrackingGranular.RegisterAction(address, size, _externalFlushDelegate);
+                    }
+                    else
+                    {
+                        _memoryTracking.RegisterAction(_externalFlushDelegate);
+                    }
+                });
+            }
+        }
+
+        /// <summary>
+        /// Determine if a given region of the buffer has been modified, and must be flushed.
+        /// </summary>
+        /// <param name="address">The start address of the region</param>
+        /// <param name="size">The size of the region</param>
+        /// <returns></returns>
+        public bool IsModified(ulong address, ulong size)
+        {
+            if (_modifiedRanges != null)
+            {
+                return _modifiedRanges.HasRange(address, size);
+            }
+
+            return false;
+        }
+
         /// <summary>
         /// Indicate that a region of the buffer was modified, and must be loaded from memory.
         /// </summary>
@@ -141,6 +279,23 @@ namespace Ryujinx.Graphics.Gpu.Memory
                 mSize = maxSize;
             }
 
+            if (_modifiedRanges != null)
+            {
+                _modifiedRanges.ExcludeModifiedRegions(mAddress, mSize, _loadDelegate);
+            }
+            else
+            {
+                LoadRegion(mAddress, mSize);
+            }
+        }
+
+        /// <summary>
+        /// Load a region of the buffer from memory.
+        /// </summary>
+        /// <param name="mAddress">Start address of the modified region</param>
+        /// <param name="mSize">Size of the modified region</param>
+        private void LoadRegion(ulong mAddress, ulong mSize)
+        {
             int offset = (int)(mAddress - Address);
 
             _context.Renderer.SetBufferData(Handle, offset, _context.PhysicalMemory.GetSpan(mAddress, (int)mSize));
@@ -172,15 +327,62 @@ namespace Ryujinx.Graphics.Gpu.Memory
             _context.PhysicalMemory.WriteUntracked(address, data);
         }
 
+        /// <summary>
+        /// Align a given address and size region to page boundaries.
+        /// </summary>
+        /// <param name="address">The start address of the region</param>
+        /// <param name="size">The size of the region</param>
+        /// <returns>The page aligned address and size</returns>
+        private static (ulong address, ulong size) PageAlign(ulong address, ulong size)
+        {
+            ulong pageMask = MemoryManager.PageMask;
+            ulong rA = address & ~pageMask;
+            ulong rS = ((address + size + pageMask) & ~pageMask) - rA;
+            return (rA, rS);
+        }
+
+        /// <summary>
+        /// Flush modified ranges of the buffer from another thread.
+        /// This will flush all modifications made before the active SyncNumber was set, and may block to wait for GPU sync.
+        /// </summary>
+        /// <param name="address">Address of the memory action</param>
+        /// <param name="size">Size in bytes</param>
+        public void ExternalFlush(ulong address, ulong size)
+        {
+            _context.Renderer.BackgroundContextAction(() =>
+            {
+                var ranges = _modifiedRanges;
+
+                if (ranges != null)
+                {
+                    (address, size) = PageAlign(address, size);
+                    ranges.WaitForAndGetRanges(address, size, Flush);
+                }
+            });
+        }
+
+        /// <summary>
+        /// Called when part of the memory for this buffer has been unmapped.
+        /// Calls are from non-GPU threads.
+        /// </summary>
+        /// <param name="address">Start address of the unmapped region</param>
+        /// <param name="size">Size of the unmapped region</param>
+        public void Unmapped(ulong address, ulong size)
+        {
+            _modifiedRanges?.Clear(address, size);
+        }
+
         /// <summary>
         /// Disposes the host buffer.
         /// </summary>
         public void Dispose()
         {
-            _context.Renderer.DeleteBuffer(Handle);
+            _modifiedRanges?.Clear();
 
             _memoryTrackingGranular?.Dispose();
             _memoryTracking?.Dispose();
+
+            _context.Renderer.DeleteBuffer(Handle);
         }
     }
 }

Ryujinx.Graphics.Gpu/Memory/BufferBounds.cs

@@ -1,3 +1,5 @@
+using Ryujinx.Graphics.Shader;
+
 namespace Ryujinx.Graphics.Gpu.Memory
 {
     /// <summary>
@@ -15,15 +17,22 @@ namespace Ryujinx.Graphics.Gpu.Memory
         /// </summary>
         public ulong Size { get; }
 
+        /// <summary>
+        /// Buffer usage flags.
+        /// </summary>
+        public BufferUsageFlags Flags { get; }
+
         /// <summary>
         /// Creates a new buffer region.
         /// </summary>
         /// <param name="address">Region address</param>
         /// <param name="size">Region size</param>
-        public BufferBounds(ulong address, ulong size)
+        /// <param name="flags">Buffer usage flags</param>
+        public BufferBounds(ulong address, ulong size, BufferUsageFlags flags = BufferUsageFlags.None)
         {
             Address = address;
             Size = size;
+            Flags = flags;
         }
     }
 }

Ryujinx.Graphics.Gpu/Memory/BufferManager.cs

@@ -68,9 +68,10 @@ namespace Ryujinx.Graphics.Gpu.Memory
             /// <param name="index">Buffer slot</param>
             /// <param name="address">Region virtual address</param>
             /// <param name="size">Region size in bytes</param>
-            public void SetBounds(int index, ulong address, ulong size)
+            /// <param name="flags">Buffer usage flags</param>
+            public void SetBounds(int index, ulong address, ulong size, BufferUsageFlags flags = BufferUsageFlags.None)
             {
-                Buffers[index] = new BufferBounds(address, size);
+                Buffers[index] = new BufferBounds(address, size, flags);
             }
 
