Fix VIC out of bounds copy (#3386)

* Fix VIC out of bounds copy

* Update the assert

ARMeilleure/IntermediateRepresentation/Operand.cs

@@ -14,10 +14,11 @@ namespace ARMeilleure.IntermediateRepresentation
             public byte Kind;
             public byte Type;
             public byte SymbolType;
+            public byte Padding; // Unused space.
             public ushort AssignmentsCount;
             public ushort AssignmentsCapacity;
-            public ushort UsesCount;
-            public ushort UsesCapacity;
+            public uint UsesCount;
+            public uint UsesCapacity;
             public Operation* Assignments;
             public Operation* Uses;
             public ulong Value;
@@ -84,11 +85,11 @@ namespace ARMeilleure.IntermediateRepresentation
             {
                 Debug.Assert(Kind != OperandKind.Memory);
 
-                return new ReadOnlySpan<Operation>(_data->Uses, _data->UsesCount);
+                return new ReadOnlySpan<Operation>(_data->Uses, (int)_data->UsesCount);
             }
         }
 
-        public int UsesCount => _data->UsesCount;
+        public int UsesCount => (int)_data->UsesCount;
         public int AssignmentsCount => _data->AssignmentsCount;
 
         public bool Relocatable => Symbol.Type != SymbolType.None;
@@ -178,7 +179,7 @@ namespace ARMeilleure.IntermediateRepresentation
                 {
                     Add(operation, ref addr._data->Assignments, ref addr._data->AssignmentsCount, ref addr._data->AssignmentsCapacity);
                 }
-                
+
                 if (index != default)
                 {
                     Add(operation, ref index._data->Assignments, ref index._data->AssignmentsCount, ref index._data->AssignmentsCapacity);
@@ -265,6 +266,13 @@ namespace ARMeilleure.IntermediateRepresentation
             data = Allocators.References.Allocate<T>(initialCapacity);
         }
 
+        private static void New<T>(ref T* data, ref uint count, ref uint capacity, uint initialCapacity) where T : unmanaged
+        {
+            count = 0;
+            capacity = initialCapacity;
+            data = Allocators.References.Allocate<T>(initialCapacity);
+        }
+
         private static void Add<T>(T item, ref T* data, ref ushort count, ref ushort capacity) where T : unmanaged
         {
             if (count < capacity)
@@ -294,6 +302,40 @@ namespace ARMeilleure.IntermediateRepresentation
             }
         }
 
+        private static void Add<T>(T item, ref T* data, ref uint count, ref uint capacity) where T : unmanaged
+        {
+            if (count < capacity)
+            {
+                data[count++] = item;
+
+                return;
+            }
+
+            // Could not add item in the fast path, fallback onto the slow path.
+            ExpandAdd(item, ref data, ref count, ref capacity);
+
+            static void ExpandAdd(T item, ref T* data, ref uint count, ref uint capacity)
+            {
+                uint newCount = checked(count + 1);
+                uint newCapacity = (uint)Math.Min(capacity * 2, int.MaxValue);
+
+                if (newCapacity <= capacity)
+                {
+                    throw new OverflowException();
+                }
+
+                var oldSpan = new Span<T>(data, (int)count);
+
+                capacity = newCapacity;
+                data = Allocators.References.Allocate<T>(capacity);
+
+                oldSpan.CopyTo(new Span<T>(data, (int)count));
+
+                data[count] = item;
+                count = newCount;
+            }
+        }
+
         private static void Remove<T>(in T item, ref T* data, ref ushort count) where T : unmanaged
         {
             var span = new Span<T>(data, count);
@@ -314,6 +356,26 @@ namespace ARMeilleure.IntermediateRepresentation
             }
         }
 
+        private static void Remove<T>(in T item, ref T* data, ref uint count) where T : unmanaged
+        {
+            var span = new Span<T>(data, (int)count);
+
+            for (int i = 0; i < span.Length; i++)
+            {
+                if (EqualityComparer<T>.Default.Equals(span[i], item))
+                {
+                    if (i + 1 < count)
+                    {
+                        span.Slice(i + 1).CopyTo(span.Slice(i));
+                    }
+
+                    count--;
+
+                    return;
+                }
+            }
+        }
+
         public override int GetHashCode()
         {
             if (Kind == OperandKind.LocalVariable)

Ryujinx.Graphics.GAL/IPipeline.cs

@@ -10,9 +10,10 @@ namespace Ryujinx.Graphics.GAL
 
         void ClearBuffer(BufferHandle destination, int offset, int size, uint value);
 
-        void ClearRenderTargetColor(int index, uint componentMask, ColorF color);
+        void ClearRenderTargetColor(int index, int layer, uint componentMask, ColorF color);
 
         void ClearRenderTargetDepthStencil(
+            int layer,
             float depthValue,
             bool depthMask,
             int stencilValue,

Ryujinx.Graphics.GAL/Multithreading/Commands/ClearRenderTargetColorCommand.cs

@@ -4,19 +4,21 @@
     {
         public CommandType CommandType => CommandType.ClearRenderTargetColor;
         private int _index;
+        private int _layer;
         private uint _componentMask;
         private ColorF _color;
 
-        public void Set(int index, uint componentMask, ColorF color)
+        public void Set(int index, int layer, uint componentMask, ColorF color)
         {
             _index = index;
+            _layer = layer;
             _componentMask = componentMask;
             _color = color;
         }
 
         public static void Run(ref ClearRenderTargetColorCommand command, ThreadedRenderer threaded, IRenderer renderer)
         {
-            renderer.Pipeline.ClearRenderTargetColor(command._index, command._componentMask, command._color);
+            renderer.Pipeline.ClearRenderTargetColor(command._index, command._layer, command._componentMask, command._color);
         }
     }
 }

