Start building more accurate vertex as compute usage info

- Fixes

Ryujinx.slnx

@@ -21,6 +21,7 @@
   <Project Path="src/Ryujinx.Graphics.Nvdec.Vp9/Ryujinx.Graphics.Nvdec.Vp9.csproj" />
   <Project Path="src/Ryujinx.Graphics.Nvdec/Ryujinx.Graphics.Nvdec.csproj" />
   <Project Path="src/Ryujinx.Graphics.OpenGL/Ryujinx.Graphics.OpenGL.csproj" />
+  <Project Path="src/Ryujinx.Graphics.RenderDocApi/Ryujinx.Graphics.RenderDocApi.csproj" />
   <Project Path="src/Ryujinx.Graphics.Shader/Ryujinx.Graphics.Shader.csproj" />
   <Project Path="src/Ryujinx.Graphics.Texture/Ryujinx.Graphics.Texture.csproj" />
   <Project Path="src/Ryujinx.Graphics.Vic/Ryujinx.Graphics.Vic.csproj" />

src/ARMeilleure/Translation/Translator.cs

@@ -413,6 +413,10 @@ namespace ARMeilleure.Translation
 
                         context.CurrOp = opCode;
 
+                        Operand nativeContext = context.LoadArgument(OperandType.I64, 0);
+                        Operand dispAddressAddr = context.Add(nativeContext, Const((ulong)NativeContext.GetDispatchAddressOffset()));
+                        context.Store(dispAddressAddr, Const(opCode.Address));
+
                         bool isLastOp = opcIndex == block.OpCodes.Count - 1;
 
                         if (isLastOp)

src/Ryujinx.Cpu/LightningJit/Arm32/Decoder.cs

@@ -8,7 +8,7 @@ namespace Ryujinx.Cpu.LightningJit.Arm32
 {
     static class Decoder<T> where T : IInstEmit
     {
-        public static MultiBlock DecodeMulti(CpuPreset cpuPreset, IMemoryManager memoryManager, ulong address, bool isThumb)
+        public static MultiBlock DecodeMulti(CpuPreset cpuPreset, IMemoryManager memoryManager, ulong address, bool singleBlock, bool isThumb)
         {
             List<Block> blocks = [];
             List<ulong> branchTargets = [];
@@ -24,7 +24,7 @@ namespace Ryujinx.Cpu.LightningJit.Arm32
 
                 blocks.Add(block);
 
-                if (block.IsTruncated || !HasNextBlock(block, block.EndAddress - 4UL, branchTargets))
+                if (block.IsTruncated || (singleBlock && block.EndsWithBranch) || !HasNextBlock(block, block.EndAddress - 4UL, branchTargets))
                 {
                     break;
                 }

src/Ryujinx.Cpu/LightningJit/Arm32/Target/Arm64/Compiler.cs

@@ -227,7 +227,7 @@ namespace Ryujinx.Cpu.LightningJit.Arm32.Target.Arm64
 
         public static CompiledFunction Compile(CpuPreset cpuPreset, IMemoryManager memoryManager, ulong address, IAddressTable<ulong> funcTable, nint dispatchStubPtr, bool isThumb)
         {
-            MultiBlock multiBlock = Decoder<InstEmit>.DecodeMulti(cpuPreset, memoryManager, address, isThumb);
+            MultiBlock multiBlock = Decoder<InstEmit>.DecodeMulti(cpuPreset, memoryManager, address, singleBlock: true, isThumb);
 
             Dictionary<ulong, int> targets = new();
 

src/Ryujinx.Cpu/LightningJit/Arm64/Target/Arm64/Compiler.cs

@@ -305,7 +305,7 @@ namespace Ryujinx.Cpu.LightningJit.Arm64.Target.Arm64
 
         public static CompiledFunction Compile(CpuPreset cpuPreset, IMemoryManager memoryManager, ulong address, IAddressTable<ulong> funcTable, nint dispatchStubPtr)
         {
-            MultiBlock multiBlock = Decoder.DecodeMulti(cpuPreset, memoryManager, address);
+            MultiBlock multiBlock = Decoder.DecodeMulti(cpuPreset, memoryManager, address, singleBlock: true);
 
             Dictionary<ulong, int> targets = new();
             List<PendingBranch> pendingBranches = [];

src/Ryujinx.Cpu/LightningJit/Arm64/Target/Arm64/Decoder.cs

@@ -13,7 +13,7 @@ namespace Ryujinx.Cpu.LightningJit.Arm64.Target.Arm64
         private const uint NzcvFlags = 0xfu << 28;
         private const uint CFlag = 0x1u << 29;
 
-        public static MultiBlock DecodeMulti(CpuPreset cpuPreset, IMemoryManager memoryManager, ulong address)
+        public static MultiBlock DecodeMulti(CpuPreset cpuPreset, IMemoryManager memoryManager, ulong address, bool singleBlock)
         {
             List<Block> blocks = [];
             List<ulong> branchTargets = [];
@@ -35,7 +35,7 @@ namespace Ryujinx.Cpu.LightningJit.Arm64.Target.Arm64
 
                 blocks.Add(block);
 
-                if (block.IsTruncated || !HasNextBlock(block, block.EndAddress - 4UL, branchTargets))
+                if (block.IsTruncated || (singleBlock && block.EndsWithBranch) || !HasNextBlock(block, block.EndAddress - 4UL, branchTargets))
                 {
                     break;
                 }

src/Ryujinx.Graphics.GAL/Capabilities.cs

@@ -32,6 +32,7 @@ namespace Ryujinx.Graphics.GAL
         public readonly bool SupportsGeometryShader;
         public readonly bool SupportsGeometryShaderPassthrough;
         public readonly bool SupportsTransformFeedback;
+        public readonly bool SupportsImageBufferPixelAlignment;
         public readonly bool SupportsImageLoadFormatted;
         public readonly bool SupportsLayerVertexTessellation;
         public readonly bool SupportsMismatchingViewFormat;
@@ -43,6 +44,7 @@ namespace Ryujinx.Graphics.GAL
         public readonly bool SupportsShaderBarrierDivergence;
         public readonly bool SupportsShaderFloat64;
         public readonly bool SupportsShaderNonUniformIndexing;
+        public readonly bool SupportsTextureBufferPixelAlignment;
         public readonly bool SupportsTextureGatherOffsets;
         public readonly bool SupportsTextureShadowLod;
         public readonly bool SupportsVertexStoreAndAtomics;
@@ -101,6 +103,7 @@ namespace Ryujinx.Graphics.GAL
             bool supportsGeometryShader,
             bool supportsGeometryShaderPassthrough,
             bool supportsTransformFeedback,
+            bool supportsImageBufferPixelAlignment,
             bool supportsImageLoadFormatted,
             bool supportsLayerVertexTessellation,
             bool supportsMismatchingViewFormat,
@@ -112,6 +115,7 @@ namespace Ryujinx.Graphics.GAL
             bool supportsShaderBarrierDivergence,
             bool supportsShaderFloat64,
             bool supportsShaderNonUniformIndexing,
+            bool supportsTextureBufferPixelAlignment,
             bool supportsTextureGatherOffsets,
             bool supportsTextureShadowLod,
             bool supportsVertexStoreAndAtomics,
@@ -164,6 +168,7 @@ namespace Ryujinx.Graphics.GAL
             SupportsGeometryShader = supportsGeometryShader;
             SupportsGeometryShaderPassthrough = supportsGeometryShaderPassthrough;
             SupportsTransformFeedback = supportsTransformFeedback;
+            SupportsImageBufferPixelAlignment = supportsImageBufferPixelAlignment;
             SupportsImageLoadFormatted = supportsImageLoadFormatted;
             SupportsLayerVertexTessellation = supportsLayerVertexTessellation;
             SupportsMismatchingViewFormat = supportsMismatchingViewFormat;
@@ -175,6 +180,7 @@ namespace Ryujinx.Graphics.GAL
             SupportsShaderBarrierDivergence = supportsShaderBarrierDivergence;
             SupportsShaderFloat64 = supportsShaderFloat64;
             SupportsShaderNonUniformIndexing = supportsShaderNonUniformIndexing;
+            SupportsTextureBufferPixelAlignment = supportsTextureBufferPixelAlignment;
             SupportsTextureGatherOffsets = supportsTextureGatherOffsets;
             SupportsTextureShadowLod = supportsTextureShadowLod;
             SupportsVertexStoreAndAtomics = supportsVertexStoreAndAtomics;

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

@@ -466,6 +466,7 @@ namespace Ryujinx.Graphics.Gpu.Engine.Threed
 
             engine.UpdateState(ulong.MaxValue & ~(1UL << StateUpdater.ShaderStateIndex));
 
+            _channel.TextureManager.SignalRenderTargetsModifiable();
             _channel.TextureManager.UpdateRenderTargets();
 
             int textureId = _state.State.DrawTextureTextureId;
@@ -803,7 +804,7 @@ namespace Ryujinx.Graphics.Gpu.Engine.Threed
             int index = (argument >> 6) & 0xf;
             int layer = (argument >> 10) & 0x3ff;
 
-            RenderTargetUpdateFlags updateFlags = RenderTargetUpdateFlags.SingleColor;
+            RenderTargetUpdateFlags updateFlags = RenderTargetUpdateFlags.SingleColor | RenderTargetUpdateFlags.ForClear;
 
             if (layer != 0 || layerCount > 1)
             {

src/Ryujinx.Graphics.Gpu/Engine/Threed/RenderTargetUpdateFlags.cs

@@ -38,6 +38,11 @@ namespace Ryujinx.Graphics.Gpu.Engine.Threed
         /// </summary>
         DiscardClip = 1 << 4,
 
+        /// <summary>
+        /// Indicates that the render target will be used for a clear operation.
+        /// </summary>
+        ForClear = 1 << 5,
+
         /// <summary>
         /// Default update flags for draw.
         /// </summary>

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

@@ -45,6 +45,7 @@ namespace Ryujinx.Graphics.Gpu.Engine.Threed
         private ProgramPipelineState _pipeline;
 
         private bool _fsReadsFragCoord;
+        private bool _fsAlwaysDiscards;
         private bool _vsUsesDrawParameters;
         private bool _vtgWritesRtLayer;
         private byte _vsClipDistancesWritten;
@@ -491,6 +492,8 @@ namespace Ryujinx.Graphics.Gpu.Engine.Threed
 
             Span<RtColorState> rtColorStateSpan = _state.State.RtColorState.AsSpan();
 
+            bool rtModifiable = updateFlags.HasFlag(RenderTargetUpdateFlags.ForClear) || !_fsAlwaysDiscards;
+
             for (int index = 0; index < Constants.TotalRenderTargets; index++)
             {
                 int rtIndex = useControl ? rtControl.UnpackPermutationIndex(index) : index;
@@ -499,7 +502,7 @@ namespace Ryujinx.Graphics.Gpu.Engine.Threed
 
                 if (index >= count || !IsRtEnabled(colorState) || (singleColor && index != singleUse))
                 {
-                    changedScale |= _channel.TextureManager.SetRenderTargetColor(index, null);
+                    changedScale |= _channel.TextureManager.SetRenderTargetColor(index, null, rtModifiable);
 
                     continue;
                 }
@@ -518,7 +521,7 @@ namespace Ryujinx.Graphics.Gpu.Engine.Threed
                     samplesInY,
                     sizeHint);
 
-                changedScale |= _channel.TextureManager.SetRenderTargetColor(index, color);
+                changedScale |= _channel.TextureManager.SetRenderTargetColor(index, color, rtModifiable);
 
                 if (color != null)
                 {
@@ -572,7 +575,7 @@ namespace Ryujinx.Graphics.Gpu.Engine.Threed
                 }
             }
 
-            changedScale |= _channel.TextureManager.SetRenderTargetDepthStencil(depthStencil);
+            changedScale |= _channel.TextureManager.SetRenderTargetDepthStencil(depthStencil, rtModifiable);
 
             if (changedScale)
             {
@@ -774,6 +777,11 @@ namespace Ryujinx.Graphics.Gpu.Engine.Threed
                 float width = scaleX * 2;
                 float height = scaleY * 2;
 
+                if (yNegate)
+                {
+                    y += _state.State.ScreenScissorState.Height - MathF.Abs(height);
+                }
+
                 float scale = _channel.TextureManager.RenderTargetScale;
                 if (scale != 1f)
                 {
@@ -1517,20 +1525,38 @@ namespace Ryujinx.Graphics.Gpu.Engine.Threed
                 _currentProgramInfo[stageIndex] = info;
             }
 
-            if (gs.Shaders[5]?.Info.UsesFragCoord == true)
-            {
-                // Make sure we update the viewport size on the support buffer if it will be consumed on the new shader.
+            _fsReadsFragCoord = false;
 
-                if (!_fsReadsFragCoord && (_state.State.YControl & YControl.NegateY) != 0)
+            ShaderProgramInfo fragmentShaderInfo = gs.Shaders[5]?.Info;
+
+            if (fragmentShaderInfo != null)
+            {
+                if (fragmentShaderInfo.UsesFragCoord)
                 {
-                    UpdateSupportBufferViewportSize();
+                    // Make sure we update the viewport size on the support buffer if it will be consumed on the new shader.
+
+                    if (!_fsReadsFragCoord && (_state.State.YControl & YControl.NegateY) != 0)
+                    {
+                        UpdateSupportBufferViewportSize();
+                    }
+
+                    _fsReadsFragCoord = true;
                 }
 
-                _fsReadsFragCoord = true;
+                if (_fsAlwaysDiscards != fragmentShaderInfo.HasUnconditionalDiscard)
+                {
+                    _fsAlwaysDiscards = fragmentShaderInfo.HasUnconditionalDiscard;
+
+                    if (!_fsAlwaysDiscards)
+                    {
+                        _channel.TextureManager.RefreshModifiedTextures();
+                        _channel.TextureManager.SignalRenderTargetsModifiable();
+                    }
+                }
             }
             else
             {
-                _fsReadsFragCoord = false;
+                _fsAlwaysDiscards = false;
             }
 
             if (gs.VertexAsCompute != null)

src/Ryujinx.Graphics.Gpu/Image/Texture.cs

@@ -1572,11 +1572,6 @@ namespace Ryujinx.Graphics.Gpu.Image
         /// <param name="bound">True if the texture has been bound, false if it has been unbound</param>
         public void SignalModifying(bool bound)
         {
-            if (bound)
-            {
-                _scaledSetScore = Math.Max(0, _scaledSetScore - 1);
-            }
-
             if (_modifiedStale || Group.HasCopyDependencies || Group.HasFlushBuffer)
             {
                 _modifiedStale = false;
@@ -1584,9 +1579,18 @@ namespace Ryujinx.Graphics.Gpu.Image
             }
 