             /// <summary>
@@ -219,7 +220,8 @@ namespace Ryujinx.Graphics.Gpu.Memory
         /// <param name="index">Index of the storage buffer</param>
         /// <param name="gpuVa">Start GPU virtual address of the buffer</param>
         /// <param name="size">Size in bytes of the storage buffer</param>
-        public void SetComputeStorageBuffer(int index, ulong gpuVa, ulong size)
+        /// <param name="flags">Buffer usage flags</param>
+        public void SetComputeStorageBuffer(int index, ulong gpuVa, ulong size, BufferUsageFlags flags)
         {
             size += gpuVa & ((ulong)_context.Capabilities.StorageBufferOffsetAlignment - 1);
 
@@ -227,7 +229,7 @@ namespace Ryujinx.Graphics.Gpu.Memory
 
             ulong address = TranslateAndCreateBuffer(gpuVa, size);
 
-            _cpStorageBuffers.SetBounds(index, address, size);
+            _cpStorageBuffers.SetBounds(index, address, size, flags);
         }
 
         /// <summary>
@@ -238,7 +240,8 @@ namespace Ryujinx.Graphics.Gpu.Memory
         /// <param name="index">Index of the storage buffer</param>
         /// <param name="gpuVa">Start GPU virtual address of the buffer</param>
         /// <param name="size">Size in bytes of the storage buffer</param>
-        public void SetGraphicsStorageBuffer(int stage, int index, ulong gpuVa, ulong size)
+        /// <param name="flags">Buffer usage flags</param>
+        public void SetGraphicsStorageBuffer(int stage, int index, ulong gpuVa, ulong size, BufferUsageFlags flags)
         {
             size += gpuVa & ((ulong)_context.Capabilities.StorageBufferOffsetAlignment - 1);
 
@@ -252,7 +255,7 @@ namespace Ryujinx.Graphics.Gpu.Memory
                 _gpStorageBuffersDirty = true;
             }
 
-            _gpStorageBuffers[stage].SetBounds(index, address, size);
+            _gpStorageBuffers[stage].SetBounds(index, address, size, flags);
         }
 
         /// <summary>
@@ -385,6 +388,30 @@ namespace Ryujinx.Graphics.Gpu.Memory
             return mask;
         }
 
+        /// <summary>
+        /// Handles removal of buffers written to a memory region being unmapped.
+        /// </summary>
+        /// <param name="sender">Sender object</param>
+        /// <param name="e">Event arguments</param>
+        public void MemoryUnmappedHandler(object sender, UnmapEventArgs e)
+        {
+            Buffer[] overlaps = new Buffer[10];
+            int overlapCount;
+
+            ulong address = _context.MemoryManager.Translate(e.Address);
+            ulong size = e.Size;
+
+            lock (_buffers)
+            {
+                overlapCount = _buffers.FindOverlaps(address, size, ref overlaps);
+            }
+
+            for (int i = 0; i < overlapCount; i++)
+            {
+                overlaps[i].Unmapped(address, size);
+            }
+        }
+
         /// <summary>
         /// Performs address translation of the GPU virtual address, and creates a
         /// new buffer, if needed, for the specified range.
@@ -443,7 +470,12 @@ namespace Ryujinx.Graphics.Gpu.Memory
         /// <param name="size">Size in bytes of the buffer</param>
         private void CreateBufferAligned(ulong address, ulong size)
         {
-            int overlapsCount = _buffers.FindOverlapsNonOverlapping(address, size, ref _bufferOverlaps);
+            int overlapsCount;
+
+            lock (_buffers)
+            {
+                overlapsCount = _buffers.FindOverlapsNonOverlapping(address, size, ref _bufferOverlaps);
+            }
 
             if (overlapsCount != 0)
             {
@@ -463,15 +495,19 @@ namespace Ryujinx.Graphics.Gpu.Memory
                         address    = Math.Min(address,    buffer.Address);
                         endAddress = Math.Max(endAddress, buffer.EndAddress);
 
-                        buffer.SynchronizeMemory(buffer.Address, buffer.Size);
+                        lock (_buffers)
-
+                        {
                             _buffers.Remove(buffer);
                         }
+                    }
 
                     Buffer newBuffer = new Buffer(_context, address, endAddress - address);
                     newBuffer.SynchronizeMemory(address, endAddress - address);
 
+                    lock (_buffers)
+                    {
                         _buffers.Add(newBuffer);
+                    }
 
                     for (int index = 0; index < overlapsCount; index++)
                     {
@@ -479,7 +515,10 @@ namespace Ryujinx.Graphics.Gpu.Memory
 
                         int dstOffset = (int)(buffer.Address - newBuffer.Address);
 
+                        buffer.SynchronizeMemory(buffer.Address, buffer.Size);
+
                         buffer.CopyTo(newBuffer, dstOffset);
+                        newBuffer.InheritModifiedRanges(buffer);
 
                         buffer.Dispose();
                     }
@@ -493,8 +532,11 @@ namespace Ryujinx.Graphics.Gpu.Memory
                 // No overlap, just create a new buffer.
                 Buffer buffer = new Buffer(_context, address, size);
 
+                lock (_buffers)
+                {
                     _buffers.Add(buffer);
                 }
+            }
 
             ShrinkOverlapsBufferIfNeeded();
         }
@@ -549,7 +591,7 @@ namespace Ryujinx.Graphics.Gpu.Memory
 
                 if (bounds.Address != 0)
                 {
-                    sRanges[bindingInfo.Binding] = GetBufferRange(bounds.Address, bounds.Size);
+                    sRanges[bindingInfo.Binding] = GetBufferRange(bounds.Address, bounds.Size, bounds.Flags.HasFlag(BufferUsageFlags.Write));
                 }
             }
 
@@ -722,7 +764,7 @@ namespace Ryujinx.Graphics.Gpu.Memory
 
                     if (bounds.Address != 0)
                     {
-                        ranges[bindingInfo.Binding] = GetBufferRange(bounds.Address, bounds.Size);
+                        ranges[bindingInfo.Binding] = GetBufferRange(bounds.Address, bounds.Size, bounds.Flags.HasFlag(BufferUsageFlags.Write));
                     }
                 }
             }
@@ -818,7 +860,17 @@ namespace Ryujinx.Graphics.Gpu.Memory
                 dstOffset,
                 (int)size);
 