Ryujinx.Graphics.GAL/Multithreading/Commands/ClearRenderTargetDepthStencilCommand.cs

@@ -3,13 +3,15 @@
     struct ClearRenderTargetDepthStencilCommand : IGALCommand
     {
         public CommandType CommandType => CommandType.ClearRenderTargetDepthStencil;
+        private int _layer;
         private float _depthValue;
         private bool _depthMask;
         private int _stencilValue;
         private int _stencilMask;
 
-        public void Set(float depthValue, bool depthMask, int stencilValue, int stencilMask)
+        public void Set(int layer, float depthValue, bool depthMask, int stencilValue, int stencilMask)
         {
+            _layer = layer;
             _depthValue = depthValue;
             _depthMask = depthMask;
             _stencilValue = stencilValue;
@@ -18,7 +20,7 @@
 
         public static void Run(ref ClearRenderTargetDepthStencilCommand command, ThreadedRenderer threaded, IRenderer renderer)
         {
-            renderer.Pipeline.ClearRenderTargetDepthStencil(command._depthValue, command._depthMask, command._stencilValue, command._stencilMask);
+            renderer.Pipeline.ClearRenderTargetDepthStencil(command._layer, command._depthValue, command._depthMask, command._stencilValue, command._stencilMask);
         }
     }
 }

Ryujinx.Graphics.GAL/Multithreading/ThreadedPipeline.cs

@@ -40,15 +40,15 @@ namespace Ryujinx.Graphics.GAL.Multithreading
             _renderer.QueueCommand();
         }
 
-        public void ClearRenderTargetColor(int index, uint componentMask, ColorF color)
+        public void ClearRenderTargetColor(int index, int layer, uint componentMask, ColorF color)
         {
-            _renderer.New<ClearRenderTargetColorCommand>().Set(index, componentMask, color);
+            _renderer.New<ClearRenderTargetColorCommand>().Set(index, layer, componentMask, color);
             _renderer.QueueCommand();
         }
 
-        public void ClearRenderTargetDepthStencil(float depthValue, bool depthMask, int stencilValue, int stencilMask)
+        public void ClearRenderTargetDepthStencil(int layer, float depthValue, bool depthMask, int stencilValue, int stencilMask)
         {
-            _renderer.New<ClearRenderTargetDepthStencilCommand>().Set(depthValue, depthMask, stencilValue, stencilMask);
+            _renderer.New<ClearRenderTargetDepthStencilCommand>().Set(layer, depthValue, depthMask, stencilValue, stencilMask);
             _renderer.QueueCommand();
         }
 

Ryujinx.Graphics.Gpu/Engine/Threed/DrawManager.cs

@@ -18,6 +18,7 @@ namespace Ryujinx.Graphics.Gpu.Engine.Threed
 
         private bool _instancedDrawPending;
         private bool _instancedIndexed;
+        private bool _instancedIndexedInline;
 
         private int _instancedFirstIndex;
         private int _instancedFirstVertex;
@@ -134,13 +135,16 @@ namespace Ryujinx.Graphics.Gpu.Engine.Threed
             {
                 _instancedDrawPending = true;
 
+                int ibCount = _drawState.IbStreamer.InlineIndexCount;
+
                 _instancedIndexed = _drawState.DrawIndexed;
+                _instancedIndexedInline = ibCount != 0;
 
                 _instancedFirstIndex = firstIndex;
                 _instancedFirstVertex = (int)_state.State.FirstVertex;
                 _instancedFirstInstance = (int)_state.State.FirstInstance;
 
-                _instancedIndexCount = indexCount;
+                _instancedIndexCount = ibCount != 0 ? ibCount : indexCount;
 
                 var drawState = _state.State.VertexBufferDrawState;
 
@@ -451,8 +455,18 @@ namespace Ryujinx.Graphics.Gpu.Engine.Threed
             {
                 _instancedDrawPending = false;
 
-                if (_instancedIndexed)
+                bool indexedInline = _instancedIndexedInline;
+
+                if (_instancedIndexed || indexedInline)
                 {
+                    if (indexedInline)
+                    {
+                        int inlineIndexCount = _drawState.IbStreamer.GetAndResetInlineIndexCount();
+                        BufferRange br = new BufferRange(_drawState.IbStreamer.GetInlineIndexBuffer(), 0, inlineIndexCount * 4);
+
+                        _channel.BufferManager.SetIndexBuffer(br, IndexType.UInt);
+                    }
+
                     _context.Renderer.Pipeline.DrawIndexed(
                         _instancedIndexCount,
                         _instanceIndex + 1,
@@ -491,8 +505,9 @@ namespace Ryujinx.Graphics.Gpu.Engine.Threed
             }
 
             int index = (argument >> 6) & 0xf;
+            int layer = (argument >> 10) & 0x3ff;
 
-            engine.UpdateRenderTargetState(useControl: false, singleUse: index);
+            engine.UpdateRenderTargetState(useControl: false, layered: layer != 0, singleUse: index);
 
             // If there is a mismatch on the host clip region and the one explicitly defined by the guest
             // on the screen scissor state, then we need to force only one texture to be bound to avoid
@@ -567,7 +582,7 @@ namespace Ryujinx.Graphics.Gpu.Engine.Threed
 
                 ColorF color = new ColorF(clearColor.Red, clearColor.Green, clearColor.Blue, clearColor.Alpha);
 
-                _context.Renderer.Pipeline.ClearRenderTargetColor(index, componentMask, color);
+                _context.Renderer.Pipeline.ClearRenderTargetColor(index, layer, componentMask, color);
             }
 
             if (clearDepth || clearStencil)
@@ -588,6 +603,7 @@ namespace Ryujinx.Graphics.Gpu.Engine.Threed
                 }
 