             _physicalMemory.TextureCache.Lift(this);
+        }
 
+        /// <summary>
+        /// Signals that a render target texture has been either bound or unbound.
+        /// </summary>
+        /// <param name="bound">True if the texture has been bound, false if it has been unbound</param>
+        public void SignalBindingChange(bool bound)
+        {
             if (bound)
             {
+                _scaledSetScore = Math.Max(0, _scaledSetScore - 1);
+
                 IncrementReferenceCount();
             }
             else

src/Ryujinx.Graphics.Gpu/Image/TextureBindingsManager.cs

@@ -4,6 +4,7 @@ using Ryujinx.Graphics.Gpu.Engine.Types;
 using Ryujinx.Graphics.Gpu.Memory;
 using Ryujinx.Graphics.Gpu.Shader;
 using Ryujinx.Graphics.Shader;
+using Ryujinx.Memory.Range;
 using System;
 using System.Runtime.CompilerServices;
 using System.Runtime.InteropServices;
@@ -67,6 +68,9 @@ namespace Ryujinx.Graphics.Gpu.Image
 
         private int _lastFragmentTotal;
 
+        private readonly int _bufferTextureAlignment;
+        private readonly int _bufferImageAlignment;
+
         /// <summary>
         /// Constructs a new instance of the texture bindings manager.
         /// </summary>
@@ -108,6 +112,10 @@ namespace Ryujinx.Graphics.Gpu.Image
             }
 
             _textureCounts = [];
+
+            int alignment = context.Capabilities.TextureBufferOffsetAlignment;
+            _bufferTextureAlignment = context.Capabilities.SupportsTextureBufferPixelAlignment ? 0 : alignment;
+            _bufferImageAlignment = context.Capabilities.SupportsImageBufferPixelAlignment ? 0 : alignment;
         }
 
         /// <summary>
@@ -521,10 +529,12 @@ namespace Ryujinx.Graphics.Gpu.Image
 
                 if (hostTexture != null && texture.Target == Target.TextureBuffer)
                 {
+                    MultiRange range = GetAlignedBufferTextureRange(texture, index, stageIndex, false);
+
                     // Ensure that the buffer texture is using the correct buffer as storage.
                     // Buffers are frequently re-created to accommodate larger data, so we need to re-bind
                     // to ensure we're not using a old buffer that was already deleted.
-                    _channel.BufferManager.SetBufferTextureStorage(stage, hostTexture, texture.Range, bindingInfo, false);
+                    _channel.BufferManager.SetBufferTextureStorage(stage, hostTexture, range, bindingInfo, false);
 
                     // Cache is not used for buffer texture, it must always rebind.
                     state.CachedTexture = null;
@@ -659,7 +669,9 @@ namespace Ryujinx.Graphics.Gpu.Image
                     // Buffers are frequently re-created to accommodate larger data, so we need to re-bind
                     // to ensure we're not using a old buffer that was already deleted.
 
-                    _channel.BufferManager.SetBufferTextureStorage(stage, hostTexture, texture.Range, bindingInfo, true);
+                    MultiRange range = GetAlignedBufferTextureRange(texture, index, stageIndex, true);
+
+                    _channel.BufferManager.SetBufferTextureStorage(stage, hostTexture, range, bindingInfo, true);
 
                     // Cache is not used for buffer texture, it must always rebind.
                     state.CachedTexture = null;
@@ -692,6 +704,61 @@ namespace Ryujinx.Graphics.Gpu.Image
             return specStateMatches;
         }
 
+        /// <summary>
+        /// Gets the aligned address of the texture according to the host requirements.
+        /// </summary>
+        /// <param name="texture">Texture to have its address aligned</param>
+        /// <param name="index">Index of the texture in the shader</param>
+        /// <param name="stageIndex">Index of the shader stage</param>
+        /// <param name="isImage">True if the texture is bound as image, false for sampled textures</param>
+        /// <returns>Aligned address and size of the buffer texture</returns>
+        private MultiRange GetAlignedBufferTextureRange(Texture texture, int index, int stageIndex, bool isImage)
+        {
+            MultiRange range = texture.Range;
+            int alignment = isImage ? _bufferImageAlignment : _bufferTextureAlignment;
+
+            if (alignment != 0)
+            {
+                MemoryRange firstRange = range.GetSubRange(0);
+                ulong misalign = firstRange.Address & ((ulong)alignment - 1);
+                int offset = misalign != 0 ? (int)misalign / texture.Info.FormatInfo.BytesPerPixel : 0;
+
+                if (misalign != 0)
+                {
+                    firstRange = new MemoryRange(firstRange.Address - misalign, firstRange.Size + misalign);
+
+                    if (range.Count > 1)
+                    {
+                        MemoryRange[] ranges = new MemoryRange[range.Count];
+
+                        ranges[0] = firstRange;
+
+                        for (int i = 1; i < range.Count; i++)
+                        {
+                            ranges[i] = range.GetSubRange(i);
+                        }
+
+                        range = new MultiRange(ranges);
+                    }
+                    else
+                    {
+                        range = new MultiRange(firstRange.Address, firstRange.Size);
+                    }
+                }
+
+                if (isImage)
+                {
+                    _context.SupportBufferUpdater.UpdateBufferImageOffset(stageIndex, index, offset);
+                }
+                else
+                {
+                    _context.SupportBufferUpdater.UpdateBufferTextureOffset(stageIndex, index, offset);
+                }
+            }
+
+            return range;
+        }
+
         /// <summary>
         /// Gets the texture descriptor for a given texture handle.
         /// </summary>

src/Ryujinx.Graphics.Gpu/Image/TextureManager.cs

@@ -19,12 +19,39 @@ namespace Ryujinx.Graphics.Gpu.Image
         private readonly TexturePoolCache _texturePoolCache;
         private readonly SamplerPoolCache _samplerPoolCache;
 
+        /// <summary>
+        /// Bound render target texture modification report state.
+        /// </summary>
+        [Flags]
+        private enum BindState : byte
+        {
+            /// <summary>
+            /// Render target texture has not been signalled for modification, and can't be modified on the next render operations.
+            /// </summary>
+            None = 0,
+
+            /// <summary>
+            /// Render target texture has been signalled for modification.
+            /// </summary>
+            Bound = 1 << 0,
+
+            /// <summary>
+            /// Render target texture might be modified on the next render operations.
+            /// </summary>
+            Modified = 1 << 1,
+
+            /// <summary>
+            /// Render target texture has been signalled for modification and might be modified on the next render operations.
+            /// </summary>
+            BoundModified = Bound | Modified,
+        }
+
         private readonly Texture[] _rtColors;
         private readonly ITexture[] _rtHostColors;
-        private readonly bool[] _rtColorsBound;
+        private readonly BindState[] _rtColorsBound;
         private Texture _rtDepthStencil;
         private ITexture _rtHostDs;
-        private bool _rtDsBound;
+        private BindState _rtDsBound;
 
         public int ClipRegionWidth { get; private set; }
         public int ClipRegionHeight { get; private set; }
@@ -55,7 +82,7 @@ namespace Ryujinx.Graphics.Gpu.Image
 
             _rtColors = new Texture[Constants.TotalRenderTargets];
             _rtHostColors = new ITexture[Constants.TotalRenderTargets];
-            _rtColorsBound = new bool[Constants.TotalRenderTargets];
+            _rtColorsBound = new BindState[Constants.TotalRenderTargets];
         }
 
         /// <summary>
@@ -151,27 +178,39 @@ namespace Ryujinx.Graphics.Gpu.Image
         /// </summary>
         /// <param name="index">The index of the color buffer to set (up to 8)</param>
         /// <param name="color">The color buffer texture</param>
+        /// <param name="modified">Indicates if the following render operations will modidify <paramref name="color"/> contents</param>
         /// <returns>True if render target scale must be updated.</returns>
-        public bool SetRenderTargetColor(int index, Texture color)
+        public bool SetRenderTargetColor(int index, Texture color, bool modified)
         {
             bool hasValue = color != null;
             bool changesScale = (hasValue != (_rtColors[index] != null)) || (hasValue && RenderTargetScale != color.ScaleFactor);
 
             if (_rtColors[index] != color)
             {
-                if (_rtColorsBound[index])
+                Texture oldColor = _rtColors[index];
+
+                if (oldColor != null)
                 {
-                    _rtColors[index]?.SignalModifying(false);
-                }
-                else
-                {
-                    _rtColorsBound[index] = true;
+                    if (_rtColorsBound[index].HasFlag(BindState.Bound))
+                    {
+                        oldColor.SignalModifying(false);
+                    }
+
+                    oldColor.SignalBindingChange(false);
                 }
 
+                _rtColorsBound[index] = modified ? BindState.BoundModified : BindState.None;
+
                 if (color != null)
                 {
                     color.SynchronizeMemory();
-                    color.SignalModifying(true);
+
+                    if (modified)
+                    {
+                        color.SignalModifying(true);
+                    }
+
+                    color.SignalBindingChange(true);
                 }
 
                 _rtColors[index] = color;
@@ -184,27 +223,39 @@ namespace Ryujinx.Graphics.Gpu.Image
         /// Sets the render target depth-stencil buffer.
         /// </summary>
         /// <param name="depthStencil">The depth-stencil buffer texture</param>
+        /// <param name="modified">Indicates if the following render operations will modidify <paramref name="depthStencil"/> contents</param>
         /// <returns>True if render target scale must be updated.</returns>
-        public bool SetRenderTargetDepthStencil(Texture depthStencil)
+        public bool SetRenderTargetDepthStencil(Texture depthStencil, bool modified)
         {
             bool hasValue = depthStencil != null;
             bool changesScale = (hasValue != (_rtDepthStencil != null)) || (hasValue && RenderTargetScale != depthStencil.ScaleFactor);
 
             if (_rtDepthStencil != depthStencil)
             {
-                if (_rtDsBound)
+                Texture oldDepthStencil = _rtDepthStencil;
+
+                if (oldDepthStencil != null)
                 {
-                    _rtDepthStencil?.SignalModifying(false);
-                }
-                else
-                {
-                    _rtDsBound = true;
+                    if (_rtDsBound.HasFlag(BindState.Bound))
+                    {
+                        oldDepthStencil.SignalModifying(false);
+                    }
+
+                    oldDepthStencil.SignalBindingChange(false);
                 }
 
+                _rtDsBound = modified ? BindState.BoundModified : BindState.None;
+
                 if (depthStencil != null)
                 {
                     depthStencil.SynchronizeMemory();
-                    depthStencil.SignalModifying(true);
+
+                    if (modified)
+                    {
+                        depthStencil.SignalModifying(true);
+                    }
+
+                    depthStencil.SignalBindingChange(true);
                 }
 
                 _rtDepthStencil = depthStencil;
@@ -443,10 +494,10 @@ namespace Ryujinx.Graphics.Gpu.Image
             {
                 hostDsTexture = dsTexture.HostTexture;
 
-                if (!_rtDsBound)
+                if (_rtDsBound == BindState.Modified)
                 {
                     dsTexture.SignalModifying(true);
-                    _rtDsBound = true;
+                    _rtDsBound |= BindState.Bound;
                 }
             }
 
@@ -470,10 +521,10 @@ namespace Ryujinx.Graphics.Gpu.Image
                 {
                     hostTexture = texture.HostTexture;
 
-                    if (!_rtColorsBound[index])
+                    if (_rtColorsBound[index] == BindState.Modified)
                     {
                         texture.SignalModifying(true);
-                        _rtColorsBound[index] = true;
+                        _rtColorsBound[index] |= BindState.Bound;
                     }
                 }
 
@@ -518,24 +569,37 @@ namespace Ryujinx.Graphics.Gpu.Image
         {
             Texture dsTexture = _rtDepthStencil;
 
-            if (dsTexture != null && _rtDsBound)
+            if (dsTexture != null && _rtDsBound.HasFlag(BindState.Bound))
             {
                 dsTexture.SignalModifying(false);
-                _rtDsBound = false;
+                _rtDsBound &= ~BindState.Bound;
             }
 
             for (int index = 0; index < _rtColors.Length; index++)
             {
                 Texture texture = _rtColors[index];
 
-                if (texture != null && _rtColorsBound[index])
+                if (texture != null && _rtColorsBound[index].HasFlag(BindState.Bound))
                 {
                     texture.SignalModifying(false);
-                    _rtColorsBound[index] = false;
+                    _rtColorsBound[index] &= ~BindState.Bound;
                 }
             }
         }
 
+        /// <summary>
+        /// Indicates that the currently bound render targets might be modified, if they are used on the next render operation.
+        /// </summary>
+        public void SignalRenderTargetsModifiable()
+        {
+            _rtDsBound |= BindState.Modified;
+
+            for (int index = 0; index < _rtColorsBound.Length; index++)
+            {
+                _rtColorsBound[index] |= BindState.Modified;
+            }
+        }
+
         /// <summary>
         /// Forces the texture and sampler pools to be re-loaded from the cache on next use.
         /// </summary>
@@ -572,19 +636,11 @@ namespace Ryujinx.Graphics.Gpu.Image
 
             for (int i = 0; i < _rtColors.Length; i++)
             {
-                if (_rtColorsBound[i])
-                {
-                    _rtColors[i]?.DecrementReferenceCount();
-                }
-
+                _rtColors[i]?.DecrementReferenceCount();
                 _rtColors[i] = null;
             }
 
-            if (_rtDsBound)
-            {
-                _rtDepthStencil?.DecrementReferenceCount();
-            }
-
+            _rtDepthStencil?.DecrementReferenceCount();
             _rtDepthStencil = null;
         }
     }

src/Ryujinx.Graphics.Gpu/Memory/SupportBufferUpdater.cs

@@ -129,6 +129,38 @@ namespace Ryujinx.Graphics.Gpu.Memory
             }
         }
 