-            dstBuffer.Flush(dstAddress, size);
+            if (srcBuffer.IsModified(srcAddress, size))
+            {
+                dstBuffer.SignalModified(dstAddress, size);
+            }
+            else
+            {
+                // Optimization: If the data being copied is already in memory, then copy it directly instead of flushing from GPU.
+
+                dstBuffer.ClearModified(dstAddress, size);
+                _context.PhysicalMemory.WriteUntracked(dstAddress, _context.PhysicalMemory.GetSpan(srcAddress, (int)size));
+            }
         }
 
         /// <summary>
@@ -840,7 +892,7 @@ namespace Ryujinx.Graphics.Gpu.Memory
 
             _context.Renderer.Pipeline.ClearBuffer(buffer.Handle, offset, (int)size, value);
 
-            buffer.Flush(address, size);
+            buffer.SignalModified(address, size);
         }
 
         /// <summary>
@@ -848,10 +900,11 @@ namespace Ryujinx.Graphics.Gpu.Memory
         /// </summary>
         /// <param name="address">Start address of the memory range</param>
         /// <param name="size">Size in bytes of the memory range</param>
+        /// <param name="write">Whether the buffer will be written to by this use</param>
         /// <returns>The buffer sub-range for the given range</returns>
-        private BufferRange GetBufferRange(ulong address, ulong size)
+        private BufferRange GetBufferRange(ulong address, ulong size, bool write = false)
         {
-            return GetBuffer(address, size).GetRange(address, size);
+            return GetBuffer(address, size, write).GetRange(address, size);
         }
 
         /// <summary>
@@ -860,21 +913,33 @@ namespace Ryujinx.Graphics.Gpu.Memory
         /// </summary>
         /// <param name="address">Start address of the memory range</param>
         /// <param name="size">Size in bytes of the memory range</param>
+        /// <param name="write">Whether the buffer will be written to by this use</param>
         /// <returns>The buffer where the range is fully contained</returns>
-        private Buffer GetBuffer(ulong address, ulong size)
+        private Buffer GetBuffer(ulong address, ulong size, bool write = false)
         {
             Buffer buffer;
 
             if (size != 0)
+            {
+                lock (_buffers)
                 {
                     buffer = _buffers.FindFirstOverlap(address, size);
+                }
 
                 buffer.SynchronizeMemory(address, size);
+
+                if (write)
+                {
+                    buffer.SignalModified(address, size);
+                }
             }
             else
+            {
+                lock (_buffers)
                 {
                     buffer = _buffers.FindFirstOverlap(address, 1);
                 }
+            }
 
             return buffer;
         }
@@ -888,7 +953,12 @@ namespace Ryujinx.Graphics.Gpu.Memory
         {
             if (size != 0)
             {
-                Buffer buffer = _buffers.FindFirstOverlap(address, size);
+                Buffer buffer;
+
+                lock (_buffers)
+                {
+                    buffer = _buffers.FindFirstOverlap(address, size);
+                }
 
                 buffer.SynchronizeMemory(address, size);
             }
@@ -899,6 +969,8 @@ namespace Ryujinx.Graphics.Gpu.Memory
         /// It's an error to use the buffer manager after disposal.
         /// </summary>
         public void Dispose()
+        {
+            lock (_buffers)
             {
                 foreach (Buffer buffer in _buffers)
                 {
@@ -907,3 +979,4 @@ namespace Ryujinx.Graphics.Gpu.Memory
             }
         }
     }
+}