                 _context.Renderer.Pipeline.ClearRenderTargetDepthStencil(
+                    layer,
                     depthValue,
                     clearDepth,
                     stencilValue,

Ryujinx.Graphics.Gpu/Engine/Threed/IbStreamer.cs

@@ -20,6 +20,11 @@ namespace Ryujinx.Graphics.Gpu.Engine.Threed
         /// </summary>
         public bool HasInlineIndexData => _inlineIndexCount != 0;
 
+        /// <summary>
+        /// Total numbers of indices that have been pushed.
+        /// </summary>
+        public int InlineIndexCount => _inlineIndexCount;
+
         /// <summary>
         /// Gets the handle for the host buffer currently holding the inline index buffer data.
         /// </summary>

Ryujinx.Graphics.Gpu/Engine/Threed/StateUpdater.cs

@@ -362,8 +362,9 @@ namespace Ryujinx.Graphics.Gpu.Engine.Threed
         /// Updates render targets (color and depth-stencil buffers) based on current render target state.
         /// </summary>
         /// <param name="useControl">Use draw buffers information from render target control register</param>
+        /// <param name="layered">Indicates if the texture is layered</param>
         /// <param name="singleUse">If this is not -1, it indicates that only the given indexed target will be used.</param>
-        public void UpdateRenderTargetState(bool useControl, int singleUse = -1)
+        public void UpdateRenderTargetState(bool useControl, bool layered = false, int singleUse = -1)
         {
             var memoryManager = _channel.MemoryManager;
             var rtControl = _state.State.RtControl;
@@ -399,7 +400,7 @@ namespace Ryujinx.Graphics.Gpu.Engine.Threed
                 Image.Texture color = memoryManager.Physical.TextureCache.FindOrCreateTexture(
                     memoryManager,
                     colorState,
-                    _vtgWritesRtLayer,
+                    _vtgWritesRtLayer || layered,
                     samplesInX,
                     samplesInY,
                     sizeHint);
@@ -433,6 +434,7 @@ namespace Ryujinx.Graphics.Gpu.Engine.Threed
                     memoryManager,
                     dsState,
                     dsSize,
+                    _vtgWritesRtLayer || layered,
                     samplesInX,
                     samplesInY,
                     sizeHint);

Ryujinx.Graphics.Gpu/Engine/Threed/ThreedClass.cs

@@ -131,10 +131,11 @@ namespace Ryujinx.Graphics.Gpu.Engine.Threed
         /// Updates render targets (color and depth-stencil buffers) based on current render target state.
         /// </summary>
         /// <param name="useControl">Use draw buffers information from render target control register</param>
+        /// <param name="layered">Indicates if the texture is layered</param>
         /// <param name="singleUse">If this is not -1, it indicates that only the given indexed target will be used.</param>
-        public void UpdateRenderTargetState(bool useControl, int singleUse = -1)
+        public void UpdateRenderTargetState(bool useControl, bool layered = false, int singleUse = -1)
         {
-            _stateUpdater.UpdateRenderTargetState(useControl, singleUse);
+            _stateUpdater.UpdateRenderTargetState(useControl, layered, singleUse);
         }
 
         /// <summary>

Ryujinx.Graphics.Gpu/Image/TextureCache.cs

@@ -349,6 +349,7 @@ namespace Ryujinx.Graphics.Gpu.Image
         /// <param name="memoryManager">GPU memory manager where the texture is mapped</param>
         /// <param name="dsState">Depth-stencil buffer texture to find or create</param>
         /// <param name="size">Size of the depth-stencil texture</param>
+        /// <param name="layered">Indicates if the texture might be accessed with a non-zero layer index</param>
         /// <param name="samplesInX">Number of samples in the X direction, for MSAA</param>
         /// <param name="samplesInY">Number of samples in the Y direction, for MSAA</param>
         /// <param name="sizeHint">A hint indicating the minimum used size for the texture</param>
@@ -357,6 +358,7 @@ namespace Ryujinx.Graphics.Gpu.Image
             MemoryManager memoryManager,
             RtDepthStencilState dsState,
             Size3D size,
+            bool layered,
             int samplesInX,
             int samplesInY,
             Size sizeHint)
@@ -364,9 +366,24 @@ namespace Ryujinx.Graphics.Gpu.Image
             int gobBlocksInY = dsState.MemoryLayout.UnpackGobBlocksInY();
             int gobBlocksInZ = dsState.MemoryLayout.UnpackGobBlocksInZ();
 
-            Target target = (samplesInX | samplesInY) != 1
-                ? Target.Texture2DMultisample
-                : Target.Texture2D;
+            Target target;
+
+            if (dsState.MemoryLayout.UnpackIsTarget3D())
+            {
+                target = Target.Texture3D;
+            }
+            else if ((samplesInX | samplesInY) != 1)
+            {
+                target = size.Depth > 1 && layered
+                    ? Target.Texture2DMultisampleArray
+                    : Target.Texture2DMultisample;
+            }
+            else
+            {
+                target = size.Depth > 1 && layered
+                    ? Target.Texture2DArray
+                    : Target.Texture2D;
+            }
 
             FormatInfo formatInfo = dsState.Format.Convert();
 

Ryujinx.Graphics.OpenGL/Framebuffer.cs

@@ -9,10 +9,13 @@ namespace Ryujinx.Graphics.OpenGL
     class Framebuffer : IDisposable
     {
         public int Handle { get; private set; }
+        private int _clearFbHandle;
+        private bool _clearFbInitialized;
 
         private FramebufferAttachment _lastDsAttachment;
 
         private readonly TextureView[] _colors;
+        private TextureView _depthStencil;
 
         private int _colorsCount;
         private bool _dualSourceBlend;
@@ -20,6 +23,7 @@ namespace Ryujinx.Graphics.OpenGL
         public Framebuffer()
         {
             Handle = GL.GenFramebuffer();
+            _clearFbHandle = GL.GenFramebuffer();
 