+        /// <summary>
+        /// Updates the offset used for accessing buffer textures that are not aligned to the host requirements.
+        /// </summary>
+        /// <param name="stageIndex">Index of the shader stage where the texture is used</param>
+        /// <param name="bindingIndex">Index of the texture binding in the shader</param>
+        /// <param name="offset">Offset for the misaligned part of the address</param>
+        public void UpdateBufferTextureOffset(int stageIndex, int bindingIndex, int offset)
+        {
+            if (_data.BufferTextureOffset[stageIndex][bindingIndex].X != offset)
+            {
+                _data.BufferTextureOffset[stageIndex][bindingIndex].X = offset;
+                int index = stageIndex * SupportBuffer.TextureCount + bindingIndex;
+                MarkDirty(SupportBuffer.BufferTextureOffsetOffset + index * sizeof(int) * 4, sizeof(int));
+            }
+        }
+
+        /// <summary>
+        /// Updates the offset used for accessing buffer images that are not aligned to the host requirements.
+        /// </summary>
+        /// <param name="stageIndex">Index of the shader stage where the image is used</param>
+        /// <param name="bindingIndex">Index of the image binding in the shader</param>
+        /// <param name="offset">Offset for the misaligned part of the address</param>
+        public void UpdateBufferImageOffset(int stageIndex, int bindingIndex, int offset)
+        {
+            if (_data.BufferTextureOffset[stageIndex][bindingIndex].Y != offset)
+            {
+                _data.BufferTextureOffset[stageIndex][bindingIndex].Y = offset;
+                int index = stageIndex * SupportBuffer.TextureCount + bindingIndex;
+                MarkDirty(SupportBuffer.BufferTextureOffsetOffset + index * sizeof(int) * 4 + sizeof(int), sizeof(int));
+            }
+        }
+
         /// <summary>
         /// Sets whether the format of a given render target is a BGRA format.
         /// </summary>

src/Ryujinx.Graphics.Gpu/Shader/DiskCache/DiskCacheHostStorage.cs

@@ -22,7 +22,7 @@ namespace Ryujinx.Graphics.Gpu.Shader.DiskCache
         private const ushort FileFormatVersionMajor = 1;
         private const ushort FileFormatVersionMinor = 2;
         private const uint FileFormatVersionPacked = ((uint)FileFormatVersionMajor << 16) | FileFormatVersionMinor;
-        private const uint CodeGenVersion = 7354;
+        private const uint CodeGenVersion = 6371;
 
         private const string SharedTocFileName = "shared.toc";
         private const string SharedDataFileName = "shared.data";
@@ -170,6 +170,11 @@ namespace Ryujinx.Graphics.Gpu.Shader.DiskCache
             /// </summary>
             public bool UsesRtLayer;
 
+            /// <summary>
+            /// Indicates that the fragment shader always discards the fragment, not producing any output for the bound render targets.
+            /// </summary>
+            public bool HasUnconditionalDiscard;
+
             /// <summary>
             /// Bit mask with the clip distances written on the vertex stage.
             /// </summary>
@@ -806,6 +811,7 @@ namespace Ryujinx.Graphics.Gpu.Shader.DiskCache
                 dataInfo.UsesInstanceId,
                 dataInfo.UsesDrawParameters,
                 dataInfo.UsesRtLayer,
+                dataInfo.HasUnconditionalDiscard,
                 dataInfo.ClipDistancesWritten,
                 dataInfo.FragmentOutputMap);
         }
@@ -836,6 +842,7 @@ namespace Ryujinx.Graphics.Gpu.Shader.DiskCache
                 UsesInstanceId = info.UsesInstanceId,
                 UsesDrawParameters = info.UsesDrawParameters,
                 UsesRtLayer = info.UsesRtLayer,
+                HasUnconditionalDiscard = info.HasUnconditionalDiscard,
                 ClipDistancesWritten = info.ClipDistancesWritten,
                 FragmentOutputMap = info.FragmentOutputMap,
             };

src/Ryujinx.Graphics.Gpu/Shader/GpuAccessorBase.cs

@@ -207,6 +207,10 @@ namespace Ryujinx.Graphics.Gpu.Shader
 
         public bool QueryHostSupportsBgraFormat() => _context.Capabilities.SupportsBgraFormat;
 
+        public bool QueryHostSupportsBufferImagePixelAlignment() => _context.Capabilities.SupportsImageBufferPixelAlignment;
+
+        public bool QueryHostSupportsBufferTexturePixelAlignment() => _context.Capabilities.SupportsTextureBufferPixelAlignment;
+
         public bool QueryHostSupportsFragmentShaderInterlock() => _context.Capabilities.SupportsFragmentShaderInterlock;
 
         public bool QueryHostSupportsFragmentShaderOrderingIntel() => _context.Capabilities.SupportsFragmentShaderOrderingIntel;

src/Ryujinx.Graphics.Gpu/Shader/ShaderCache.cs

@@ -430,7 +430,8 @@ namespace Ryujinx.Graphics.Gpu.Shader
 
                             TranslatorContext lastInVertexPipeline = geometryToCompute ? translatorContexts[4] ?? currentStage : currentStage;
 
-                            program = lastInVertexPipeline.GenerateVertexPassthroughForCompute();
+                            (program, ShaderProgramInfo vacInfo) = lastInVertexPipeline.GenerateVertexPassthroughForCompute();
+                            infoBuilder.AddStageInfoVac(vacInfo);
                         }
                         else
                         {
@@ -535,7 +536,7 @@ namespace Ryujinx.Graphics.Gpu.Shader
         private ShaderAsCompute CreateHostVertexAsComputeProgram(ShaderProgram program, TranslatorContext context, bool tfEnabled)
         {
             ShaderSource source = new(program.Code, program.BinaryCode, ShaderStage.Compute, program.Language);
-            ShaderInfo info = ShaderInfoBuilder.BuildForVertexAsCompute(_context, program.Info, tfEnabled);
+            ShaderInfo info = ShaderInfoBuilder.BuildForVertexAsCompute(_context, program.Info, context.GetVertexAsComputeInfo(), tfEnabled);
 
             return new(_context.Renderer.CreateProgram([source], info), program.Info, context.GetResourceReservations());
         }

src/Ryujinx.Graphics.Gpu/Shader/ShaderInfoBuilder.cs

@@ -95,7 +95,7 @@ namespace Ryujinx.Graphics.Gpu.Shader
         private void PopulateDescriptorAndUsages(ResourceStages stages, ResourceType type, int setIndex, int start, int count, bool write = false)
         {
             AddDescriptor(stages, type, setIndex, start, count);
-            AddUsage(stages, type, setIndex, start, count, write);
+            // AddUsage(stages, type, setIndex, start, count, write);
         }
 
         /// <summary>
@@ -159,6 +159,25 @@ namespace Ryujinx.Graphics.Gpu.Shader
             AddUsage(info.Images, stages, isImage: true);
         }
 
+        public void AddStageInfoVac(ShaderProgramInfo info)
+        {
+            ResourceStages stages = info.Stage switch
+            {
+                ShaderStage.Compute => ResourceStages.Compute,
+                ShaderStage.Vertex => ResourceStages.Vertex,
+                ShaderStage.TessellationControl => ResourceStages.TessellationControl,
+                ShaderStage.TessellationEvaluation => ResourceStages.TessellationEvaluation,
+                ShaderStage.Geometry => ResourceStages.Geometry,
+                ShaderStage.Fragment => ResourceStages.Fragment,
+                _ => ResourceStages.None,
+            };
+
+            AddUsage(info.CBuffers, stages, isStorage: false);
+            AddUsage(info.SBuffers, stages, isStorage: true);
+            AddUsage(info.Textures, stages, isImage: false);
+            AddUsage(info.Images, stages, isImage: true);
+        }
+
         /// <summary>
         /// Adds a resource descriptor to the list of descriptors.
         /// </summary>
@@ -422,10 +441,11 @@ namespace Ryujinx.Graphics.Gpu.Shader
         /// <param name="tfEnabled">Indicates if the graphics shader is used with transform feedback enabled</param>
         /// <param name="fromCache">True if the compute shader comes from a disk cache, false otherwise</param>
         /// <returns>Shader information</returns>
-        public static ShaderInfo BuildForVertexAsCompute(GpuContext context, ShaderProgramInfo info, bool tfEnabled, bool fromCache = false)
+        public static ShaderInfo BuildForVertexAsCompute(GpuContext context, ShaderProgramInfo info, ShaderProgramInfo info2, bool tfEnabled, bool fromCache = false)
         {
             ShaderInfoBuilder builder = new(context, tfEnabled, vertexAsCompute: true);
 
+            builder.AddStageInfoVac(info2);
             builder.AddStageInfo(info, vertexAsCompute: true);
 
             return builder.Build(null, fromCache);

src/Ryujinx.Graphics.OpenGL/OpenGLRenderer.cs

@@ -174,6 +174,7 @@ namespace Ryujinx.Graphics.OpenGL
                 supportsGeometryShader: true,
                 supportsGeometryShaderPassthrough: HwCapabilities.SupportsGeometryShaderPassthrough,
                 supportsTransformFeedback: true,
+                supportsImageBufferPixelAlignment: false,
                 supportsImageLoadFormatted: HwCapabilities.SupportsImageLoadFormatted,
                 supportsLayerVertexTessellation: HwCapabilities.SupportsShaderViewportLayerArray,
                 supportsMismatchingViewFormat: HwCapabilities.SupportsMismatchingViewFormat,
@@ -185,6 +186,7 @@ namespace Ryujinx.Graphics.OpenGL
                 supportsShaderBarrierDivergence: !(intelWindows || intelUnix),
                 supportsShaderFloat64: true,
                 supportsShaderNonUniformIndexing: false,
+                supportsTextureBufferPixelAlignment: false,
                 supportsTextureGatherOffsets: true,
                 supportsTextureShadowLod: HwCapabilities.SupportsTextureShadowLod,
                 supportsVertexStoreAndAtomics: true,

src/Ryujinx.Graphics.RenderDocApi/Capture.cs

@@ -0,0 +1,12 @@
+using System;
+
+namespace Ryujinx.Graphics.RenderDocApi
+{
+    public readonly record struct Capture(int Index, string FileName, DateTime Timestamp)
+    {
+        public void SetComments(string comments)
+        {
+            RenderDoc.SetCaptureFileComments(FileName, comments);
+        }
+    }
+}

src/Ryujinx.Graphics.RenderDocApi/CaptureOption.cs

@@ -0,0 +1,100 @@
+// ReSharper disable UnusedMember.Global
+
+namespace Ryujinx.Graphics.RenderDocApi
+{
+    public enum CaptureOption
+    {
+        /// <summary>
+        /// specifies whether the application is allowed to enable vsync. Default is on.
+        /// </summary>
+        AllowVsync = 0,
+        /// <summary>
+        /// specifies whether the application is allowed to enter exclusive fullscreen. Default is on.
+        /// </summary>
+        AllowFullscreen = 1,
+        /// <summary>
+        /// specifies whether (where possible) API-specific debugging is enabled. Default is off.
+        /// </summary>
+        ApiValidation = 2,
+        /// <summary>
+        /// specifies whether each API call should save a callstack. Default is off.
+        /// </summary>
+        CaptureCallstacks = 3,
+        /// <summary>
+        /// specifies whether, if <see cref="CaptureCallstacks"/> is enabled, callstacks are only saved on actions. Default is off.
+        /// </summary>
+        CaptureCallstacksOnlyDraws = 4,
+        /// <summary>
+        /// specifies a delay in seconds after launching a process to pause, to allow debuggers to attach. <br/>
+        /// This will only apply to child processes since the delay happens at process startup. Default is 0.
+        /// </summary>
+        DelayForDebugger = 5,
+        /// <summary>
+        /// specifies whether any mapped memory updates should be bounds-checked for overruns,
+        /// and uninitialised buffers are initialized to <code>0xDDDDDDDD</code> to catch use of uninitialised data.
+        /// Only supported on D3D11 and OpenGL. Default is off.
+        /// </summary>
+        /// <remarks>
+        /// This option is only valid for OpenGL and D3D11. Explicit APIs such as D3D12 and Vulkan do
+        /// not do the same kind of interception &amp; checking, and undefined contents are really undefined.
+        /// </remarks>
+        VerifyBufferAccess = 6,
+        /// <summary>
+        /// Hooks any system API calls that create child processes, and injects
+        /// RenderDoc into them recursively with the same options.
+        /// </summary>
+        HookIntoChildren = 7,
+        /// <summary>
+        /// specifies whether all live resources at the time of capture should be included in the capture,
+        /// even if they are not referenced by the frame. Default is off.
+        /// </summary>
+        RefAllSources = 8,
+        /// <summary>
+        /// By default, RenderDoc skips saving initial states for resources where the
+        /// previous contents don't appear to be used, assuming that writes before
+        /// reads indicate previous contents aren't used.
+        /// </summary>
+        /// <remarks>
+        /// **NOTE**: As of RenderDoc v1.1 this option has been deprecated. Setting or
+        /// getting it will be ignored, to allow compatibility with older versions.
+        /// In v1.1 the option acts as if it's always enabled.
+        /// </remarks>
+        SaveAllInitials = 9,
+        /// <summary>
+        /// In APIs that allow for the recording of command lists to be replayed later,
+        /// RenderDoc may choose to not capture command lists before a frame capture is
+        /// triggered, to reduce overheads. This means any command lists recorded once
+        /// and replayed many times will not be available and may cause a failure to
+        /// capture.
+        /// </summary>
+        /// <remarks>
+        /// NOTE: This is only true for APIs where multithreading is difficult or
+        /// discouraged. Newer APIs like Vulkan and D3D12 will ignore this option
+        /// and always capture all command lists since the API is heavily oriented
+        /// around it and the overheads have been reduced by API design.
+        /// </remarks>
+        CaptureAllCmdLists = 10,
+        /// <summary>
+        /// Mute API debugging output when the <see cref="ApiValidation"/> option is enabled.
+        /// </summary>
+        DebugOutputMute = 11,
+        /// <summary>
+        /// Allow vendor extensions to be used even when they may be
+        /// incompatible with RenderDoc and cause corrupted replays or crashes.
+        /// </summary>
+        AllowUnsupportedVendorExtensions = 12,
+        /// <summary>
+        /// Define a soft memory limit which some APIs may aim to keep overhead under where
+        /// possible. Anything above this limit will where possible be saved directly to disk during
+        /// capture.<br/>
+        /// This will cause increased disk space use (which may cause a capture to fail if disk space is
+        /// exhausted) as well as slower capture times.
+        /// <br/><br/>
+        /// Not all memory allocations may be deferred like this so it is not a guarantee of a memory
+        /// limit.
+        /// <br/><br/>
+        /// Units are in MBs, suggested values would range from 200MB to 1000MB.
+        /// </summary>
+        SoftMemoryLimit = 13,
+    }
+}

src/Ryujinx.Graphics.RenderDocApi/InputButton.cs

@@ -0,0 +1,83 @@
+
+// ReSharper disable UnusedMember.Global
+namespace Ryujinx.Graphics.RenderDocApi
+{
+    public enum InputButton
+    {
+        // '0' - '9' matches ASCII values
+        Key0 = 0x30,
+        Key1 = 0x31,
+        Key2 = 0x32,
+        Key3 = 0x33,
+        Key4 = 0x34,
+        Key5 = 0x35,
+        Key6 = 0x36,
+        Key7 = 0x37,
+        Key8 = 0x38,
+        Key9 = 0x39,
+
+        // 'A' - 'Z' matches ASCII values
+        A = 0x41,
+        B = 0x42,
+        C = 0x43,
+        D = 0x44,
+        E = 0x45,
+        F = 0x46,
+        G = 0x47,
+        H = 0x48,
+        I = 0x49,
+        J = 0x4A,
+        K = 0x4B,
+        L = 0x4C,
+        M = 0x4D,
+        N = 0x4E,
+        O = 0x4F,
+        P = 0x50,
+        Q = 0x51,
+        R = 0x52,
+        S = 0x53,
+        T = 0x54,
+        U = 0x55,
+        V = 0x56,
+        W = 0x57,
+        X = 0x58,
+        Y = 0x59,
+        Z = 0x5A,
+
+        // leave the rest of the ASCII range free
+        // in case we want to use it later
+        NonPrintable = 0x100,
+
+        Divide,
+        Multiply,
+        Subtract,
+        Plus,
+
+        F1,
+        F2,
+        F3,
+        F4,
+        F5,
+        F6,
+        F7,
+        F8,
+        F9,
+        F10,
+        F11,
+        F12,
+
+        Home,
+        End,
+        Insert,
+        Delete,
+        PageUp,
+        PageDn,
+
+        Backspace,
+        Tab,
+        PrtScrn,
+        Pause,
+
+        Max,
+    }
+}