Ryujinx.Graphics.Gpu/Memory/BufferModifiedRangeList.cs

@@ -0,0 +1,367 @@
+using Ryujinx.Memory.Range;
+using System;
+using System.Linq;
+
+namespace Ryujinx.Graphics.Gpu.Memory
+{
+    /// <summary>
+    /// A range within a buffer that has been modified by the GPU.
+    /// </summary>
+    class BufferModifiedRange : IRange
+    {
+        /// <summary>
+        /// Start address of the range in guest memory.
+        /// </summary>
+        public ulong Address { get; }
+
+        /// <summary>
+        /// Size of the range in bytes.
+        /// </summary>
+        public ulong Size { get; }
+
+        /// <summary>
+        /// End address of the range in guest memory.
+        /// </summary>
+        public ulong EndAddress => Address + Size;
+
+        /// <summary>
+        /// The GPU sync number at the time of the last modification.
+        /// </summary>
+        public ulong SyncNumber { get; internal set; }
+
+        /// <summary>
+        /// Creates a new instance of a modified range.
+        /// </summary>
+        /// <param name="address">Start address of the range</param>
+        /// <param name="size">Size of the range in bytes</param>
+        /// <param name="syncNumber">The GPU sync number at the time of creation</param>
+        public BufferModifiedRange(ulong address, ulong size, ulong syncNumber)
+        {
+            Address = address;
+            Size = size;
+            SyncNumber = syncNumber;
+        }
+
+        /// <summary>
+        /// Checks if a given range overlaps with the modified range.
+        /// </summary>
+        /// <param name="address">Start address of the range</param>
+        /// <param name="size">Size in bytes of the range</param>
+        /// <returns>True if the range overlaps, false otherwise</returns>
+        public bool OverlapsWith(ulong address, ulong size)
+        {
+            return Address < address + size && address < EndAddress;
+        }
+    }
+
+    /// <summary>
+    /// A structure used to track GPU modified ranges within a buffer.
+    /// </summary>
+    class BufferModifiedRangeList : RangeList<BufferModifiedRange>
+    {
+        private GpuContext _context;
+
+        private object _lock = new object();
+
+        // The list can be accessed from both the GPU thread, and a background thread.
+        private BufferModifiedRange[] _foregroundOverlaps = new BufferModifiedRange[1];
+        private BufferModifiedRange[] _backgroundOverlaps = new BufferModifiedRange[1];
+
+        /// <summary>
+        /// Creates a new instance of a modified range list.
+        /// </summary>
+        /// <param name="context">GPU context that the buffer range list belongs to</param>
+        public BufferModifiedRangeList(GpuContext context)
+        {
+            _context = context;
+        }
+
+        /// <summary>
+        /// Given an input range, calls the given action with sub-ranges which exclude any of the modified regions.
+        /// </summary>
+        /// <param name="address">Start address of the query range</param>
+        /// <param name="size">Size of the query range in bytes</param>
+        /// <param name="action">Action to perform for each remaining sub-range of the input range</param>
+        public void ExcludeModifiedRegions(ulong address, ulong size, Action<ulong, ulong> action)
+        {
+            lock (_lock)
+            {
+                // Slices a given region using the modified regions in the list. Calls the action for the new slices.
+                int count = FindOverlapsNonOverlapping(address, size, ref _foregroundOverlaps);
+
+                for (int i = 0; i < count; i++)
+                {
+                    BufferModifiedRange overlap = _foregroundOverlaps[i];
+                    
+                    if (overlap.Address > address)
+                    {
+                        // The start of the remaining region is uncovered by this overlap. Call the action for it.
+                        action(address, overlap.Address - address);
+                    }
+
+                    // Remaining region is after this overlap.
+                    size -= overlap.EndAddress - address;
+                    address = overlap.EndAddress;
+                }
+
+                if ((long)size > 0)
+                {
+                    // If there is any region left after removing the overlaps, signal it.
+                    action(address, size);
+                }
+            }
+        }
+
+        /// <summary>
+        /// Signal that a region of the buffer has been modified, and add the new region to the range list.
+        /// Any overlapping ranges will be (partially) removed.
+        /// </summary>
+        /// <param name="address">Start address of the modified region</param>
+        /// <param name="size">Size of the modified region in bytes</param>
+        public void SignalModified(ulong address, ulong size)
+        {
+            // Must lock, as this can affect flushes from the background thread.
+            lock (_lock)
+            {
+                // We may overlap with some existing modified regions. They must be cut into by the new entry.
+                int count = FindOverlapsNonOverlapping(address, size, ref _foregroundOverlaps);
+
+                ulong endAddress = address + size;
+                ulong syncNumber = _context.SyncNumber;
+
+                for (int i = 0; i < count; i++)
+                {
+                    // The overlaps must be removed or split.
+
+                    BufferModifiedRange overlap = _foregroundOverlaps[i];
+
+                    if (overlap.Address == address && overlap.Size == size)
+                    {
+                        // Region already exists. Just update the existing sync number.
+                        overlap.SyncNumber = syncNumber;
+
+                        return;
+                    }
+
+                    Remove(overlap);
+
+                    if (overlap.Address < address && overlap.EndAddress > address)
+                    {
+                        // A split item must be created behind this overlap.
+
+                        Add(new BufferModifiedRange(overlap.Address, address - overlap.Address, overlap.SyncNumber));
+                    }
+
+                    if (overlap.Address < endAddress && overlap.EndAddress > endAddress)
+                    {
+                        // A split item must be created after this overlap.
+
+                        Add(new BufferModifiedRange(endAddress, overlap.EndAddress - endAddress, overlap.SyncNumber));
+                    }
+                }
+
+                Add(new BufferModifiedRange(address, size, syncNumber));
+            }
+        }
+
+        /// <summary>
+        /// Gets modified ranges within the specified region, and then fires the given action for each range individually.
+        /// </summary>
+        /// <param name="address">Start address to query</param>
+        /// <param name="size">Size to query</param>
+        /// <param name="rangeAction">The action to call for each modified range</param>
+        public void GetRanges(ulong address, ulong size, Action<ulong, ulong> rangeAction)
+        {
+            int count = 0;
+
+            // Range list must be consistent for this operation.
+            lock (_lock)
+            {
+                count = FindOverlapsNonOverlapping(address, size, ref _foregroundOverlaps);
+            }
+
+            for (int i = 0; i < count; i++)
+            {
+                BufferModifiedRange overlap = _foregroundOverlaps[i];
+                rangeAction(overlap.Address, overlap.Size);
+            }
+        }
+
+        /// <summary>
+        /// Queries if a range exists within the specified region.
+        /// </summary>
+        /// <param name="address">Start address to query</param>
+        /// <param name="size">Size to query</param>
+        /// <returns>True if a range exists in the specified region, false otherwise</returns>
+        public bool HasRange(ulong address, ulong size)
+        {
+            // Range list must be consistent for this operation.
+            lock (_lock)
+            {
+                return FindOverlapsNonOverlapping(address, size, ref _foregroundOverlaps) > 0;
+            }
+        }
+
+        /// <summary>
+        /// Gets modified ranges within the specified region, waits on ones from a previous sync number,
+        /// and then fires the given action for each range individually.
+        /// </summary>
+        /// <remarks>
+        /// This function assumes it is called from the background thread.
+        /// Modifications from the current sync number are ignored because the guest should not expect them to be available yet.
+        /// They will remain reserved, so that any data sync prioritizes the data in the GPU.
+        /// </remarks>
+        /// <param name="address">Start address to query</param>
+        /// <param name="size">Size to query</param>
+        /// <param name="rangeAction">The action to call for each modified range</param>
+        public void WaitForAndGetRanges(ulong address, ulong size, Action<ulong, ulong> rangeAction)
+        {
+            ulong endAddress = address + size;
+            ulong currentSync = _context.SyncNumber;
+
+            int rangeCount = 0;
+
+            // Range list must be consistent for this operation
+            lock (_lock)
+            {
+                rangeCount = FindOverlapsNonOverlapping(address, size, ref _backgroundOverlaps);
+            }
+
+            if (rangeCount == 0)
+            {
+                return;
+            }
+
+            // First, determine which syncpoint to wait on.
+            // This is the latest syncpoint that is not equal to the current sync.
+
+            long highestDiff = long.MinValue;
+
+            for (int i = 0; i < rangeCount; i++)
+            {
+                BufferModifiedRange overlap = _backgroundOverlaps[i];
+
+                long diff = (long)(overlap.SyncNumber - currentSync);
+
+                if (diff < 0 && diff > highestDiff)
+                {
+                    highestDiff = diff;
+                }
+            }
+
+            if (highestDiff == long.MinValue)
+            {
+                return;
+            }
+
+            // Wait for the syncpoint.
+            _context.Renderer.WaitSync(currentSync + (ulong)highestDiff);
+
+            // Flush and remove all regions with the older syncpoint.
+            lock (_lock)
+            {
+                for (int i = 0; i < rangeCount; i++)
+                {
+                    BufferModifiedRange overlap = _backgroundOverlaps[i];
+
+                    long diff = (long)(overlap.SyncNumber - currentSync);
+
+                    if (diff <= highestDiff)
+                    {
+                        ulong clampAddress = Math.Max(address, overlap.Address);
+                        ulong clampEnd = Math.Min(endAddress, overlap.EndAddress);
+
+                        ClearPart(overlap, clampAddress, clampEnd);
+
+                        rangeAction(clampAddress, clampEnd - clampAddress);
+                    }
+                }
+            }
+        }
+
+        /// <summary>
+        /// Inherit ranges from another modified range list.
+        /// </summary>
+        /// <param name="ranges">The range list to inherit from</param>
+        /// <param name="rangeAction">The action to call for each modified range</param>
+        public void InheritRanges(BufferModifiedRangeList ranges, Action<ulong, ulong> rangeAction)
+        {
+            BufferModifiedRange[] inheritRanges;
+
+            lock (ranges._lock)
+            {
+                inheritRanges = ranges.ToArray();
+            }
+
+            lock (_lock)
+            {
+                foreach (BufferModifiedRange range in inheritRanges)
+                {
+                    Add(range);
+                }
+            }
+
+            ulong currentSync = _context.SyncNumber;
+            foreach (BufferModifiedRange range in inheritRanges)
+            {
+                if (range.SyncNumber != currentSync)
+                {
+                    rangeAction(range.Address, range.Size);
+                }
+            }
+        }
+
+        private void ClearPart(BufferModifiedRange overlap, ulong address, ulong endAddress)
+        {
+            Remove(overlap);
+
+            // If the overlap extends outside of the clear range, make sure those parts still exist.
+
+            if (overlap.Address < address)
+            {
+                Add(new BufferModifiedRange(overlap.Address, address - overlap.Address, overlap.SyncNumber));
+            }
+
+            if (overlap.EndAddress > endAddress)
+            {
+                Add(new BufferModifiedRange(endAddress, overlap.EndAddress - endAddress, overlap.SyncNumber));
+            }
+        }
+
+        /// <summary>
+        /// Clear modified ranges within the specified area.
+        /// </summary>
+        /// <param name="address">Start address to clear</param>
+        /// <param name="size">Size to clear</param>
+        public void Clear(ulong address, ulong size)
+        {
+            lock (_lock)
+            {
+                // This function can be called from any thread, so it cannot use the arrays for background or foreground.
+                BufferModifiedRange[] toClear = new BufferModifiedRange[1];
+
+                int rangeCount = FindOverlapsNonOverlapping(address, size, ref toClear);
+
+                ulong endAddress = address + size;
+
+                for (int i = 0; i < rangeCount; i++)
+                {
+                    BufferModifiedRange overlap = toClear[i];
+
+                    ClearPart(overlap, address, endAddress);
+                }
+            }
+        }
+
+        /// <summary>
+        /// Clear all modified ranges.
+        /// </summary>
+        public void Clear()
+        {
+            lock (_lock)
+            {
+                Items.Clear();
+            }
+        }
+    }
+}