             _colors = new TextureView[8];
         }
@@ -55,20 +59,7 @@ namespace Ryujinx.Graphics.OpenGL
 
             if (depthStencil != null)
             {
-                FramebufferAttachment attachment;
-
-                if (IsPackedDepthStencilFormat(depthStencil.Format))
-                {
-                    attachment = FramebufferAttachment.DepthStencilAttachment;
-                }
-                else if (IsDepthOnlyFormat(depthStencil.Format))
-                {
-                    attachment = FramebufferAttachment.DepthAttachment;
-                }
-                else
-                {
-                    attachment = FramebufferAttachment.StencilAttachment;
-                }
+                FramebufferAttachment attachment = GetAttachment(depthStencil.Format);
 
                 GL.FramebufferTexture(
                     FramebufferTarget.Framebuffer,
@@ -82,6 +73,8 @@ namespace Ryujinx.Graphics.OpenGL
             {
                 _lastDsAttachment = 0;
             }
+
+            _depthStencil = depthStencil;
         }
 
         public void SetDualSourceBlend(bool enable)
@@ -124,6 +117,22 @@ namespace Ryujinx.Graphics.OpenGL
             GL.DrawBuffers(colorsCount, drawBuffers);
         }
 
+        private static FramebufferAttachment GetAttachment(Format format)
+        {
+            if (IsPackedDepthStencilFormat(format))
+            {
+                return FramebufferAttachment.DepthStencilAttachment;
+            }
+            else if (IsDepthOnlyFormat(format))
+            {
+                return FramebufferAttachment.DepthAttachment;
+            }
+            else
+            {
+                return FramebufferAttachment.StencilAttachment;
+            }
+        }
+
         private static bool IsPackedDepthStencilFormat(Format format)
         {
             return format == Format.D24UnormS8Uint ||
@@ -136,6 +145,78 @@ namespace Ryujinx.Graphics.OpenGL
             return format == Format.D16Unorm || format == Format.D32Float;
         }
 
+        public void AttachColorLayerForClear(int index, int layer)
+        {
+            TextureView color = _colors[index];
+
+            if (!IsLayered(color))
+            {
+                return;
+            }
+
+            BindClearFb();
+            GL.FramebufferTextureLayer(FramebufferTarget.Framebuffer, FramebufferAttachment.ColorAttachment0 + index, color.Handle, 0, layer);
+        }
+
+        public void DetachColorLayerForClear(int index)
+        {
+            TextureView color = _colors[index];
+
+            if (!IsLayered(color))
+            {
+                return;
+            }
+
+            GL.FramebufferTexture(FramebufferTarget.Framebuffer, FramebufferAttachment.ColorAttachment0 + index, 0, 0);
+            Bind();
+        }
+
+        public void AttachDepthStencilLayerForClear(int layer)
+        {
+            TextureView depthStencil = _depthStencil;
+
+            if (!IsLayered(depthStencil))
+            {
+                return;
+            }
+
+            BindClearFb();
+            GL.FramebufferTextureLayer(FramebufferTarget.Framebuffer, GetAttachment(depthStencil.Format), depthStencil.Handle, 0, layer);
+        }
+
+        public void DetachDepthStencilLayerForClear()
+        {
+            TextureView depthStencil = _depthStencil;
+
+            if (!IsLayered(depthStencil))
+            {
+                return;
+            }
+
+            GL.FramebufferTexture(FramebufferTarget.Framebuffer, GetAttachment(depthStencil.Format), 0, 0);
+            Bind();
+        }
+
+        private void BindClearFb()
+        {
+            GL.BindFramebuffer(FramebufferTarget.Framebuffer, _clearFbHandle);
+
+            if (!_clearFbInitialized)
+            {
+                SetDrawBuffersImpl(Constants.MaxRenderTargets);
+                _clearFbInitialized = true;
+            }
+        }
+
+        private static bool IsLayered(TextureView view)
+        {
+            return view != null &&
+                   view.Target != Target.Texture1D &&
+                   view.Target != Target.Texture2D &&
+                   view.Target != Target.Texture2DMultisample &&
+                   view.Target != Target.TextureBuffer;
+        }
+
         public void Dispose()
         {
             if (Handle != 0)
@@ -144,6 +225,13 @@ namespace Ryujinx.Graphics.OpenGL
 
                 Handle = 0;
             }
+
+            if (_clearFbHandle != 0)
+            {
+                GL.DeleteFramebuffer(_clearFbHandle);
+
+                _clearFbHandle = 0;
+            }
         }
     }
 }

Ryujinx.Graphics.OpenGL/Pipeline.cs

@@ -110,7 +110,7 @@ namespace Ryujinx.Graphics.OpenGL
             Buffer.Clear(destination, offset, size, value);
         }
 
-        public void ClearRenderTargetColor(int index, uint componentMask, ColorF color)
+        public void ClearRenderTargetColor(int index, int layer, uint componentMask, ColorF color)
         {
             GL.ColorMask(
                 index,
@@ -119,14 +119,18 @@ namespace Ryujinx.Graphics.OpenGL
                 (componentMask & 4) != 0,
                 (componentMask & 8) != 0);
 
+            _framebuffer.AttachColorLayerForClear(index, layer);
+
             float[] colors = new float[] { color.Red, color.Green, color.Blue, color.Alpha };
 
             GL.ClearBuffer(OpenTK.Graphics.OpenGL.ClearBuffer.Color, index, colors);
 
+            _framebuffer.DetachColorLayerForClear(index);
+
             RestoreComponentMask(index);
         }
 