src/Ryujinx.Graphics.RenderDocApi/OverlayBits.cs

@@ -0,0 +1,39 @@
+// ReSharper disable UnusedMember.Global
+
+using System;
+
+namespace Ryujinx.Graphics.RenderDocApi
+{
+    [Flags]
+    public enum OverlayBits
+    {
+        /// <summary>
+        /// This single bit controls whether the overlay is enabled or disabled globally
+        /// </summary>
+        Enabled = 1 << 0,
+        /// <summary>
+        /// Show the average framerate over several seconds as well as min/max
+        /// </summary>
+        FrameRate = 1 << 1,
+        /// <summary>
+        /// Show the current frame number
+        /// </summary>
+        FrameNumber = 1 << 2,
+        /// <summary>
+        /// Show a list of recent captures, and how many captures have been made
+        /// </summary>
+        CaptureList = 1 << 3,
+        /// <summary>
+        /// Default values for the overlay mask
+        /// </summary>
+        Default = Enabled | FrameRate | FrameNumber | CaptureList,
+        /// <summary>
+        /// Enable all bits
+        /// </summary>
+        All = ~0,
+        /// <summary>
+        /// Disable all bits
+        /// </summary>
+        None = 0
+    }
+}

src/Ryujinx.Graphics.RenderDocApi/README.md

@@ -0,0 +1,5 @@
+# Ryujinx.Graphics.RenderDocApi
+
+This is a C# binding for RenderDoc's application API.
+This is a source-inclusion of https://github.com/utkumaden/RenderdocSharp. 
+I didn't use the NuGet package as I had a few minor changes I wanted to make, and I want to learn from it as well via hands-on experience.