Ryujinx.Graphics.Gpu/Memory/MemoryManager.cs

@@ -61,6 +61,7 @@ namespace Ryujinx.Graphics.Gpu.Memory
         /// </summary>
         /// <param name="va">GPU virtual address where the data is located</param>
         /// <param name="size">Size of the data</param>
+        /// <param name="tracked">True if read tracking is triggered on the span</param>
         /// <returns>The span of the data at the specified memory location</returns>
         public ReadOnlySpan<byte> GetSpan(ulong va, int size, bool tracked = false)
         {

Ryujinx.Graphics.Gpu/Shader/ShaderCache.cs

@@ -34,7 +34,7 @@ namespace Ryujinx.Graphics.Gpu.Shader
         /// <summary>
         /// Version of the codegen (to be changed when codegen or guest format change).
         /// </summary>
-        private const ulong ShaderCodeGenVersion = 1910;
+        private const ulong ShaderCodeGenVersion = 1790;
 
         /// <summary>
         /// Creates a new instance of the shader cache.

Ryujinx.Graphics.OpenGL/Renderer.cs

@@ -26,6 +26,8 @@ namespace Ryujinx.Graphics.OpenGL
         private TextureCopy _backgroundTextureCopy;
         internal TextureCopy TextureCopy => BackgroundContextWorker.InBackground ? _backgroundTextureCopy : _textureCopy;
 
+        private Sync _sync;
+
         internal ResourcePool ResourcePool { get; }
 
         public string GpuVendor { get; private set; }
@@ -39,6 +41,7 @@ namespace Ryujinx.Graphics.OpenGL
             _window = new Window(this);
             _textureCopy = new TextureCopy(this);
             _backgroundTextureCopy = new TextureCopy(this);
+            _sync = new Sync();
             ResourcePool = new ResourcePool();
         }
 
@@ -108,6 +111,7 @@ namespace Ryujinx.Graphics.OpenGL
 
         public void PreFrame()
         {
+            _sync.Cleanup();
             ResourcePool.Tick();
         }
 