-        public void ClearRenderTargetDepthStencil(float depthValue, bool depthMask, int stencilValue, int stencilMask)
+        public void ClearRenderTargetDepthStencil(int layer, float depthValue, bool depthMask, int stencilValue, int stencilMask)
         {
             bool stencilMaskChanged =
                 stencilMask != 0 &&
@@ -144,6 +148,8 @@ namespace Ryujinx.Graphics.OpenGL
                 GL.DepthMask(depthMask);
             }
 
+            _framebuffer.AttachDepthStencilLayerForClear(layer);
+
             if (depthMask && stencilMask != 0)
             {
                 GL.ClearBuffer(ClearBufferCombined.DepthStencil, 0, depthValue, stencilValue);
@@ -157,6 +163,8 @@ namespace Ryujinx.Graphics.OpenGL
                 GL.ClearBuffer(OpenTK.Graphics.OpenGL.ClearBuffer.Stencil, 0, ref stencilValue);
             }
 
+            _framebuffer.DetachDepthStencilLayerForClear();
+
             if (stencilMaskChanged)
             {
                 GL.StencilMaskSeparate(StencilFace.Front, _stencilFrontMask);
@@ -597,6 +605,8 @@ namespace Ryujinx.Graphics.OpenGL
                         GL.EndTransformFeedback();
                     }
 
+                    GL.ClipControl(ClipOrigin.UpperLeft, ClipDepthMode.NegativeOneToOne);
+
                     _drawTexture.Draw(
                         view,
                         samp,
@@ -627,6 +637,8 @@ namespace Ryujinx.Graphics.OpenGL
                     {
                         GL.BeginTransformFeedback(_tfTopology);
                     }
+
+                    RestoreClipControl();
                 }
             }
         }

Ryujinx.Graphics.Vic/Blender.cs

@@ -10,17 +10,22 @@ namespace Ryujinx.Graphics.Vic
 {
     static class Blender
     {
-        public static void BlendOne(Surface dst, Surface src, ref SlotStruct slot)
+        public static void BlendOne(Surface dst, Surface src, ref SlotStruct slot, Rectangle targetRect)
         {
-            if (Sse41.IsSupported && (dst.Width & 3) == 0)
+            int x1 = targetRect.X;
+            int y1 = targetRect.Y;
+            int x2 = Math.Min(src.Width, x1 + targetRect.Width);
+            int y2 = Math.Min(src.Height, y1 + targetRect.Height);
+
+            if (Sse41.IsSupported && ((x1 | x2) & 3) == 0)
             {
-                BlendOneSse41(dst, src, ref slot);
+                BlendOneSse41(dst, src, ref slot, x1, y1, x2, y2);
                 return;
             }
 
-            for (int y = 0; y < dst.Height; y++)
+            for (int y = y1; y < y2; y++)
             {
-                for (int x = 0; x < dst.Width; x++)
+                for (int x = x1; x < x2; x++)
                 {
                     int inR = src.GetR(x, y);
                     int inG = src.GetG(x, y);
@@ -40,9 +45,9 @@ namespace Ryujinx.Graphics.Vic
             }
         }
 
-        private unsafe static void BlendOneSse41(Surface dst, Surface src, ref SlotStruct slot)
+        private unsafe static void BlendOneSse41(Surface dst, Surface src, ref SlotStruct slot, int x1, int y1, int x2, int y2)
         {
-            Debug.Assert((dst.Width & 3) == 0);
+            Debug.Assert(((x1 | x2) & 3) == 0);
 
             ref MatrixStruct mtx = ref slot.ColorMatrixStruct;
 
@@ -62,9 +67,9 @@ namespace Ryujinx.Graphics.Vic
                 Pixel* ip = srcPtr;
                 Pixel* op = dstPtr;
 
-                for (int y = 0; y < dst.Height; y++, ip += src.Width, op += dst.Width)
+                for (int y = y1; y < y2; y++, ip += src.Width, op += dst.Width)
                 {
-                    for (int x = 0; x < dst.Width; x += 4)
+                    for (int x = x1; x < x2; x += 4)
                     {
                         Vector128<int> pixel1 = Sse41.ConvertToVector128Int32((ushort*)(ip + (uint)x));
                         Vector128<int> pixel2 = Sse41.ConvertToVector128Int32((ushort*)(ip + (uint)x + 1));

Ryujinx.Graphics.Vic/Rectangle.cs

@@ -0,0 +1,18 @@
+namespace Ryujinx.Graphics.Vic
+{
+    struct Rectangle
+    {
+        public readonly int X;
+        public readonly int Y;
+        public readonly int Width;
+        public readonly int Height;
+
+        public Rectangle(int x, int y, int width, int height)
+        {
+            X = x;
+            Y = y;
+            Width = width;
+            Height = height;
+        }
+    }
+}

Ryujinx.Graphics.Vic/VicDevice.cs

@@ -2,6 +2,7 @@
 using Ryujinx.Graphics.Gpu.Memory;
 using Ryujinx.Graphics.Vic.Image;
 using Ryujinx.Graphics.Vic.Types;
+using System;
 using System.Collections.Generic;
 
 namespace Ryujinx.Graphics.Vic
@@ -47,7 +48,19 @@ namespace Ryujinx.Graphics.Vic
 
                 using Surface src = SurfaceReader.Read(_rm, ref slot.SlotConfig, ref slot.SlotSurfaceConfig, ref offsets);
 