src/Ryujinx.Graphics.RenderDocApi/RenderDoc.cs

@@ -0,0 +1,639 @@
+using System;
+using System.Collections.Immutable;
+using System.Diagnostics;
+using System.Diagnostics.CodeAnalysis;
+using System.Reflection;
+using System.Runtime.CompilerServices;
+using System.Runtime.InteropServices;
+using System.Text;
+using System.Text.RegularExpressions;
+
+namespace Ryujinx.Graphics.RenderDocApi
+{
+    public static unsafe partial class RenderDoc
+    {
+        /// <summary>
+        /// True if the API is available.
+        /// </summary>
+        public static bool IsAvailable => Api != null;
+
+        /// <summary>
+        /// Set the minimum version of the API you require.
+        /// </summary>
+        /// <remarks>Set this before you do anything else with the RenderDoc API, including <see cref="IsAvailable"/>.</remarks>
+        public static RenderDocVersion MinimumRequired { get; set; } = RenderDocVersion.Version_1_0_0;
+
+        /// <summary>
+        /// Set to true to assert versions.
+        /// </summary>
+        public static bool AssertVersionEnabled { get; set; } = true;
+
+        /// <summary>
+        /// Version of the API available.
+        /// </summary>
+        [MemberNotNullWhen(true, nameof(IsAvailable))]
+        public static Version? Version
+        {
+            get
+            {
+                if (!IsAvailable)
+                    return null;
+
+                int major, minor, build;
+                Api->GetApiVersion(&major, &minor, &build);
+                return new Version(major, minor, build);
+            }
+        }
+
+        /// <summary>
+        /// The current mask which determines what sections of the overlay render on each window.
+        /// </summary>
+        [RenderDocApiVersion(1, 0)]
+        public static OverlayBits OverlayBits
+        {
+            get => Api->GetOverlayBits();
+            set
+            {
+                Api->MaskOverlayBits(~value, value);
+            }
+        }
+
+        /// <summary>
+        /// The template for new captures.<br/>
+        /// The template can either be a relative or absolute path, which determines where captures will be saved and how they will be named.
+        /// If the path template is 'my_captures/example', then captures saved will be e.g.
+        /// 'my_captures/example_frame123.rdc' and 'my_captures/example_frame456.rdc'.<br/>
+        /// Relative paths will be saved relative to the process’s current working directory.<br/>
+        /// </summary>
+        /// <remarks>The default template is in a folder controlled by the UI - initially the system temporary folder, and the filename is the executable’s filename.</remarks>
+        [RenderDocApiVersion(1, 0)]
+        public static string CaptureFilePathTemplate
+        {
+            get
+            {
+                byte* ptr = Api->GetCaptureFilePathTemplate();
+                return Marshal.PtrToStringUTF8((nint)ptr)!;
+            }
+            set
+            {
+                fixed (byte* ptr = value.ToNullTerminatedByteArray())
+                {
+                    Api->SetCaptureFilePathTemplate(ptr);
+                }
+            }
+        }
+
+        /// <summary>
+        /// The amount of frame captures that have been made.
+        /// </summary>
+        [RenderDocApiVersion(1, 0)]
+        public static int CaptureCount => Api->GetNumCaptures();
+
+        /// <summary>
+        /// Checks if the RenderDoc UI is currently connected to this process.
+        /// </summary>
+        [RenderDocApiVersion(1, 0)]
+        public static bool IsTargetControlConnected => Api is not null && Api->IsTargetControlConnected() != 0;
+
+        /// <summary>
+        /// Checks if the current frame is capturing.
+        /// </summary>
+        [RenderDocApiVersion(1, 0)]
+        public static bool IsFrameCapturing => Api is not null && Api->IsFrameCapturing() != 0;
+
+        /// <summary>
+        ///     Set one of the options for tweaking some behaviors of capturing.
+        /// </summary>
+        /// <param name="option">specifies which capture option should be set.</param>
+        /// <param name="integer">the unsigned integer value to set for the option.</param>
+        /// <remarks>Note that each option only takes effect from after it is set - so it is advised to set these options as early as possible, ideally before any graphics API has been initialized.</remarks>
+        /// <returns>
+        /// true, if the <paramref name="option"/> is valid, and the value set on the option is within valid ranges.<br/>
+        /// false, if the option is not a <see cref="CaptureOption"/>, or the value is not valid for the option.
+        /// </returns>
+        [RenderDocApiVersion(1, 0)]
+        public static bool SetCaptureOption(CaptureOption option, uint integer)
+        {
+            return Api is not null && Api->SetCaptureOptionU32(option, integer) != 0;
+        }
+
+        /// <summary>
+        /// Set one of the options for tweaking some behaviors of capturing.
+        /// </summary>
+        /// <param name="option">specifies which capture option should be set.</param>
+        /// <param name="boolean">the value to set for the option, converted to a 0 or 1 before setting.</param>
+        /// <remarks>Note that each option only takes effect from after it is set - so it is advised to set these options as early as possible, ideally before any graphics API has been initialized.</remarks>
+        /// <returns>
+        /// true, if the <paramref name="option"/> is valid, and the value set on the option is within valid ranges.<br/>
+        /// false, if the option is not a <see cref="CaptureOption"/>, or the value is not valid for the option.
+        /// </returns>
+        [RenderDocApiVersion(1, 0)]
+        public static bool SetCaptureOption(CaptureOption option, bool boolean)
+            => SetCaptureOption(option, boolean ? 1 : 0);
+
+        /// <summary>
+        /// Set one of the options for tweaking some behaviors of capturing.
+        /// </summary>
+        /// <param name="option">specifies which capture option should be set.</param>
+        /// <param name="single">the floating point value to set for the option.</param>
+        /// <remarks>Note that each option only takes effect from after it is set - so it is advised to set these options as early as possible, ideally before any graphics API has been initialized.</remarks>
+        /// <returns>
+        /// true, if the <paramref name="option"/> is valid, and the value set on the option is within valid ranges.<br/>
+        /// false, if the option is not a <see cref="CaptureOption"/>, or the value is not valid for the option.
+        /// </returns>
+        [RenderDocApiVersion(1, 0)]
+        public static bool SetCaptureOption(CaptureOption option, float single)
+        {
+            return Api is not null && Api->SetCaptureOptionF32(option, single) != 0;
+        }
+
+        /// <summary>
+        /// Gets the current value of one of the different options in <see cref="CaptureOption"/>, writing it to an out parameter.
+        /// </summary>
+        /// <param name="option">specifies which capture option should be retrieved.</param>
+        /// <param name="integer">the value of the capture option, if the option is a valid <see cref="CaptureOption"/> enum. Otherwise, <see cref="int.MaxValue"/>.</param>
+        [RenderDocApiVersion(1, 0)]
+        public static void GetCaptureOption(CaptureOption option, out uint integer)
+        {
+            integer = Api->GetCaptureOptionU32(option);
+        }
+
+        /// <summary>
+        ///     Gets the current value of one of the different options in <see cref="CaptureOption"/>, writing it to an out parameter.
+        /// </summary>
+        /// <param name="option">specifies which capture option should be retrieved.</param>
+        /// <param name="single">the value of the capture option, if the option is a valid <see cref="CaptureOption"/> enum. Otherwise, -<see cref="float.MaxValue"/>.</param>
+        [RenderDocApiVersion(1, 0)]
+        public static void GetCaptureOption(CaptureOption option, out float single)
+        {
+            single = Api->GetCaptureOptionF32(option);
+        }
+
+        /// <summary>
+        /// Gets the current value of one of the different options in <see cref="CaptureOption"/>,
+        /// converted to a boolean.
+        /// </summary>
+        /// <param name="option">specifies which capture option should be retrieved.</param>
+        /// <returns>
+        /// the value of the capture option, converted to bool, if the option is a valid <see cref="CaptureOption"/> enum.
+        /// Otherwise, returns null.
+        /// </returns>
+        [RenderDocApiVersion(1, 0)]
+        public static bool? GetCaptureOptionBool(CaptureOption option)
+        {
+            if (Api is null) return false;
+
+            uint returnVal = GetCaptureOptionU32(option);
+            if (returnVal == uint.MaxValue)
+                return null;
+
+            return returnVal is not 0;
+        }
+
+        /// <summary>
+        /// Gets the current value of one of the different options in <see cref="CaptureOption"/>.
+        /// </summary>
+        /// <param name="option">specifies which capture option should be retrieved.</param>
+        /// <returns>
+        /// the value of the capture option, if the option is a valid <see cref="CaptureOption"/> enum.
+        /// Otherwise, returns <see cref="int.MaxValue"/>.
+        /// </returns>
+        [RenderDocApiVersion(1, 0)]
+        public static uint GetCaptureOptionU32(CaptureOption option) => Api->GetCaptureOptionU32(option);
+
+        /// <summary>
+        /// Gets the current value of one of the different options in <see cref="CaptureOption"/>.
+        /// </summary>
+        /// <param name="option">specifies which capture option should be retrieved.</param>
+        /// <returns>
+        /// the value of the capture option, if the option is a valid <see cref="CaptureOption"/> enum.
+        /// Otherwise, returns -<see cref="float.MaxValue"/>.
+        /// </returns>
+        [RenderDocApiVersion(1, 0)]
+        public static float GetCaptureOptionF32(CaptureOption option) => Api->GetCaptureOptionF32(option);
+
+        /// <summary>
+        /// Changes the key bindings in-application for changing the focussed window.
+        /// </summary>
+        /// <param name="buttons">lists the keys to bind.</param>
+        [RenderDocApiVersion(1, 0)]
+        public static void SetFocusToggleKeys(ReadOnlySpan<InputButton> buttons)
+        {
+            if (Api is null) return;
+
+            fixed (InputButton* ptr = buttons)
+            {
+                Api->SetFocusToggleKeys(ptr, buttons.Length);
+            }
+        }
+
+        /// <summary>
+        /// Changes the key bindings in-application for triggering a capture on the current window.
+        /// </summary>
+        /// <param name="buttons">lists the keys to bind.</param>
+        [RenderDocApiVersion(1, 0)]
+        public static void SetCaptureKeys(ReadOnlySpan<InputButton> buttons)
+        {
+            if (Api is null) return;
+
+            fixed (InputButton* ptr = buttons)
+            {
+                Api->SetCaptureKeys(ptr, buttons.Length);
+            }
+        }
+
+        /// <summary>
+        /// Attempts to remove RenderDoc and its hooks from the target process.<br/>
+        /// It must be called as early as possible in the process, and will have undefined results
+        /// if any graphics API functions have been called.
+        /// </summary>
+        [RenderDocApiVersion(1, 0)]
+        public static void RemoveHooks()
+        {
+            if (Api is null) return;
+
+            Api->RemoveHooks();
+        }
+
+        /// <summary>
+        /// Remove RenderDoc’s crash handler from the target process.<br/>
+        /// If you have your own crash handler that you want to handle any exceptions,
+        /// RenderDoc’s handler could interfere; so it can be disabled.
+        /// </summary>
+        [RenderDocApiVersion(1, 0)]
+        public static void UnloadCrashHandler()
+        {
+            if (Api is null) return;
+
+            Api->UnloadCrashHandler();
+        }
+
+        /// <summary>
+        /// Trigger a capture as if the user had pressed one of the capture hotkeys.<br/>
+        /// The capture will be taken from the next frame presented to whichever window is considered current.
+        /// </summary>
+        [RenderDocApiVersion(1, 0)]
+        public static void TriggerCapture()
+        {
+            if (Api is null) return;
+
+            Api->TriggerCapture();
+        }
+
+
+        /// <summary>
+        /// Gets the details of all frame capture in the current session.
+        /// This simply calls <see cref="GetCapture"/> for each index available as specified by <see cref="CaptureCount"/>.
+        /// </summary>
+        /// <returns>An immutable array of structs representing RenderDoc Captures.</returns>
+        public static ImmutableArray<Capture> GetCaptures()
+        {
+            if (Api is null) return [];
+            int captureCount = CaptureCount;
+            if (captureCount is 0) return [];
+
+            ImmutableArray<Capture>.Builder captures = ImmutableArray.CreateBuilder<Capture>(captureCount);
+
+            for (int captureIndex = 0; captureIndex < captureCount; captureIndex++)
+            {
+                if (GetCapture(captureIndex) is { } capture)
+                    captures.Add(capture);
+            }
+
+            return captures.DrainToImmutable();
+        }
+
+        /// <summary>
+        /// Gets the details of a particular frame capture, as specified by an index from 0 to <see cref="CaptureCount"/> - 1.
+        /// </summary>
+        /// <param name="index">specifies which capture to return the details of. Must be less than the value returned by <see cref="CaptureCount"/>.</param>
+        /// <returns>A struct representing a RenderDoc Capture.</returns>
+        [RenderDocApiVersion(1, 0)]
+        public static Capture? GetCapture(int index)
+        {
+            if (Api is null) return null;
+
+            int length = 0;
+            if (Api->GetCapture(index, null, &length, null) == 0)
+            {
+                return null;
+            }
+
+            Span<byte> bytes = stackalloc byte[length + 1];
+            long timestamp;
+
+            fixed (byte* ptr = bytes)
+                Api->GetCapture(index, ptr, &length, &timestamp);
+
+            string fileName = Encoding.UTF8.GetString(bytes[length..]);
+            return new Capture(index, fileName, DateTimeOffset.FromUnixTimeSeconds(timestamp).DateTime);
+        }
+
+        /// <summary>
+        /// Determine the closest matching replay UI executable for the current RenderDoc module, and launch it.
+        /// </summary>
+        /// <param name="connectTargetControl">if the UI should immediately connect to the application.</param>
+        /// <param name="commandLine">string to be appended to the command line, e.g. a capture filename. If this parameter is null, the command line will be unmodified.</param>
+        /// <returns>true if the UI was successfully launched; false otherwise.</returns>
+        [RenderDocApiVersion(1, 0)]
+        public static bool LaunchReplayUI(bool connectTargetControl, string? commandLine = null)
+        {
+            if (Api is null) return false;
+
+            if (commandLine == null)
+            {
+                return Api->LaunchReplayUI(connectTargetControl ? 1u : 0u, null) != 0;
+            }
+
+            fixed (byte* ptr = commandLine.ToNullTerminatedByteArray())
+            {
+                return Api->LaunchReplayUI(connectTargetControl ? 1u : 0u, ptr) != 0;
+            }
+        }
+
+        /// <summary>
+        /// Explicitly sets which window is considered active.<br/>
+        /// The active window is the one that will be captured when the keybind to trigger a capture is pressed.
+        /// </summary>
+        /// <param name="hDevice">a handle to the API ‘device’ object that will be set active. Must be valid.</param>
+        /// <param name="hWindow">a handle to the platform window handle that will be set active. Must be valid.</param>
+        [RenderDocApiVersion(1, 0)]
+        public static void SetActiveWindow(nint hDevice, nint hWindow)
+        {
+            if (Api is null) return;
+
+            Api->SetActiveWindow((void*)hDevice, (void*)hWindow);
+        }
+
+        /// <summary>
+        /// Immediately begin a capture for the specified device/window combination.
+        /// </summary>
+        /// <param name="hDevice">a handle to the API ‘device’ object that will be set active. May be <see cref="nint.Zero"/> to wildcard match.</param>
+        /// <param name="hWindow">a handle to the platform window handle that will be set active. May be <see cref="nint.Zero"/> to wildcard match.</param>
+        [RenderDocApiVersion(1, 0)]
+        public static void StartFrameCapture(nint hDevice, nint hWindow)
+        {
+            if (Api is null) return;
+
+            Api->StartFrameCapture((void*)hDevice, (void*)hWindow);
+        }
+
+        /// <summary>
+        /// Immediately end an active capture for the specified device/window combination.
+        /// </summary>
+        /// <param name="hDevice">a handle to the API ‘device’ object that will be set active. May be <see cref="nint.Zero"/> to wildcard match.</param>
+        /// <param name="hWindow">a handle to the platform window handle that will be set active. May be <see cref="nint.Zero"/> to wildcard match.</param>
+        /// <returns>true if the capture succeeded; false otherwise.</returns>
+        [RenderDocApiVersion(1, 0)]
+        public static bool EndFrameCapture(nint hDevice, nint hWindow)
+        {
+            if (Api is null) return false;
+
+            return Api->EndFrameCapture((void*)hDevice, (void*)hWindow) != 0;
+        }
+
+        /// <summary>
+        /// Trigger multiple sequential frame captures as if the user had pressed one of the capture hotkeys before each frame.<br/>
+        /// The captures will be taken from the next frames presented to whichever window is considered current.<br/>
+        /// Each capture will be taken independently and saved to a separate file, with no reference to the other frames.
+        /// </summary>
+        /// <param name="numFrames">the number of frames to capture.</param>
+        /// <remarks>Requires RenderDoc API version 1.1</remarks>
+        [RenderDocApiVersion(1, 1)]
+        public static void TriggerMultiFrameCapture(uint numFrames)
+        {
+            if (Api is null) return;
+
+            AssertAtLeast(1, 1);
+            Api->TriggerMultiFrameCapture(numFrames);
+        }
+
+        /// <summary>
+        /// Adds an arbitrary comments field to the most recent capture,
+        /// which will then be displayed in the UI to anyone opening the capture.
+        /// <br/><br/>
+        /// This is equivalent to calling <see cref="SetCaptureFileComments"/> with a null first (fileName) parameter.
+        /// </summary>
+        /// <param name="comments">the comments to set in the capture file.</param>
+        /// <remarks>Requires RenderDoc API version 1.2</remarks>
+        public static void SetMostRecentCaptureFileComments(string comments)
+        {
+            if (Api is null) return;
+
+            AssertAtLeast(1, 2);
+
+            byte[] commentBytes = comments.ToNullTerminatedByteArray();
+
+            fixed (byte* pcomment = commentBytes)
+            {
+                Api->SetCaptureFileComments((byte*)nint.Zero, pcomment);
+            }
+        }
+
+        /// <summary>
+        /// Adds an arbitrary comments field to an existing capture on disk,
+        /// which will then be displayed in the UI to anyone opening the capture.
+        /// </summary>
+        /// <param name="fileName">the path to the capture file to set comments in. If this path is null or an empty string, the most recent capture file that has been created will be used.</param>
+        /// <param name="comments">the comments to set in the capture file.</param>
+        /// <remarks>Requires RenderDoc API version 1.2</remarks>
+        [RenderDocApiVersion(1, 2)]
+        public static void SetCaptureFileComments(string? fileName, string comments)
+        {
+            if (Api is null) return;
+
+            AssertAtLeast(1, 2);
+
+            byte[] commentBytes = comments.ToNullTerminatedByteArray();
+
+            fixed (byte* pcomment = commentBytes)
+            {
+                if (fileName is null)
+                {
+                    Api->SetCaptureFileComments((byte*)nint.Zero, pcomment);
+                }
+                else
+                {
+                    byte[] fileBytes = fileName.ToNullTerminatedByteArray();
+
+                    fixed (byte* pfile = fileBytes)
+                    {
+                        Api->SetCaptureFileComments(pfile, pcomment);
+                    }
+                }
+            }
+        }
+
+        /// <summary>
+        /// Similar to <see cref="EndFrameCapture"/>, but the capture contents will be discarded immediately, and not processed and written to disk.<br/>
+        /// This will be more efficient than <see cref="EndFrameCapture"/> if the frame capture is not needed.
+        /// </summary>
+        /// <param name="hDevice">a handle to the API ‘device’ object that will be set active. May be <see cref="nint.Zero"/> to wildcard match.</param>
+        /// <param name="hWindow">a handle to the platform window handle that will be set active. May be <see cref="nint.Zero"/> to wildcard match.</param>
+        /// <returns>true if the capture was discarded; false if there was an error or no capture was in progress.</returns>
+        /// <remarks>Requires RenderDoc API version 1.4</remarks>
+        [RenderDocApiVersion(1, 4)]
+        public static bool DiscardFrameCapture(nint hDevice, nint hWindow)
+        {
+            if (Api is null) return false;
+
+            AssertAtLeast(1, 4);
+            return Api->DiscardFrameCapture((void*)hDevice, (void*)hWindow) != 0;
+        }
+
+
+        /// <summary>
+        /// Requests that the currently connected replay UI raise its window to the top.<br/>
+        /// This is only possible if an instance of the replay UI is currently connected, otherwise this method does nothing.<br/>
+        /// This can be used in conjunction with <see cref="IsTargetControlConnected"/> and <see cref="LaunchReplayUI"/>,<br/> to intelligently handle showing the UI after making a capture.<br/><br/>
+        /// Given OS differences, it is not guaranteed that the UI will be successfully raised even if the request is passed on.<br/>
+        /// On some systems it may only be highlighted or otherwise indicated to the user.
+        /// </summary>
+        /// <returns>true if the request was passed onto the UI successfully; false if there is no UI connected or some other error occurred.</returns>
+        /// <remarks>Requires RenderDoc API version 1.5</remarks>
+        [RenderDocApiVersion(1, 5)]
+        public static bool ShowReplayUI()
+        {
+            if (Api is null) return false;
+
+            AssertAtLeast(1, 5);
+            return Api->ShowReplayUI() != 0;
+        }
+
+        /// <summary>
+        /// Sets a given title for the currently in-progress capture, which will be displayed in the UI.<br/>
+        /// This can be used either with a user-defined capture using a manual start and end,
+        /// or an automatic capture triggered by <see cref="TriggerCapture"/> or a keypress.<br/>
+        /// If multiple captures are ongoing at once, the title will be applied to the first capture to end only.<br/>
+        /// Any subsequent captures will not get any title unless the function is called again.
+        /// This function can only be called while a capture is in-progress,
+        /// after <see cref="StartFrameCapture"/> and before <see cref="EndFrameCapture"/>.<br/>
+        /// If it is called elsewhere it will have no effect.
+        /// If it is called multiple times within a capture, only the last title will have any effect.
+        /// </summary>
+        /// <param name="title">The title to set for the in-progress capture.</param>
+        /// <remarks>Requires RenderDoc API version 1.6</remarks>
+        [RenderDocApiVersion(1, 6)]
+        public static void SetCaptureTitle(string title)
+        {
+            if (Api is null) return;
+
+            AssertAtLeast(1, 6);
+            fixed (byte* ptr = title.ToNullTerminatedByteArray())
+                Api->SetCaptureTitle(ptr);
+        }
+
+        #region Dynamic Library loading
+
+        /// <summary>
+        /// Reload the internal RenderDoc API structure. Useful for manually refreshing <see cref="Api"/> while using process injection.
+        /// </summary>
+        /// <param name="ignoreAlreadyLoaded">Ignores the existing API function structure and overwrites it with a re-request.</param>
+        /// <param name="requiredVersion">The version of the RenderDoc API required by your application.</param>
+        public static void ReloadApi(bool ignoreAlreadyLoaded = false, RenderDocVersion? requiredVersion = null)
+        {
+            if (_loaded && !ignoreAlreadyLoaded)
+                return;
+
+            lock (typeof(RenderDoc))
+            {
+                // Prevent double loads.
+                if (_loaded && !ignoreAlreadyLoaded)
+                    return;
+
+                if (requiredVersion.HasValue)
+                    MinimumRequired = requiredVersion.Value;
+
+                _loaded = true;
+                _api = GetApi(MinimumRequired);
+
+                if (_api != null)
+                    AssertAtLeast(MinimumRequired);
+            }
+        }
+
+        private static RenderDocApi* _api = null;
+        private static bool _loaded;
+
+        private static RenderDocApi* Api
+        {
+            get
+            {
+                ReloadApi();
+                return _api;
+            }
+        }
+
+        private static readonly Regex _dynamicLibraryPattern = RenderDocApiDynamicLibraryRegex();
+
+        private static RenderDocApi* GetApi(RenderDocVersion minimumRequired = RenderDocVersion.Version_1_0_0)
+        {
+            foreach (ProcessModule module in Process.GetCurrentProcess().Modules)
+            {
+                string moduleName = module.FileName ?? string.Empty;
+
+                if (!_dynamicLibraryPattern.IsMatch(moduleName))
+                    continue;
+
+                if (!NativeLibrary.TryLoad(moduleName, out nint moduleHandle))
+                    return null;
+
+                if (!NativeLibrary.TryGetExport(moduleHandle, "RENDERDOC_GetAPI", out nint procAddress))
+                    return null;
+
+                var RENDERDOC_GetApi = (delegate* unmanaged[Cdecl]<RenderDocVersion, RenderDocApi**, int>)procAddress;
+
+                RenderDocApi* api;
+                return RENDERDOC_GetApi(minimumRequired, &api) != 0 ? api : null;
+            }
+
+            return null;
+        }
+
+        private static void AssertAtLeast(RenderDocVersion rdv, [CallerMemberName] string callee = "")
+        {
+            Version ver = rdv.SystemVersion;
+            AssertAtLeast(ver.Major, ver.Minor, ver.Build, callee);
+        }
+
+        private static void AssertAtLeast(int major, int minor, int patch = 0, [CallerMemberName] string callee = "")
+        {
+            if (!AssertVersionEnabled)
+                return;
+
+            if (Version!.Major < major)
+                goto fail;
+
+            if (Version.Major > major)
+                goto success;
+            if (Version.Minor < minor)
+                goto fail;
+            if (Version.Minor > minor)
+                goto success;
+            if (Version.Build < patch)
+                goto fail;
+
+            success:
+            return;
+
+            fail:
+            Version minVersion =
+                typeof(RenderDoc).GetMethod(callee)!.GetCustomAttribute<RenderDocApiVersionAttribute>()!.MinVersion;
+            throw new NotSupportedException(
+                $"This API was introduced in RenderDoc API {minVersion}. Current API version is {Version}.");
+        }
+
+        private static byte[] ToNullTerminatedByteArray(this string str, Encoding? encoding = null)
+        {
+            encoding ??= Encoding.UTF8;
+
+            return encoding.GetBytes(str + '\0');
+        }
+
+        [GeneratedRegex(@"(lib)?renderdoc(\.dll|\.so|\.dylib)(\.\d+)?",
+            RegexOptions.IgnoreCase | RegexOptions.CultureInvariant)]
+        private static partial Regex RenderDocApiDynamicLibraryRegex();
+
+        #endregion
+    }
+}