@@ -164,6 +168,7 @@ namespace Ryujinx.Graphics.OpenGL
             _pipeline.Dispose();
             _window.Dispose();
             _counters.Dispose();
+            _sync.Dispose();
         }
 
         public IProgram LoadProgramBinary(byte[] programBinary)
@@ -179,5 +184,15 @@ namespace Ryujinx.Graphics.OpenGL
 
             return null;
         }
+
+        public void CreateSync(ulong id)
+        {
+            _sync.Create(id);
+        }
+
+        public void WaitSync(ulong id)
+        {
+            _sync.Wait(id);
+        }
     }
 }

Ryujinx.Graphics.OpenGL/Sync.cs

@@ -0,0 +1,129 @@
+using OpenTK.Graphics.OpenGL;
+using Ryujinx.Common.Logging;
+using System;
+using System.Collections.Generic;
+using System.Linq;
+
+namespace Ryujinx.Graphics.OpenGL
+{
+    class Sync : IDisposable
+    {
+        private class SyncHandle
+        {
+            public ulong ID;
+            public IntPtr Handle;
+        }
+
+        private ulong _firstHandle = 0;
+
+        private List<SyncHandle> Handles = new List<SyncHandle>();
+
+        public void Create(ulong id)
+        {
+            SyncHandle handle = new SyncHandle
+            {
+                ID = id,
+                Handle = GL.FenceSync(SyncCondition.SyncGpuCommandsComplete, WaitSyncFlags.None)
+            };
+
+            lock (Handles)
+            {
+                Handles.Add(handle);
+            }
+        }
+
+        public void Wait(ulong id)
+        {
+            SyncHandle result = null;
+
+            lock (Handles)
+            {
+                if ((long)(_firstHandle - id) > 0)
+                {
+                    return; // The handle has already been signalled or deleted.
+                }
+
+                foreach (SyncHandle handle in Handles)
+                {
+                    if (handle.ID == id)
+                    {
+                        result = handle;
+                        break;
+                    }
+                }
+            }
+
+            if (result != null)
+            {
+                lock (result)
+                {
+                    if (result.Handle == IntPtr.Zero)
+                    {
+                        return;
+                    }
+
+                    WaitSyncStatus syncResult = GL.ClientWaitSync(result.Handle, ClientWaitSyncFlags.SyncFlushCommandsBit, 1000000000);
+                    
+                    if (syncResult == WaitSyncStatus.TimeoutExpired)
+                    {
+                        Logger.Error?.PrintMsg(LogClass.Gpu, $"GL Sync Object {result.ID} failed to signal within 1000ms. Continuing...");
+                    }
+                }
+            }
+        }
+
+        public void Cleanup()
+        {
+            // Iterate through handles and remove any that have already been signalled.
+
+            while (true)
+            {
+                SyncHandle first = null;
+                lock (Handles)
+                {
+                    first = Handles.FirstOrDefault();
+                }
+
+                if (first == null) break;
+
+                WaitSyncStatus syncResult = GL.ClientWaitSync(first.Handle, ClientWaitSyncFlags.SyncFlushCommandsBit, 0);
+
+                if (syncResult == WaitSyncStatus.AlreadySignaled)
+                {
+                    // Delete the sync object.
+                    lock (Handles)
+                    {
+                        lock (first)
+                        {
+                            _firstHandle = first.ID + 1;
+                            Handles.RemoveAt(0);
+                            GL.DeleteSync(first.Handle);
+                            first.Handle = IntPtr.Zero;
+                        }
+                    }
+                } else
+                {
+                    // This sync handle and any following have not been reached yet.
+                    break;
+                }
+            }
+        }
+
+        public void Dispose()
+        {
+            lock (Handles)
+            {
+                foreach (SyncHandle handle in Handles)
+                {
+                    lock (handle)
+                    {
+                        GL.DeleteSync(handle.Handle);
+                        handle.Handle = IntPtr.Zero;
+                    }
+                }
+
+                Handles.Clear();
+            }
+        }
+    }
+}

Ryujinx.Graphics.Shader/BufferDescriptor.cs

@@ -4,11 +4,21 @@ namespace Ryujinx.Graphics.Shader
     {
         public readonly int Binding;
         public readonly int Slot;
+        public BufferUsageFlags Flags;
 
         public BufferDescriptor(int binding, int slot)
         {
             Binding = binding;
             Slot = slot;
+
+            Flags = BufferUsageFlags.None;
+        }
+
+        public BufferDescriptor SetFlag(BufferUsageFlags flag)
+        {
+            Flags |= flag;
+
+            return this;
         }
     }
 }

Ryujinx.Graphics.Shader/BufferUsageFlags.cs

@@ -0,0 +1,18 @@
+using System;
+
+namespace Ryujinx.Graphics.Shader
+{
+    /// <summary>
+    /// Flags that indicate how a buffer will be used in a shader.
+    /// </summary>
+    [Flags]
+    public enum BufferUsageFlags
+    {
+        None = 0,
+
+        /// <summary>
+        /// Buffer is written to.
+        /// </summary>
+        Write = 1 << 0
+    }
+}

Ryujinx.Graphics.Shader/CodeGen/Glsl/Instructions/InstGenMemory.cs

@@ -298,6 +298,7 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Glsl.Instructions
 
             string src = TypeConversion.ReinterpretCast(context, src3, srcType, VariableType.U32);
 
+            SetStorageWriteFlag(context, src1, context.Config.Stage);
             string sb = GetStorageBufferAccessor(indexExpr, offsetExpr, context.Config.Stage);
 
             return $"{sb} = {src}";
@@ -629,6 +630,32 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Glsl.Instructions
             }
         }
 