-                Blender.BlendOne(output, src, ref slot);
+                int x1 = config.OutputConfig.TargetRectLeft;
+                int y1 = config.OutputConfig.TargetRectTop;
+                int x2 = config.OutputConfig.TargetRectRight + 1;
+                int y2 = config.OutputConfig.TargetRectBottom + 1;
+
+                int targetX = Math.Min(x1, x2);
+                int targetY = Math.Min(y1, y2);
+                int targetW = Math.Min(output.Width - targetX, Math.Abs(x2 - x1));
+                int targetH = Math.Min(output.Height - targetY, Math.Abs(y2 - y1));
+
+                Rectangle targetRect = new Rectangle(targetX, targetY, targetW, targetH);
+
+                Blender.BlendOne(output, src, ref slot, targetRect);
             }
 
             SurfaceWriter.Write(_rm, output, ref config.OutputSurfaceConfig, ref _state.State.SetOutputSurface);

Ryujinx.HLE/HOS/Kernel/Process/KProcess.cs

@@ -735,11 +735,12 @@ namespace Ryujinx.HLE.HOS.Kernel.Process
             ulong argsPtr,
             ulong stackTop,
             int priority,
-            int cpuCore)
+            int cpuCore,
+            ThreadStart customThreadStart = null)
         {
             lock (_processLock)
             {
-                return thread.Initialize(entrypoint, argsPtr, stackTop, priority, cpuCore, this, ThreadType.User, null);
+                return thread.Initialize(entrypoint, argsPtr, stackTop, priority, cpuCore, this, ThreadType.User, customThreadStart);
             }
         }
 

Ryujinx.HLE/HOS/Kernel/SupervisorCall/Syscall.cs

@@ -2350,6 +2350,18 @@ namespace Ryujinx.HLE.HOS.Kernel.SupervisorCall
             [PointerSized] ulong stackTop,
             int priority,
             int cpuCore)
+        {
+            return CreateThread(out handle, entrypoint, argsPtr, stackTop, priority, cpuCore, null);
+        }
+
+        public KernelResult CreateThread(
+            out int handle,
+            ulong entrypoint,
+            ulong argsPtr,
+            ulong stackTop,
+            int priority,
+            int cpuCore,
+            ThreadStart customThreadStart)
         {
             handle = 0;
 
@@ -2386,7 +2398,8 @@ namespace Ryujinx.HLE.HOS.Kernel.SupervisorCall
                 argsPtr,
                 stackTop,
                 priority,
-                cpuCore);
+                cpuCore,
+                customThreadStart);
 
             if (result == KernelResult.Success)
             {

Ryujinx.HLE/HOS/Services/ServerBase.cs

@@ -1,3 +1,4 @@
+using Ryujinx.Common.Logging;
 using Ryujinx.HLE.HOS.Ipc;
 using Ryujinx.HLE.HOS.Kernel;
 using Ryujinx.HLE.HOS.Kernel.Common;
@@ -38,15 +39,15 @@ namespace Ryujinx.HLE.HOS.Services
         private readonly Dictionary<int, Func<IpcService>> _ports = new Dictionary<int, Func<IpcService>>();
 
         public ManualResetEvent InitDone { get; }
-        public Func<IpcService> SmObjectFactory { get; }
         public string Name { get; }
+        public Func<IpcService> SmObjectFactory { get; }
 
         public ServerBase(KernelContext context, string name, Func<IpcService> smObjectFactory = null)
         {
             InitDone = new ManualResetEvent(false);
+            _context = context;
             Name = name;
             SmObjectFactory = smObjectFactory;
-            _context = context;
 
             const ProcessCreationFlags flags =
                 ProcessCreationFlags.EnableAslr |
@@ -56,7 +57,7 @@ namespace Ryujinx.HLE.HOS.Services
 
             ProcessCreationInfo creationInfo = new ProcessCreationInfo("Service", 1, 0, 0x8000000, 1, flags, 0, 0);
 
-            KernelStatic.StartInitialProcess(context, creationInfo, DefaultCapabilities, 44, ServerLoop);
+            KernelStatic.StartInitialProcess(context, creationInfo, DefaultCapabilities, 44, Main);
         }
 
         private void AddPort(int serverPortHandle, Func<IpcService> objectFactory)
@@ -80,6 +81,11 @@ namespace Ryujinx.HLE.HOS.Services
             _sessions.Add(serverSessionHandle, obj);
         }
 
+        private void Main()
+        {
+            ServerLoop();
+        }
+
         private void ServerLoop()
         {
             _selfProcess = KernelStatic.GetCurrentProcess();

Ryujinx.HLE/HOS/Services/Sockets/Bsd/Impl/EventFileDescriptor.cs

@@ -8,7 +8,6 @@ namespace Ryujinx.HLE.HOS.Services.Sockets.Bsd
     {
         private ulong _value;
         private readonly EventFdFlags _flags;
-        private AutoResetEvent _event;
 
         private object _lock = new object();
 
@@ -19,9 +18,13 @@ namespace Ryujinx.HLE.HOS.Services.Sockets.Bsd
 
         public EventFileDescriptor(ulong value, EventFdFlags flags)
         {
+            // FIXME: We should support blocking operations.
+            // Right now they can't be supported because it would cause the
+            // service to lock up as we only have one thread processing requests.
+            flags |= EventFdFlags.NonBlocking;
+
             _value = value;
             _flags = flags;
-            _event = new AutoResetEvent(false);
 
             WriteEvent = new ManualResetEvent(true);
             ReadEvent = new ManualResetEvent(true);
@@ -31,7 +34,6 @@ namespace Ryujinx.HLE.HOS.Services.Sockets.Bsd
 
         public void Dispose()
         {
-            _event.Dispose();
             WriteEvent.Dispose();
             ReadEvent.Dispose();
         }
@@ -57,7 +59,7 @@ namespace Ryujinx.HLE.HOS.Services.Sockets.Bsd
                     {
                         while (_value == 0)
                         {
-                            _event.WaitOne();
+                            Monitor.Wait(_lock);
                         }
                     }
                     else
@@ -106,7 +108,7 @@ namespace Ryujinx.HLE.HOS.Services.Sockets.Bsd
                 {
                     if (Blocking)
                     {
-                        _event.WaitOne();
+                        Monitor.Wait(_lock);
                     }
                     else
                     {
@@ -119,7 +121,7 @@ namespace Ryujinx.HLE.HOS.Services.Sockets.Bsd
                 writeSize = sizeof(ulong);
 