src/Ryujinx.Graphics.RenderDocApi/RenderDocApi.cs

@@ -0,0 +1,51 @@
+namespace Ryujinx.Graphics.RenderDocApi
+{
+#pragma warning disable CS0649
+    internal unsafe struct RenderDocApi
+    {
+        public delegate* unmanaged[Cdecl]<int*, int*, int*, void> GetApiVersion;
+
+        public delegate* unmanaged[Cdecl]<CaptureOption, uint, int> SetCaptureOptionU32;
+        public delegate* unmanaged[Cdecl]<CaptureOption, float, int> SetCaptureOptionF32;
+        public delegate* unmanaged[Cdecl]<CaptureOption, uint> GetCaptureOptionU32;
+        public delegate* unmanaged[Cdecl]<CaptureOption, float> GetCaptureOptionF32;
+
+        public delegate* unmanaged[Cdecl]<InputButton*, int, void> SetFocusToggleKeys;
+        public delegate* unmanaged[Cdecl]<InputButton*, int, void> SetCaptureKeys;
+
+        public delegate* unmanaged[Cdecl]<OverlayBits> GetOverlayBits;
+        public delegate* unmanaged[Cdecl]<OverlayBits, OverlayBits, void> MaskOverlayBits;
+
+        public delegate* unmanaged[Cdecl]<void> RemoveHooks;
+        public delegate* unmanaged[Cdecl]<void> UnloadCrashHandler;
+        public delegate* unmanaged[Cdecl]<byte*, void> SetCaptureFilePathTemplate;
+        public delegate* unmanaged[Cdecl]<byte*> GetCaptureFilePathTemplate;
+
+        public delegate* unmanaged[Cdecl]<int> GetNumCaptures;
+        public delegate* unmanaged[Cdecl]<int, byte*, int*, long*, uint> GetCapture;
+        public delegate* unmanaged[Cdecl]<void> TriggerCapture;
+        public delegate* unmanaged[Cdecl]<uint> IsTargetControlConnected;
+        public delegate* unmanaged[Cdecl]<uint, byte*, uint> LaunchReplayUI;
+
+        public delegate* unmanaged[Cdecl]<void*, void*, void> SetActiveWindow;
+        public delegate* unmanaged[Cdecl]<void*, void*, void> StartFrameCapture;
+        public delegate* unmanaged[Cdecl]<uint> IsFrameCapturing;
+        public delegate* unmanaged[Cdecl]<void*, void*, uint> EndFrameCapture;
+
+        // 1.1
+        public delegate* unmanaged[Cdecl]<uint, void> TriggerMultiFrameCapture;
+
+        // 1.2
+        public delegate* unmanaged[Cdecl]<byte*, byte*, void> SetCaptureFileComments;
+
+        // 1.3
+        public delegate* unmanaged[Cdecl]<void*, void*, uint> DiscardFrameCapture;
+
+        // 1.5
+        public delegate* unmanaged[Cdecl]<uint> ShowReplayUI;
+
+        // 1.6
+        public delegate* unmanaged[Cdecl]<byte*, void> SetCaptureTitle;
+    }
+#pragma warning restore CS0649
+}

src/Ryujinx.Graphics.RenderDocApi/RenderDocApiVersionAttribute.cs

@@ -0,0 +1,16 @@
+
+using System;
+
+namespace Ryujinx.Graphics.RenderDocApi
+{
+    [AttributeUsage(AttributeTargets.Method | AttributeTargets.Property)]
+    public sealed class RenderDocApiVersionAttribute : Attribute
+    {
+        public Version MinVersion { get; }
+
+        public RenderDocApiVersionAttribute(int major, int minor, int patch = 0)
+        {
+            MinVersion = new Version(major, minor, patch);
+        }
+    }
+}

src/Ryujinx.Graphics.RenderDocApi/RenderDocVersion.cs

@@ -0,0 +1,47 @@
+using System;
+
+namespace Ryujinx.Graphics.RenderDocApi
+{
+    public enum RenderDocVersion
+    {
+        Version_1_0_0 = 10000,
+        Version_1_0_1 = 10001,
+        Version_1_0_2 = 10002,
+        Version_1_1_0 = 10100,
+        Version_1_1_1 = 10101,
+        Version_1_1_2 = 10102,
+        Version_1_2_0 = 10200,
+        Version_1_3_0 = 10300,
+        Version_1_4_0 = 10400,
+        Version_1_4_1 = 10401,
+        Version_1_4_2 = 10402,
+        Version_1_5_0 = 10500,
+        Version_1_6_0 = 10600,
+    }
+    
+    public static partial class Helpers
+    {
+        extension(RenderDocVersion rdv)
+        {
+            public Version SystemVersion
+            {
+                get
+                {
+                    int i = (int)rdv;
+                    return new (i / 10000, (i % 10000) / 100, i % 100);
+                }
+            }
+        }
+
+        extension(Version sv)
+        {
+            public RenderDocVersion RenderDocVersion
+            {
+                get
+                {
+                    return (RenderDocVersion)(sv.Major * 10000 + sv.Minor * 100 + sv.Build);
+                }
+            }
+        }
+    }
+}

src/Ryujinx.Graphics.RenderDocApi/Ryujinx.Graphics.RenderDocApi.csproj

@@ -0,0 +1,9 @@
+<Project Sdk="Microsoft.NET.Sdk">
+
+    <PropertyGroup>
+        <TargetFramework>net10.0</TargetFramework>
+        <ImplicitUsings>disable</ImplicitUsings>
+        <Nullable>enable</Nullable>
+        <AllowUnsafeBlocks>true</AllowUnsafeBlocks>
+    </PropertyGroup>
+</Project>

src/Ryujinx.Graphics.Shader/Decoders/InstDecoders.cs

@@ -522,6 +522,7 @@ namespace Ryujinx.Graphics.Shader.Decoders
 
     enum PMode
     {
+        Idx = 0,
         F4e = 1,
         B4e = 2,
         Rc8 = 3,

src/Ryujinx.Graphics.Shader/IGpuAccessor.cs

@@ -234,6 +234,24 @@ namespace Ryujinx.Graphics.Shader
             return true;
         }
 
+        /// <summary>
+        /// Queries host support for buffer image access with the buffer offset aligned to a single pixel rather than a fixed alignment.
+        /// </summary>
+        /// <returns>True if the host supports buffer image access with pixel alignment, false otherwise</returns>
+        bool QueryHostSupportsBufferImagePixelAlignment()
+        {
+            return true;
+        }
+
+        /// <summary>
+        /// Queries host support for buffer texture access with the buffer offset aligned to a single pixel rather than a fixed alignment.
+        /// </summary>
+        /// <returns>True if the host supports buffer texture access with pixel alignment, false otherwise</returns>
+        bool QueryHostSupportsBufferTexturePixelAlignment()
+        {
+            return true;
+        }
+
         /// <summary>
         /// Queries host support for fragment shader ordering critical sections on the shader code.
         /// </summary>

src/Ryujinx.Graphics.Shader/Instructions/InstEmit.cs

@@ -215,34 +215,6 @@ namespace Ryujinx.Graphics.Shader.Instructions
             context.TranslatorContext.GpuAccessor.Log("Shader instruction Pret is not implemented.");
         }
 
-        public static void PrmtR(EmitterContext context)
-        {
-            context.GetOp<InstPrmtR>();
-
-            context.TranslatorContext.GpuAccessor.Log("Shader instruction PrmtR is not implemented.");
-        }
-
-        public static void PrmtI(EmitterContext context)
-        {
-            context.GetOp<InstPrmtI>();
-
-            context.TranslatorContext.GpuAccessor.Log("Shader instruction PrmtI is not implemented.");
-        }
-
-        public static void PrmtC(EmitterContext context)
-        {
-            context.GetOp<InstPrmtC>();
-
-            context.TranslatorContext.GpuAccessor.Log("Shader instruction PrmtC is not implemented.");
-        }
-
-        public static void PrmtRc(EmitterContext context)
-        {
-            context.GetOp<InstPrmtRc>();
-
-            context.TranslatorContext.GpuAccessor.Log("Shader instruction PrmtRc is not implemented.");
-        }
-
         public static void R2b(EmitterContext context)
         {
             context.GetOp<InstR2b>();

src/Ryujinx.Graphics.Shader/Instructions/InstEmitMove.cs

@@ -1,7 +1,7 @@
 using Ryujinx.Graphics.Shader.Decoders;
 using Ryujinx.Graphics.Shader.IntermediateRepresentation;
 using Ryujinx.Graphics.Shader.Translation;
-
+using System;
 using static Ryujinx.Graphics.Shader.Instructions.InstEmitHelper;
 using static Ryujinx.Graphics.Shader.IntermediateRepresentation.OperandHelper;
 
@@ -37,6 +37,38 @@ namespace Ryujinx.Graphics.Shader.Instructions
             context.Copy(GetDest(op.Dest), GetSrcImm(context, op.Imm32));
         }
 
+        public static void PrmtR(EmitterContext context)
+        {
+            context.GetOp<InstPrmtR>();
+
+            context.TranslatorContext.GpuAccessor.Log("Shader instruction PrmtR is not implemented.");
+        }
+
+        public static void PrmtI(EmitterContext context)
+        {
+            InstPrmtI op = context.GetOp<InstPrmtI>();
+
+            Operand op1 = GetSrcReg(context, op.SrcA);
+            Operand control = GetSrcImm(context, op.Imm20);
+            Operand op2 = GetSrcReg(context, op.SrcC);
+
+            EmitPrmt(context, op1, control, op2, op.PMode, op.Dest);
+        }
+
+        public static void PrmtC(EmitterContext context)
+        {
+            context.GetOp<InstPrmtC>();
+
+            context.TranslatorContext.GpuAccessor.Log("Shader instruction PrmtC is not implemented.");
+        }
+
+        public static void PrmtRc(EmitterContext context)
+        {
+            context.GetOp<InstPrmtRc>();
+
+            context.TranslatorContext.GpuAccessor.Log("Shader instruction PrmtRc is not implemented.");
+        }
+
         public static void R2pR(EmitterContext context)
         {
             InstR2pR op = context.GetOp<InstR2pR>();
@@ -169,6 +201,39 @@ namespace Ryujinx.Graphics.Shader.Instructions
             context.Copy(GetDest(op.Dest), src);
         }
 
+        public static void SelR(EmitterContext context)
+        {
+            InstSelR op = context.GetOp<InstSelR>();
+
+            Operand srcA = GetSrcReg(context, op.SrcA);
+            Operand srcB = GetSrcReg(context, op.SrcB);
+            Operand srcPred = GetPredicate(context, op.SrcPred, op.SrcPredInv);
+
+            EmitSel(context, srcA, srcB, srcPred, op.Dest);
+        }
+
+        public static void SelI(EmitterContext context)
+        {
+            InstSelI op = context.GetOp<InstSelI>();
+
+            Operand srcA = GetSrcReg(context, op.SrcA);
+            Operand srcB = GetSrcImm(context, Imm20ToSInt(op.Imm20));
+            Operand srcPred = GetPredicate(context, op.SrcPred, op.SrcPredInv);
+
+            EmitSel(context, srcA, srcB, srcPred, op.Dest);
+        }
+
+        public static void SelC(EmitterContext context)
+        {
+            InstSelC op = context.GetOp<InstSelC>();
+
+            Operand srcA = GetSrcReg(context, op.SrcA);
+            Operand srcB = GetSrcCbuf(context, op.CbufSlot, op.CbufOffset);
+            Operand srcPred = GetPredicate(context, op.SrcPred, op.SrcPredInv);
+
+            EmitSel(context, srcA, srcB, srcPred, op.Dest);
+        }
+
         private static Operand EmitLoadSubgroupLaneId(EmitterContext context)
         {
             if (context.TranslatorContext.GpuAccessor.QueryHostSubgroupSize() <= 32)
@@ -215,37 +280,34 @@ namespace Ryujinx.Graphics.Shader.Instructions
             }
         }
 
-        public static void SelR(EmitterContext context)
+        private static void EmitPrmt(EmitterContext context, Operand op1, Operand control, Operand op2, PMode pMode, int rd)
         {
-            InstSelR op = context.GetOp<InstSelR>();
+            if (pMode == PMode.Idx)
+            {
+                Operand res = Const(0);
 
-            Operand srcA = GetSrcReg(context, op.SrcA);
-            Operand srcB = GetSrcReg(context, op.SrcB);
-            Operand srcPred = GetPredicate(context, op.SrcPred, op.SrcPredInv);
+                for (int b = 0; b < 4; b++)
+                {
+                    Operand sel = context.ShiftRightU32(control, Const(b * 4));
+                    Operand byteSel = context.BitwiseAnd(sel, Const(7));
+                    Operand copySign = context.BitwiseAnd(sel, Const(8));
 
-            EmitSel(context, srcA, srcB, srcPred, op.Dest);
-        }
+                    Operand srcOp = context.ConditionalSelect(context.BitwiseAnd(byteSel, Const(4)), op2, op1);
+                    Operand srcValue = context.ShiftRightU32(srcOp, context.ShiftLeft(context.BitwiseAnd(byteSel, Const(3)), Const(3)));
+                    Operand srcSign = context.ShiftRightS32(context.ShiftLeft(srcValue, Const(24)), Const(31));
 
-        public static void SelI(EmitterContext context)
-        {
-            InstSelI op = context.GetOp<InstSelI>();
+                    srcValue = context.ConditionalSelect(copySign, srcSign, srcValue);
+                    srcValue = context.BitwiseAnd(srcValue, Const(0xff));
 
-            Operand srcA = GetSrcReg(context, op.SrcA);
-            Operand srcB = GetSrcImm(context, Imm20ToSInt(op.Imm20));
-            Operand srcPred = GetPredicate(context, op.SrcPred, op.SrcPredInv);
+                    res = context.BitfieldInsert(res, srcValue, Const(b * 8), Const(8));
+                }
 
-            EmitSel(context, srcA, srcB, srcPred, op.Dest);
-        }
-
-        public static void SelC(EmitterContext context)
-        {
-            InstSelC op = context.GetOp<InstSelC>();
-
-            Operand srcA = GetSrcReg(context, op.SrcA);
-            Operand srcB = GetSrcCbuf(context, op.CbufSlot, op.CbufOffset);
-            Operand srcPred = GetPredicate(context, op.SrcPred, op.SrcPredInv);
-
-            EmitSel(context, srcA, srcB, srcPred, op.Dest);
+                context.Copy(GetDest(rd), res);
+            }
+            else
+            {
+                throw new NotImplementedException(pMode.ToString());
+            }
         }
 
         private static void EmitR2p(EmitterContext context, Operand value, Operand mask, ByteSel byteSel, bool ccpr)

src/Ryujinx.Graphics.Shader/IntermediateRepresentation/BasicBlock.cs

@@ -23,6 +23,7 @@ namespace Ryujinx.Graphics.Shader.IntermediateRepresentation
             set => _branch = AddSuccessor(_branch, value);
         }
 
+        public bool HasSuccessor => _branch != null || _next != null;
         public bool HasBranch => _branch != null;
         public bool Reachable => Index == 0 || Predecessors.Count != 0;
 

src/Ryujinx.Graphics.Shader/ShaderProgramInfo.cs

@@ -18,6 +18,7 @@ namespace Ryujinx.Graphics.Shader
         public bool UsesInstanceId { get; }
         public bool UsesDrawParameters { get; }
         public bool UsesRtLayer { get; }
+        public bool HasUnconditionalDiscard { get; }
         public byte ClipDistancesWritten { get; }
         public int FragmentOutputMap { get; }
 