+        private static void SetStorageWriteFlag(CodeGenContext context, IAstNode indexExpr, ShaderStage stage)
+        {
+            // Attempt to find a BufferDescriptor with the given index.
+            // If it cannot be resolved or is not constant, assume that the slot expression could potentially index any of them,
+            // and set the flag on all storage buffers.
+
+            int index = -1;
+
+            if (indexExpr is AstOperand operand && operand.Type == OperandType.Constant)
+            {
+                index = context.SBufferDescriptors.FindIndex(buffer => buffer.Slot == operand.Value);
+            }
+
+            if (index != -1)
+            {
+                context.SBufferDescriptors[index] = context.SBufferDescriptors[index].SetFlag(BufferUsageFlags.Write);
+            }
+            else
+            {
+                for (int i = 0; i < context.SBufferDescriptors.Count; i++)
+                {
+                    context.SBufferDescriptors[i] = context.SBufferDescriptors[i].SetFlag(BufferUsageFlags.Write);
+                }
+            }
+        }
+
         private static string GetStorageBufferAccessor(string slotExpr, string offsetExpr, ShaderStage stage)
         {
             string sbName = OperandManager.GetShaderStagePrefix(stage);

Ryujinx.Memory.Tests/MockVirtualMemoryManager.cs

@@ -6,6 +6,8 @@ namespace Ryujinx.Memory.Tests
     {
         public bool NoMappings;
 
+        public event Action<ulong, ulong, MemoryPermission> OnProtect;
+
         public MockVirtualMemoryManager(ulong size, int pageSize)
         {
         }
@@ -82,6 +84,7 @@ namespace Ryujinx.Memory.Tests
 
         public void TrackingReprotect(ulong va, ulong size, MemoryPermission protection)
         {
+            OnProtect?.Invoke(va, size, protection);
         }
     }
 }

Ryujinx.Memory.Tests/TrackingTests.cs

@@ -421,5 +421,68 @@ namespace Ryujinx.Memory.Tests
 
             Assert.AreEqual((0, 0), _tracking.GetRegionCounts());
         }
+
+        [Test]
+        public void ReadAndWriteProtection()
+        {
+            MemoryPermission protection = MemoryPermission.ReadAndWrite;
+
+            _memoryManager.OnProtect += (va, size, newProtection) =>
+            {
+                Assert.AreEqual((0, PageSize), (va, size)); // Should protect the exact region all the operations use.
+                protection = newProtection;
+            };
+
+            RegionHandle handle = _tracking.BeginTracking(0, PageSize);
+
+            // After creating the handle, there is no protection yet.
+            Assert.AreEqual(MemoryPermission.ReadAndWrite, protection);
+
+            bool dirtyInitial = handle.Dirty;
+            Assert.True(dirtyInitial); // Handle starts dirty.
+
+            handle.Reprotect();
+
+            // After a reprotect, there is write protection, which will set a dirty flag when any write happens.
+            Assert.AreEqual(MemoryPermission.Read, protection);
+
+            (ulong address, ulong size)? readTrackingTriggered = null;
+            handle.RegisterAction((address, size) =>
+            {
+                readTrackingTriggered = (address, size);
+            });
+
+            // Registering an action adds read/write protection.
+            Assert.AreEqual(MemoryPermission.None, protection);
+
+            bool dirtyAfterReprotect = handle.Dirty;
+            Assert.False(dirtyAfterReprotect); // Handle is no longer dirty.
+
+            // First we should read, which will trigger the action. This _should not_ remove write protection on the memory.
+
+            _tracking.VirtualMemoryEvent(0, 4, false);
+
+            bool dirtyAfterRead = handle.Dirty;
+            Assert.False(dirtyAfterRead); // Not dirtied, as this was a read.
+
+            Assert.AreEqual(readTrackingTriggered, (0UL, 4UL)); // Read action was triggered.
+
+            Assert.AreEqual(MemoryPermission.Read, protection); // Write protection is still present.
+
+            readTrackingTriggered = null;
+
+            // Now, perform a write.
+
+            _tracking.VirtualMemoryEvent(0, 4, true);
+
+            bool dirtyAfterWriteAfterRead = handle.Dirty;
+            Assert.True(dirtyAfterWriteAfterRead); // Should be dirty.
+
+            Assert.AreEqual(MemoryPermission.ReadAndWrite, protection); // All protection is now be removed from the memory.
+
+            Assert.IsNull(readTrackingTriggered); // Read tracking was removed when the action fired, as it can only fire once.
+
+            handle.Dispose();
+        }
     }
 }

Ryujinx.Memory/Range/RangeList.cs

@@ -12,16 +12,16 @@ namespace Ryujinx.Memory.Range
     {
         private const int ArrayGrowthSize = 32;
 
-        private readonly List<T> _items;
+        protected readonly List<T> Items;
 
-        public int Count => _items.Count;
+        public int Count => Items.Count;
 
         /// <summary>
         /// Creates a new range list.
         /// </summary>
         public RangeList()
         {
-            _items = new List<T>();
+            Items = new List<T>();
         }
 
         /// <summary>
@@ -37,7 +37,7 @@ namespace Ryujinx.Memory.Range
                 index = ~index;
             }
 
-            _items.Insert(index, item);
+            Items.Insert(index, item);
         }
 
         /// <summary>
@@ -51,21 +51,21 @@ namespace Ryujinx.Memory.Range
 
             if (index >= 0)
             {
-                while (index > 0 && _items[index - 1].Address == item.Address)
+                while (index > 0 && Items[index - 1].Address == item.Address)
                 {
                     index--;
                 }
 
-                while (index < _items.Count)
+                while (index < Items.Count)
                 {
-                    if (_items[index].Equals(item))
+                    if (Items[index].Equals(item))
                     {
-                        _items.RemoveAt(index);
+                        Items.RemoveAt(index);
 
                         return true;
                     }
 
-                    if (_items[index].Address > item.Address)
+                    if (Items[index].Address > item.Address)
                     {
                         break;
                     }
@@ -110,7 +110,7 @@ namespace Ryujinx.Memory.Range
                 return default(T);
             }
 
-            return _items[index];
+            return Items[index];
         }
 
         /// <summary>
@@ -137,7 +137,7 @@ namespace Ryujinx.Memory.Range
 
             ulong endAddress = address + size;
 