                 _value += count;
-                _event.Set();
+                Monitor.Pulse(_lock);
 
                 WriteEvent.Set();
 

Ryujinx.Memory/MemoryBlock.cs

@@ -19,6 +19,8 @@ namespace Ryujinx.Memory
         private ConcurrentDictionary<MemoryBlock, byte> _viewStorages;
         private int _viewCount;
 
+        internal bool ForceWindows4KBView => _forceWindows4KBView;
+
         /// <summary>
         /// Pointer to the memory block data.
         /// </summary>
@@ -145,7 +147,7 @@ namespace Ryujinx.Memory
                 srcBlock.IncrementViewCount();
             }
 
-            MemoryManagement.MapView(srcBlock._sharedMemory, srcOffset, GetPointerInternal(dstOffset, size), size, _forceWindows4KBView);
+            MemoryManagement.MapView(srcBlock._sharedMemory, srcOffset, GetPointerInternal(dstOffset, size), size, this);
         }
 
         /// <summary>
@@ -156,7 +158,7 @@ namespace Ryujinx.Memory
         /// <param name="size">Size of the range to be unmapped</param>
         public void UnmapView(MemoryBlock srcBlock, ulong offset, ulong size)
         {
-            MemoryManagement.UnmapView(srcBlock._sharedMemory, GetPointerInternal(offset, size), size, _forceWindows4KBView);
+            MemoryManagement.UnmapView(srcBlock._sharedMemory, GetPointerInternal(offset, size), size, this);
         }
 
         /// <summary>

Ryujinx.Memory/MemoryManagement.cs

@@ -68,17 +68,17 @@ namespace Ryujinx.Memory
             }
         }
 
-        public static void MapView(IntPtr sharedMemory, ulong srcOffset, IntPtr address, ulong size, bool force4KBMap)
+        public static void MapView(IntPtr sharedMemory, ulong srcOffset, IntPtr address, ulong size, MemoryBlock owner)
         {
             if (OperatingSystem.IsWindows())
             {
-                if (force4KBMap)
+                if (owner.ForceWindows4KBView)
                 {
                     MemoryManagementWindows.MapView4KB(sharedMemory, srcOffset, address, (IntPtr)size);
                 }
                 else
                 {
-                    MemoryManagementWindows.MapView(sharedMemory, srcOffset, address, (IntPtr)size);
+                    MemoryManagementWindows.MapView(sharedMemory, srcOffset, address, (IntPtr)size, owner);
                 }
             }
             else if (OperatingSystem.IsLinux() || OperatingSystem.IsMacOS())
@@ -91,17 +91,17 @@ namespace Ryujinx.Memory
             }
         }
 
-        public static void UnmapView(IntPtr sharedMemory, IntPtr address, ulong size, bool force4KBMap)
+        public static void UnmapView(IntPtr sharedMemory, IntPtr address, ulong size, MemoryBlock owner)
         {
             if (OperatingSystem.IsWindows())
             {
-                if (force4KBMap)
+                if (owner.ForceWindows4KBView)
                 {
                     MemoryManagementWindows.UnmapView4KB(address, (IntPtr)size);
                 }
                 else
                 {
-                    MemoryManagementWindows.UnmapView(sharedMemory, address, (IntPtr)size);
+                    MemoryManagementWindows.UnmapView(sharedMemory, address, (IntPtr)size, owner);
                 }
             }
             else if (OperatingSystem.IsLinux() || OperatingSystem.IsMacOS())

Ryujinx.Memory/MemoryManagementWindows.cs

@@ -68,9 +68,9 @@ namespace Ryujinx.Memory
             return WindowsApi.VirtualFree(location, size, AllocationType.Decommit);
         }
 
-        public static void MapView(IntPtr sharedMemory, ulong srcOffset, IntPtr location, IntPtr size)
+        public static void MapView(IntPtr sharedMemory, ulong srcOffset, IntPtr location, IntPtr size, MemoryBlock owner)
         {
-            _placeholders.MapView(sharedMemory, srcOffset, location, size);
+            _placeholders.MapView(sharedMemory, srcOffset, location, size, owner);
         }
 
         public static void MapView4KB(IntPtr sharedMemory, ulong srcOffset, IntPtr location, IntPtr size)
@@ -106,9 +106,9 @@ namespace Ryujinx.Memory
             }
         }
 
-        public static void UnmapView(IntPtr sharedMemory, IntPtr location, IntPtr size)
+        public static void UnmapView(IntPtr sharedMemory, IntPtr location, IntPtr size, MemoryBlock owner)
         {
-            _placeholders.UnmapView(sharedMemory, location, size);
+            _placeholders.UnmapView(sharedMemory, location, size, owner);
         }
 
         public static void UnmapView4KB(IntPtr location, IntPtr size)
@@ -154,7 +154,7 @@ namespace Ryujinx.Memory
             }
             else
             {
-                _placeholders.UnmapView(IntPtr.Zero, address, size);
+                _placeholders.UnreserveRange((ulong)address, (ulong)size);
             }
 
             return WindowsApi.VirtualFree(address, IntPtr.Zero, AllocationType.Release);

Ryujinx.Memory/WindowsShared/PlaceholderManager.cs

@@ -44,6 +44,50 @@ namespace Ryujinx.Memory.WindowsShared
             }
         }
 