@@ -34,6 +35,7 @@ namespace Ryujinx.Graphics.Shader
             bool usesInstanceId,
             bool usesDrawParameters,
             bool usesRtLayer,
+            bool hasUnconditionalDiscard,
             byte clipDistancesWritten,
             int fragmentOutputMap)
         {
@@ -50,6 +52,7 @@ namespace Ryujinx.Graphics.Shader
             UsesInstanceId = usesInstanceId;
             UsesDrawParameters = usesDrawParameters;
             UsesRtLayer = usesRtLayer;
+            HasUnconditionalDiscard = hasUnconditionalDiscard;
             ClipDistancesWritten = clipDistancesWritten;
             FragmentOutputMap = fragmentOutputMap;
         }

src/Ryujinx.Graphics.Shader/SupportBuffer.cs

@@ -24,6 +24,7 @@ namespace Ryujinx.Graphics.Shader
         RenderScale,
         TfeOffset,
         TfeVertexCount,
+        BufferTextureOffset,
     }
 
     public struct SupportBuffer
@@ -42,10 +43,13 @@ namespace Ryujinx.Graphics.Shader
         public static readonly int ComputeRenderScaleOffset;
         public static readonly int TfeOffsetOffset;
         public static readonly int TfeVertexCountOffset;
+        public static readonly int BufferTextureOffsetOffset;
 
         public const int FragmentIsBgraCount = 8;
+        public const int TextureCount = 64;
+
         // One for the render target, 64 for the textures, and 8 for the images.
-        public const int RenderScaleMaxCount = 1 + 64 + 8;
+        public const int RenderScaleMaxCount = 1 + TextureCount + 8;
 
         private static int OffsetOf<T>(ref SupportBuffer storage, ref T target)
         {
@@ -68,6 +72,7 @@ namespace Ryujinx.Graphics.Shader
             ComputeRenderScaleOffset = GraphicsRenderScaleOffset + FieldSize;
             TfeOffsetOffset = OffsetOf(ref instance, ref instance.TfeOffset);
             TfeVertexCountOffset = OffsetOf(ref instance, ref instance.TfeVertexCount);
+            BufferTextureOffsetOffset = OffsetOf(ref instance, ref instance.BufferTextureOffset);
         }
 
         internal static StructureType GetStructureType()
@@ -80,7 +85,8 @@ namespace Ryujinx.Graphics.Shader
                 new StructureField(AggregateType.S32, "frag_scale_count"),
                 new StructureField(AggregateType.Array | AggregateType.FP32, "render_scale", RenderScaleMaxCount),
                 new StructureField(AggregateType.Vector4 | AggregateType.S32, "tfe_offset"),
-                new StructureField(AggregateType.S32, "tfe_vertex_count")
+                new StructureField(AggregateType.S32, "tfe_vertex_count"),
+                new StructureField(AggregateType.Array | AggregateType.Vector2 | AggregateType.S32, "buffer_texture_offset")
             ]);
         }
 
@@ -95,5 +101,7 @@ namespace Ryujinx.Graphics.Shader
 
         public Vector4<int> TfeOffset;
         public Vector4<int> TfeVertexCount;
+
+        public Array5<Array64<Vector4<int>>> BufferTextureOffset;
     }
 }

src/Ryujinx.Graphics.Shader/Translation/FeatureFlags.cs

@@ -26,5 +26,6 @@ namespace Ryujinx.Graphics.Shader.Translation
         SharedMemory = 1 << 11,
         Store = 1 << 12,
         VtgAsCompute = 1 << 13,
+        UnconditionalDiscard = 1 << 14,
     }
 }

src/Ryujinx.Graphics.Shader/Translation/FeatureIdentification.cs

@@ -0,0 +1,38 @@
+using Ryujinx.Graphics.Shader.IntermediateRepresentation;
+
+namespace Ryujinx.Graphics.Shader.Translation
+{
+    static class FeatureIdentification
+    {
+        public static void RunPass(BasicBlock[] blocks, ShaderStage stage, ref FeatureFlags usedFeatures)
+        {
+            if (stage == ShaderStage.Fragment)
+            {
+                bool endsWithDiscardOnly = true;
+
+                for (int blockIndex = 0; blockIndex < blocks.Length; blockIndex++)
+                {
+                    BasicBlock block = blocks[blockIndex];
+
+                    if (block.HasSuccessor)
+                    {
+                        continue;
+                    }
+
+                    if (block.Operations.Count == 0 ||
+                        block.Operations.Last.Value is not Operation operation ||
+                        operation.Inst != Instruction.Discard)
+                    {
+                        endsWithDiscardOnly = false;
+                        break;
+                    }
+                }
+
+                if (endsWithDiscardOnly)
+                {
+                    usedFeatures |= FeatureFlags.UnconditionalDiscard;
+                }
+            }
+        }
+    }
+}

src/Ryujinx.Graphics.Shader/Translation/ResourceManager.cs

@@ -43,6 +43,11 @@ namespace Ryujinx.Graphics.Shader.Translation
         private readonly Dictionary<TextureInfo, TextureMeta> _usedTextures;
         private readonly Dictionary<TextureInfo, TextureMeta> _usedImages;
 
+        private readonly List<BufferDefinition> _vacConstantBuffers;
+        private readonly List<BufferDefinition> _vacStorageBuffers;
+        private readonly List<TextureDefinition> _vacTextures;
+        private readonly List<TextureDefinition> _vacImages;
+
         public int LocalMemoryId { get; private set; }
         public int SharedMemoryId { get; private set; }
 
@@ -78,6 +83,11 @@ namespace Ryujinx.Graphics.Shader.Translation
             _usedTextures = new();
             _usedImages = new();
 
+            _vacConstantBuffers = new();
+            _vacStorageBuffers = new();
+            _vacTextures = new();
+            _vacImages = new();
+
             Properties.AddOrUpdateConstantBuffer(new(BufferLayout.Std140, 0, SupportBuffer.Binding, "support_buffer", SupportBuffer.GetStructureType()));
 
             LocalMemoryId = -1;
@@ -563,6 +573,76 @@ namespace Ryujinx.Graphics.Shader.Translation
             return descriptors.ToArray();
         }
 
+        public ShaderProgramInfo GetVertexAsComputeInfo(bool isVertex = false)
+        {
+            var cbDescriptors = new BufferDescriptor[_vacConstantBuffers.Count];
+            int cbDescriptorIndex = 0;
+
+            foreach (BufferDefinition definition in _vacConstantBuffers)
+            {
+                cbDescriptors[cbDescriptorIndex++] = new BufferDescriptor(definition.Set, definition.Binding, 0, 0, 0, BufferUsageFlags.None);
+            }
+
+            var sbDescriptors = new BufferDescriptor[_vacStorageBuffers.Count];
+            int sbDescriptorIndex = 0;
+
+            foreach (BufferDefinition definition in _vacStorageBuffers)
+            {
+                sbDescriptors[sbDescriptorIndex++] = new BufferDescriptor(definition.Set, definition.Binding, 0, 0, 0, BufferUsageFlags.Write);
+            }
+
+            var tDescriptors = new TextureDescriptor[_vacTextures.Count];
+            int tDescriptorIndex = 0;
+
+            foreach (TextureDefinition definition in _vacTextures)
+            {
+                tDescriptors[tDescriptorIndex++] = new TextureDescriptor(
+                    definition.Set,
+                    definition.Binding,
+                    definition.Type,
+                    definition.Format,
+                    0,
+                    0,
+                    definition.ArrayLength,
+                    definition.Separate,
+                    definition.Flags);
+            }
+
+            var iDescriptors = new TextureDescriptor[_vacImages.Count];
+            int iDescriptorIndex = 0;
+
+            foreach (TextureDefinition definition in _vacImages)
+            {
+                iDescriptors[iDescriptorIndex++] = new TextureDescriptor(
+                    definition.Set,
+                    definition.Binding,
+                    definition.Type,
+                    definition.Format,
+                    0,
+                    0,
+                    definition.ArrayLength,
+                    definition.Separate,
+                    definition.Flags);
+            }
+
+            return new ShaderProgramInfo(
+                cbDescriptors,
+                sbDescriptors,
+                tDescriptors,
+                iDescriptors,
+                isVertex ? ShaderStage.Vertex : ShaderStage.Compute,
+                0,
+                0,
+                0,
+                false,
+                false,
+                false,
+                false,
+                false,
+                0,
+                0);
+        }
+
         public bool TryGetCbufSlotAndHandleForTexture(int binding, out int cbufSlot, out int handle)
         {
             foreach ((TextureInfo info, TextureMeta meta) in _usedTextures)
@@ -627,6 +707,30 @@ namespace Ryujinx.Graphics.Shader.Translation
             Properties.AddOrUpdateStorageBuffer(new(BufferLayout.Std430, setIndex, binding, name, type));
         }
 
+        public void AddVertexAsComputeConstantBuffer(BufferDefinition definition)
+        {
+            _vacConstantBuffers.Add(definition);
+            Properties.AddOrUpdateConstantBuffer(definition);
+        }
+
+        public void AddVertexAsComputeStorageBuffer(BufferDefinition definition)
+        {
+            _vacStorageBuffers.Add(definition);
+            Properties.AddOrUpdateStorageBuffer(definition);
+        }
+
+        public void AddVertexAsComputeTexture(TextureDefinition definition)
+        {
+            _vacTextures.Add(definition);
+            Properties.AddOrUpdateTexture(definition);
+        }
+
+        public void AddVertexAsComputeImage(TextureDefinition definition)
+        {
+            _vacImages.Add(definition);
+            Properties.AddOrUpdateImage(definition);
+        }
+
         public static string GetShaderStagePrefix(ShaderStage stage)
         {
             uint index = (uint)stage;

src/Ryujinx.Graphics.Shader/Translation/Transforms/TexturePass.cs

@@ -1,4 +1,5 @@
 using Ryujinx.Graphics.Shader.IntermediateRepresentation;
+using Ryujinx.Graphics.Shader.Translation.Optimizations;
 using System.Collections.Generic;
 using System.Linq;
 using static Ryujinx.Graphics.Shader.IntermediateRepresentation.OperandHelper;
@@ -27,7 +28,19 @@ namespace Ryujinx.Graphics.Shader.Translation.Transforms
 
                     if (texOp.Type == SamplerType.TextureBuffer && !context.GpuAccessor.QueryHostSupportsSnormBufferTextureFormat())
                     {
-                        node = InsertSnormNormalization(node, context.ResourceManager, context.GpuAccessor);
+                        if (!context.GpuAccessor.QueryHostSupportsSnormBufferTextureFormat())
+                        {
+                            node = InsertSnormNormalization(node, context.ResourceManager, context.GpuAccessor);
+                        }
+
+                        if (!context.GpuAccessor.QueryHostSupportsBufferTexturePixelAlignment())
+                        {
+                            node = InsertCoordOffset(node, context.ResourceManager, context.Stage, isImage: false);
+                        }
+                    }
+                    else if (!context.GpuAccessor.QueryHostSupportsBufferTexturePixelAlignment() && texOp.Inst.IsImage())
+                    {
+                        node = InsertCoordOffset(node, context.ResourceManager, context.Stage, isImage: true);
                     }
                 }
             }
@@ -278,6 +291,87 @@ namespace Ryujinx.Graphics.Shader.Translation.Transforms
             return node;
         }
 
+        private static LinkedListNode<INode> InsertCoordOffset(LinkedListNode<INode> node, ResourceManager resourceManager, ShaderStage stage, bool isImage)
+        {
+            // Some GPUs have fixed alignment requirements for buffer textures.
+            // For those cases, we bind the aligned buffer offset, and apply the remaining offset on the shader.
+
+            TextureOperation texOp = (TextureOperation)node.Value;
+
+            if ((texOp.Type & SamplerType.Mask) != SamplerType.TextureBuffer)
+            {
+                return node;
+            }
+
+            Operand[] sources = new Operand[texOp.SourcesCount];
+
+            for (int i = 0; i < texOp.SourcesCount; i++)
+            {
+                sources[i] = texOp.GetSource(i);
+            }
+
+            bool isBindless = (texOp.Flags & TextureFlags.Bindless) != 0;
+            bool isIndexed = resourceManager.IsArrayOfTexturesOrImages(texOp.Binding, isImage);
+
+            int coordsIndex = isBindless || isIndexed ? 1 : 0;
+
+            Operand[] dests = new Operand[texOp.DestsCount];
+
+            for (int i = 0; i < texOp.DestsCount; i++)
+            {
+                dests[i] = texOp.GetDest(i);
+            }
+
+            LinkedListNode<INode> oldNode = node;
+
+            Operand source = sources[coordsIndex];
+            Operand offset = Local();
+            Operand coordPlusOffset = Local();
+
+            int stageIndex = stage switch
+            {
+                ShaderStage.TessellationControl => 1,
+                ShaderStage.TessellationEvaluation => 2,
+                ShaderStage.Geometry => 3,
+                ShaderStage.Fragment => 4,
+                _ => 0,
+            };
+
+            int bindingIndex = isImage
+                ? resourceManager.FindImageDescriptorIndex(texOp.Binding)
+                : resourceManager.FindTextureDescriptorIndex(texOp.Binding);
+
+            node.List.AddBefore(node, new Operation(
+                Instruction.Load,
+                StorageKind.ConstantBuffer,
+                offset,
+                Const(SupportBuffer.Binding),
+                Const((int)SupportBufferField.BufferTextureOffset),
+                Const(stageIndex * SupportBuffer.TextureCount + bindingIndex),
+                Const(isImage ? 1 : 0)));
+
+            node.List.AddBefore(node, new Operation(Instruction.Add, coordPlusOffset, source, offset));
+
+            sources[coordsIndex] = coordPlusOffset;
+
+            TextureOperation newTexOp = new(
+                texOp.Inst,
+                texOp.Type,
+                texOp.Format,
+                texOp.Flags,
+                texOp.Set,
+                texOp.Binding,
+                texOp.Index,
+                dests,
+                sources);
+
+            node = node.List.AddBefore(node, newTexOp);
+
+            Utils.DeleteNode(oldNode, texOp);
+
+            return node;
+        }
+
         private static LinkedListNode<INode> InsertConstOffsets(LinkedListNode<INode> node, ResourceManager resourceManager, IGpuAccessor gpuAccessor, ShaderStage stage)
         {
             // Non-constant texture offsets are not allowed (according to the spec),

src/Ryujinx.Graphics.Shader/Translation/TranslatorContext.cs

@@ -301,6 +301,11 @@ namespace Ryujinx.Graphics.Shader.Translation
 
                     Optimizer.RunPass(context);
                     TransformPasses.RunPass(context);
+
+                    if (i == 0)
+                    {
+                        FeatureIdentification.RunPass(cfg.Blocks, Definitions.Stage, ref usedFeatures);
+                    }
                 }
 
                 funcs[i] = new Function(cfg.Blocks, $"fun{i}", false, inArgumentsCount, outArgumentsCount);
@@ -353,6 +358,7 @@ namespace Ryujinx.Graphics.Shader.Translation
                 usedFeatures.HasFlag(FeatureFlags.InstanceId),
                 usedFeatures.HasFlag(FeatureFlags.DrawParameters),
                 usedFeatures.HasFlag(FeatureFlags.RtLayer),
+                usedFeatures.HasFlag(FeatureFlags.UnconditionalDiscard),
                 clipDistancesWritten,
                 originalDefinitions.OmapTargets);
 