-            foreach (T item in _items)
+            foreach (T item in Items)
             {
                 if (item.Address >= endAddress)
                 {
@@ -196,7 +196,7 @@ namespace Ryujinx.Memory.Range
 
             if (index >= 0)
             {
-                while (index > 0 && _items[index - 1].OverlapsWith(address, size))
+                while (index > 0 && Items[index - 1].OverlapsWith(address, size))
                 {
                     index--;
                 }
@@ -208,9 +208,9 @@ namespace Ryujinx.Memory.Range
                         Array.Resize(ref output, outputIndex + ArrayGrowthSize);
                     }
 
-                    output[outputIndex++] = _items[index++];
+                    output[outputIndex++] = Items[index++];
                 }
-                while (index < _items.Count && _items[index].OverlapsWith(address, size));
+                while (index < Items.Count && Items[index].OverlapsWith(address, size));
             }
 
             return outputIndex;
@@ -230,14 +230,14 @@ namespace Ryujinx.Memory.Range
 
             if (index >= 0)
             {
-                while (index > 0 && _items[index - 1].Address == address)
+                while (index > 0 && Items[index - 1].Address == address)
                 {
                     index--;
                 }
 
-                while (index < _items.Count)
+                while (index < Items.Count)
                 {
-                    T overlap = _items[index++];
+                    T overlap = Items[index++];
 
                     if (overlap.Address != address)
                     {
@@ -264,7 +264,7 @@ namespace Ryujinx.Memory.Range
         private int BinarySearch(ulong address)
         {
             int left  = 0;
-            int right = _items.Count - 1;
+            int right = Items.Count - 1;
 
             while (left <= right)
             {
@@ -272,7 +272,7 @@ namespace Ryujinx.Memory.Range
 
                 int middle = left + (range >> 1);
 
-                T item = _items[middle];
+                T item = Items[middle];
 
                 if (item.Address == address)
                 {
@@ -301,7 +301,7 @@ namespace Ryujinx.Memory.Range
         private int BinarySearch(ulong address, ulong size)
         {
             int left  = 0;
-            int right = _items.Count - 1;
+            int right = Items.Count - 1;
 
             while (left <= right)
             {
@@ -309,7 +309,7 @@ namespace Ryujinx.Memory.Range
 
                 int middle = left + (range >> 1);
 
-                T item = _items[middle];
+                T item = Items[middle];
 
                 if (item.OverlapsWith(address, size))
                 {
@@ -331,12 +331,12 @@ namespace Ryujinx.Memory.Range
 
         public IEnumerator<T> GetEnumerator()
         {
-            return _items.GetEnumerator();
+            return Items.GetEnumerator();
         }
 
         IEnumerator IEnumerable.GetEnumerator()
         {
-            return _items.GetEnumerator();
+            return Items.GetEnumerator();
         }
     }
 }

Ryujinx.Memory/Tracking/MultiRegionHandle.cs

@@ -123,6 +123,17 @@ namespace Ryujinx.Memory.Tracking
             }
         }
 
+        public void RegisterAction(ulong address, ulong size, RegionSignal action)
+        {
+            int startHandle = (int)((address - Address) / Granularity);
+            int lastHandle = (int)((address + (size - 1) - Address) / Granularity);
+
+            for (int i = startHandle; i <= lastHandle; i++)
+            {
+                _handles[i].RegisterAction(action);
+            }
+        }
+
         public void Dispose()
         {
             foreach (var handle in _handles)

Ryujinx.Memory/Tracking/RegionHandle.cs

@@ -24,6 +24,7 @@ namespace Ryujinx.Memory.Tracking
         private readonly MemoryTracking _tracking;
 
         internal MemoryPermission RequiredPermission => _preAction != null ? MemoryPermission.None : (Dirty ? MemoryPermission.ReadAndWrite : MemoryPermission.Read);
+        internal RegionSignal PreAction => _preAction;
 
         /// <summary>
         /// Create a new region handle. The handle is registered with the given tracking object,

Ryujinx.Memory/Tracking/SmartMultiRegionHandle.cs

@@ -41,6 +41,17 @@ namespace Ryujinx.Memory.Tracking
             Dirty = true;
         }
 
+        public void RegisterAction(RegionSignal action)
+        {
+            foreach (var handle in _handles)
+            {
+                if (handle != null)
+                {
+                    handle?.RegisterAction((address, size) => action(handle.Address, handle.Size));
+                }
+            }
+        }
+
         public void QueryModified(Action<ulong, ulong> modifiedAction)
         {
             if (!Dirty)
@@ -66,14 +77,23 @@ namespace Ryujinx.Memory.Tracking
             ulong size = HandlesToBytes(splitIndex - handleIndex);
 
             // First, the target handle must be removed. Its data can still be used to determine the new handles.
+            RegionSignal signal = handle.PreAction;
             handle.Dispose();
 
             RegionHandle splitLow = _tracking.BeginTracking(address, size);
             splitLow.Parent = this;
+            if (signal != null)
+            {
+                splitLow.RegisterAction(signal);
+            }
             _handles[handleIndex] = splitLow;
 
             RegionHandle splitHigh = _tracking.BeginTracking(address + size, handle.Size - size);
             splitHigh.Parent = this;
+            if (signal != null)
+            {
+                splitHigh.RegisterAction(signal);
+            }
             _handles[splitIndex] = splitHigh;
         }
 

Ryujinx.Memory/Tracking/VirtualRegion.cs

@@ -22,12 +22,12 @@ namespace Ryujinx.Memory.Tracking
 
         public override void Signal(ulong address, ulong size, bool write)
         {
-            _tracking.ProtectVirtualRegion(this, MemoryPermission.ReadAndWrite); // Remove our protection immedately.
-
             foreach (var handle in Handles)
             {
                 handle.Signal(address, size, write);
             }
+
+            UpdateProtection();
         }
 
         /// <summary>