+        /// <summary>
+        /// Unreserves a range of memory that has been previously reserved with <see cref="ReserveRange"/>.
+        /// </summary>
+        /// <param name="address">Start address of the region to unreserve</param>
+        /// <param name="size">Size in bytes of the region to unreserve</param>
+        /// <exception cref="WindowsApiException">Thrown when the Windows API returns an error unreserving the memory</exception>
+        public void UnreserveRange(ulong address, ulong size)
+        {
+            ulong endAddress = address + size;
+
+            var overlaps = Array.Empty<IntervalTreeNode<ulong, ulong>>();
+            int count;
+
+            lock (_mappings)
+            {
+                count = _mappings.Get(address, endAddress, ref overlaps);
+
+                for (int index = 0; index < count; index++)
+                {
+                    var overlap = overlaps[index];
+
+                    if (IsMapped(overlap.Value))
+                    {
+                        if (!WindowsApi.UnmapViewOfFile2(WindowsApi.CurrentProcessHandle, (IntPtr)overlap.Start, 2))
+                        {
+                            throw new WindowsApiException("UnmapViewOfFile2");
+                        }
+                    }
+
+                    _mappings.Remove(overlap);
+                }
+            }
+
+            if (count > 1)
+            {
+                CheckFreeResult(WindowsApi.VirtualFree(
+                    (IntPtr)address,
+                    (IntPtr)size,
+                    AllocationType.Release | AllocationType.CoalescePlaceholders));
+            }
+
+            RemoveProtection(address, size);
+        }
+
         /// <summary>
         /// Maps a shared memory view on a previously reserved memory region.
         /// </summary>
@@ -51,13 +95,14 @@ namespace Ryujinx.Memory.WindowsShared
         /// <param name="srcOffset">Offset in the shared memory to map</param>
         /// <param name="location">Address to map the view into</param>
         /// <param name="size">Size of the view in bytes</param>
-        public void MapView(IntPtr sharedMemory, ulong srcOffset, IntPtr location, IntPtr size)
+        /// <param name="owner">Memory block that owns the mapping</param>
+        public void MapView(IntPtr sharedMemory, ulong srcOffset, IntPtr location, IntPtr size, MemoryBlock owner)
         {
             _partialUnmapLock.AcquireReaderLock(Timeout.Infinite);
 
             try
             {
-                UnmapViewInternal(sharedMemory, location, size);
+                UnmapViewInternal(sharedMemory, location, size, owner);
                 MapViewInternal(sharedMemory, srcOffset, location, size);
             }
             finally
@@ -173,13 +218,14 @@ namespace Ryujinx.Memory.WindowsShared
         /// <param name="sharedMemory">Shared memory that the view being unmapped belongs to</param>
         /// <param name="location">Address to unmap</param>
         /// <param name="size">Size of the region to unmap in bytes</param>
-        public void UnmapView(IntPtr sharedMemory, IntPtr location, IntPtr size)
+        /// <param name="owner">Memory block that owns the mapping</param>
+        public void UnmapView(IntPtr sharedMemory, IntPtr location, IntPtr size, MemoryBlock owner)
         {
             _partialUnmapLock.AcquireReaderLock(Timeout.Infinite);
 
             try
             {
-                UnmapViewInternal(sharedMemory, location, size);
+                UnmapViewInternal(sharedMemory, location, size, owner);
             }
             finally
             {
@@ -197,8 +243,9 @@ namespace Ryujinx.Memory.WindowsShared
         /// <param name="sharedMemory">Shared memory that the view being unmapped belongs to</param>
         /// <param name="location">Address to unmap</param>
         /// <param name="size">Size of the region to unmap in bytes</param>
+        /// <param name="owner">Memory block that owns the mapping</param>
         /// <exception cref="WindowsApiException">Thrown when the Windows API returns an error unmapping or remapping the memory</exception>
-        private void UnmapViewInternal(IntPtr sharedMemory, IntPtr location, IntPtr size)
+        private void UnmapViewInternal(IntPtr sharedMemory, IntPtr location, IntPtr size, MemoryBlock owner)
         {
             ulong startAddress = (ulong)location;
             ulong unmapSize = (ulong)size;
@@ -272,7 +319,7 @@ namespace Ryujinx.Memory.WindowsShared
                 }
             }
 
-            CoalesceForUnmap(startAddress, unmapSize);
+            CoalesceForUnmap(startAddress, unmapSize, owner);
             RemoveProtection(startAddress, unmapSize);
         }
 
@@ -281,15 +328,21 @@ namespace Ryujinx.Memory.WindowsShared
         /// </summary>
         /// <param name="address">Address of the region that was unmapped</param>
         /// <param name="size">Size of the region that was unmapped in bytes</param>
-        private void CoalesceForUnmap(ulong address, ulong size)
+        /// <param name="owner">Memory block that owns the mapping</param>
+        private void CoalesceForUnmap(ulong address, ulong size, MemoryBlock owner)
         {
             ulong endAddress = address + size;
+            ulong blockAddress = (ulong)owner.Pointer;
+            ulong blockEnd = blockAddress + owner.Size;
             var overlaps = Array.Empty<IntervalTreeNode<ulong, ulong>>();
             int unmappedCount = 0;
 
             lock (_mappings)
             {
-                int count = _mappings.Get(address - MinimumPageSize, endAddress + MinimumPageSize, ref overlaps);
+                int count = _mappings.Get(
+                    Math.Max(address - MinimumPageSize, blockAddress),
+                    Math.Min(endAddress + MinimumPageSize, blockEnd), ref overlaps);
+
                 if (count < 2)
                 {
                     // Nothing to coalesce if we only have 1 or no overlaps.