@@ -393,7 +399,7 @@ namespace Ryujinx.Graphics.Shader.Translation
                 {
                     int binding = resourceManager.Reservations.GetTfeBufferStorageBufferBinding(i);
                     BufferDefinition tfeDataBuffer = new(BufferLayout.Std430, 1, binding, $"tfe_data{i}", tfeDataStruct);
-                    resourceManager.Properties.AddOrUpdateStorageBuffer(tfeDataBuffer);
+                    resourceManager.AddVertexAsComputeStorageBuffer(tfeDataBuffer);
                 }
             }
 
@@ -401,7 +407,7 @@ namespace Ryujinx.Graphics.Shader.Translation
             {
                 int vertexInfoCbBinding = resourceManager.Reservations.VertexInfoConstantBufferBinding;
                 BufferDefinition vertexInfoBuffer = new(BufferLayout.Std140, 0, vertexInfoCbBinding, "vb_info", VertexInfoBuffer.GetStructureType());
-                resourceManager.Properties.AddOrUpdateConstantBuffer(vertexInfoBuffer);
+                resourceManager.AddVertexAsComputeConstantBuffer(vertexInfoBuffer);
 
                 StructureType vertexOutputStruct = new([
                     new StructureField(AggregateType.Array | AggregateType.FP32, "data", 0)
@@ -409,13 +415,13 @@ namespace Ryujinx.Graphics.Shader.Translation
 
                 int vertexOutputSbBinding = resourceManager.Reservations.VertexOutputStorageBufferBinding;
                 BufferDefinition vertexOutputBuffer = new(BufferLayout.Std430, 1, vertexOutputSbBinding, "vertex_output", vertexOutputStruct);
-                resourceManager.Properties.AddOrUpdateStorageBuffer(vertexOutputBuffer);
+                resourceManager.AddVertexAsComputeStorageBuffer(vertexOutputBuffer);
 
                 if (Stage == ShaderStage.Vertex)
                 {
                     SetBindingPair ibSetAndBinding = resourceManager.Reservations.GetIndexBufferTextureSetAndBinding();
                     TextureDefinition indexBuffer = new(ibSetAndBinding.SetIndex, ibSetAndBinding.Binding, "ib_data", SamplerType.TextureBuffer);
-                    resourceManager.Properties.AddOrUpdateTexture(indexBuffer);
+                    resourceManager.AddVertexAsComputeTexture(indexBuffer);
 
                     int inputMap = _program.AttributeUsage.UsedInputAttributes;
 
@@ -424,7 +430,7 @@ namespace Ryujinx.Graphics.Shader.Translation
                         int location = BitOperations.TrailingZeroCount(inputMap);
                         SetBindingPair setAndBinding = resourceManager.Reservations.GetVertexBufferTextureSetAndBinding(location);
                         TextureDefinition vaBuffer = new(setAndBinding.SetIndex, setAndBinding.Binding, $"vb_data{location}", SamplerType.TextureBuffer);
-                        resourceManager.Properties.AddOrUpdateTexture(vaBuffer);
+                        resourceManager.AddVertexAsComputeTexture(vaBuffer);
 
                         inputMap &= ~(1 << location);
                     }
@@ -433,11 +439,11 @@ namespace Ryujinx.Graphics.Shader.Translation
                 {
                     SetBindingPair trbSetAndBinding = resourceManager.Reservations.GetTopologyRemapBufferTextureSetAndBinding();
                     TextureDefinition remapBuffer = new(trbSetAndBinding.SetIndex, trbSetAndBinding.Binding, "trb_data", SamplerType.TextureBuffer);
-                    resourceManager.Properties.AddOrUpdateTexture(remapBuffer);
+                    resourceManager.AddVertexAsComputeTexture(remapBuffer);
 
                     int geometryVbOutputSbBinding = resourceManager.Reservations.GeometryVertexOutputStorageBufferBinding;
                     BufferDefinition geometryVbOutputBuffer = new(BufferLayout.Std430, 1, geometryVbOutputSbBinding, "geometry_vb_output", vertexOutputStruct);
-                    resourceManager.Properties.AddOrUpdateStorageBuffer(geometryVbOutputBuffer);
+                    resourceManager.AddVertexAsComputeStorageBuffer(geometryVbOutputBuffer);
 
                     StructureType geometryIbOutputStruct = new([
                         new StructureField(AggregateType.Array | AggregateType.U32, "data", 0)
@@ -445,7 +451,7 @@ namespace Ryujinx.Graphics.Shader.Translation
 
                     int geometryIbOutputSbBinding = resourceManager.Reservations.GeometryIndexOutputStorageBufferBinding;
                     BufferDefinition geometryIbOutputBuffer = new(BufferLayout.Std430, 1, geometryIbOutputSbBinding, "geometry_ib_output", geometryIbOutputStruct);
-                    resourceManager.Properties.AddOrUpdateStorageBuffer(geometryIbOutputBuffer);
+                    resourceManager.AddVertexAsComputeStorageBuffer(geometryIbOutputBuffer);
                 }
 
                 resourceManager.SetVertexAsComputeLocalMemories(Definitions.Stage, Definitions.InputTopology);
@@ -478,12 +484,17 @@ namespace Ryujinx.Graphics.Shader.Translation
             return new ResourceReservations(GpuAccessor, IsTransformFeedbackEmulated, vertexAsCompute: true, _vertexOutput, ioUsage);
         }
 
+        public ShaderProgramInfo GetVertexAsComputeInfo()
+        {
+            return CreateResourceManager(true).GetVertexAsComputeInfo();
+        }
+
         public void SetVertexOutputMapForGeometryAsCompute(TranslatorContext vertexContext)
         {
             _vertexOutput = vertexContext._program.GetIoUsage();
         }
 
-        public ShaderProgram GenerateVertexPassthroughForCompute()
+        public (ShaderProgram, ShaderProgramInfo) GenerateVertexPassthroughForCompute()
         {
             AttributeUsage attributeUsage = new(GpuAccessor);
             ResourceManager resourceManager = new(ShaderStage.Vertex, GpuAccessor);
@@ -495,7 +506,7 @@ namespace Ryujinx.Graphics.Shader.Translation
             if (Stage == ShaderStage.Vertex)
             {
                 BufferDefinition vertexInfoBuffer = new(BufferLayout.Std140, 0, vertexInfoCbBinding, "vb_info", VertexInfoBuffer.GetStructureType());
-                resourceManager.Properties.AddOrUpdateConstantBuffer(vertexInfoBuffer);
+                resourceManager.AddVertexAsComputeConstantBuffer(vertexInfoBuffer);
             }
 
             StructureType vertexInputStruct = new([
@@ -504,7 +515,7 @@ namespace Ryujinx.Graphics.Shader.Translation
 
             int vertexDataSbBinding = reservations.VertexOutputStorageBufferBinding;
             BufferDefinition vertexOutputBuffer = new(BufferLayout.Std430, 1, vertexDataSbBinding, "vb_input", vertexInputStruct);
-            resourceManager.Properties.AddOrUpdateStorageBuffer(vertexOutputBuffer);
+            resourceManager.AddVertexAsComputeStorageBuffer(vertexOutputBuffer);
 
             EmitterContext context = new();
 
@@ -562,14 +573,14 @@ namespace Ryujinx.Graphics.Shader.Translation
                 LastInVertexPipeline = true
             };
 
-            return Generate(
+            return (Generate(
                 [function],
                 attributeUsage,
                 definitions,
                 definitions,
                 resourceManager,
                 FeatureFlags.None,
-                0);
+                0), resourceManager.GetVertexAsComputeInfo(isVertex: true));
         }
 
         public ShaderProgram GenerateGeometryPassthrough()

src/Ryujinx.Graphics.Vulkan/BackgroundResources.cs

@@ -36,6 +36,7 @@ namespace Ryujinx.Graphics.Vulkan
                         queueLock,
                         _gd.QueueFamilyIndex,
                         _gd.IsQualcommProprietary,
+                        _gd.IsTurnip,
                         isLight: true);
                 }
             }

src/Ryujinx.Graphics.Vulkan/CommandBufferPool.cs

@@ -19,6 +19,7 @@ namespace Ryujinx.Graphics.Vulkan
         private readonly Queue _queue;
         private readonly Lock _queueLock;
         private readonly bool _concurrentFenceWaitUnsupported;
+        private readonly bool _fenceAlwaysWaits;
         private readonly CommandPool _pool;
         private readonly Thread _owner;
 
@@ -66,6 +67,7 @@ namespace Ryujinx.Graphics.Vulkan
             Lock queueLock,
             uint queueFamilyIndex,
             bool concurrentFenceWaitUnsupported,
+            bool fenceAlwaysWaits,
             bool isLight = false)
         {
             _api = api;
@@ -73,6 +75,7 @@ namespace Ryujinx.Graphics.Vulkan
             _queue = queue;
             _queueLock = queueLock;
             _concurrentFenceWaitUnsupported = concurrentFenceWaitUnsupported;
+            _fenceAlwaysWaits = fenceAlwaysWaits;
             _owner = Thread.CurrentThread;
 
             CommandPoolCreateInfo commandPoolCreateInfo = new()
@@ -207,7 +210,7 @@ namespace Ryujinx.Graphics.Vulkan
 
                 ref ReservedCommandBuffer entry = ref _commandBuffers[index];
 
-                if (wait || !entry.InConsumption || entry.Fence.IsSignaled())
+                if (wait || !entry.InConsumption || entry.Fence.IsSignaledLazy())
                 {
                     WaitAndDecrementRef(index);
 
@@ -349,7 +352,7 @@ namespace Ryujinx.Graphics.Vulkan
 
             if (refreshFence)
             {
-                entry.Fence = new FenceHolder(_api, _device, _concurrentFenceWaitUnsupported);
+                entry.Fence = new FenceHolder(_api, _device, _concurrentFenceWaitUnsupported, _fenceAlwaysWaits);
             }
             else
             {

src/Ryujinx.Graphics.Vulkan/FenceHolder.cs

@@ -12,13 +12,15 @@ namespace Ryujinx.Graphics.Vulkan
         private int _referenceCount;
         private int _lock;
         private readonly bool _concurrentWaitUnsupported;
+        private readonly bool _alwaysWaits;
         private bool _disposed;
 
-        public unsafe FenceHolder(Vk api, Device device, bool concurrentWaitUnsupported)
+        public unsafe FenceHolder(Vk api, Device device, bool concurrentWaitUnsupported, bool alwaysWaits)
         {
             _api = api;
             _device = device;
             _concurrentWaitUnsupported = concurrentWaitUnsupported;
+            _alwaysWaits = alwaysWaits;
 
             FenceCreateInfo fenceCreateInfo = new()
             {
@@ -123,6 +125,16 @@ namespace Ryujinx.Graphics.Vulkan
             }
         }
 
+        public bool IsSignaledLazy()
+        {
+            if (_alwaysWaits)
+            {
+                return false;
+            }
+
+            return IsSignaled();
+        }
+
         public bool IsSignaled()
         {
             if (_concurrentWaitUnsupported)

src/Ryujinx.Graphics.Vulkan/Helpers.cs

@@ -0,0 +1,32 @@
+using Silk.NET.Vulkan;
+using System.Runtime.CompilerServices;
+
+namespace Ryujinx.Graphics.Vulkan
+{
+    public static class Helpers
+    {
+        extension(Vk api)
+        {
+            /// <summary>
+            ///     C# implementation of the RENDERDOC_DEVICEPOINTER_FROM_VKINSTANCE macro from the RenderDoc API header, since we cannot use macros from C#.
+            /// </summary>
+            /// <returns>The dispatch table pointer, which sits as the first pointer-sized object in the memory pointed to by the <see cref="Vk"/>'s <see cref="Instance"/> pointer.</returns>
+            [MethodImpl(MethodImplOptions.AggressiveInlining)]
+            public unsafe void* GetRenderDocDevicePointer() =>
+                api.CurrentInstance is not null
+                    ? api.CurrentInstance.Value.GetRenderDocDevicePointer()
+                    : null;
+        }
+
+        extension(Instance instance)
+        {
+            /// <summary>
+            ///     C# implementation of the RENDERDOC_DEVICEPOINTER_FROM_VKINSTANCE macro from the RenderDoc API header, since we cannot use macros from C#.
+            /// </summary>
+            /// <returns>The dispatch table pointer, which sits as the first pointer-sized object in the memory pointed to by the <see cref="Instance"/>'s pointer.</returns>
+            [MethodImpl(MethodImplOptions.AggressiveInlining)]
+            public unsafe void* GetRenderDocDevicePointer()
+                => (*((void**)(instance.Handle)));
+        }
+    }
+}

src/Ryujinx.Graphics.Vulkan/ShaderCollection.cs

@@ -60,7 +60,7 @@ namespace Ryujinx.Graphics.Vulkan
 
         private ProgramPipelineState _state;
         private DisposableRenderPass _dummyRenderPass;
-        private readonly Task _compileTask;
+        private ShaderCompilationRequest _compileRequest;
         private bool _firstBackgroundUse;
 
         public ShaderCollection(
@@ -139,7 +139,7 @@ namespace Ryujinx.Graphics.Vulkan
             // Updating buffer texture bindings using template updates crashes the Adreno driver on Windows.
             UpdateTexturesWithoutTemplate = gd.IsQualcommProprietary && usesBufferTextures;
 
-            _compileTask = Task.CompletedTask;
+            _compileRequest = new ShaderCompilationRequest(Task.CompletedTask);
             _firstBackgroundUse = false;
         }
 
@@ -153,7 +153,9 @@ namespace Ryujinx.Graphics.Vulkan
         {
             _state = state;
 
-            _compileTask = BackgroundCompilation();
+            _compileRequest = gd.ShaderCompilationQueue != null
+                ? gd.ShaderCompilationQueue.Add(BackgroundCompilation)
+                : new ShaderCompilationRequest(BackgroundCompilationAsync());
             _firstBackgroundUse = !fromCache;
         }
 
@@ -458,10 +460,25 @@ namespace Ryujinx.Graphics.Vulkan
             return (buffer, texture);
         }
 
-        private async Task BackgroundCompilation()
+        private async Task BackgroundCompilationAsync()
         {
             await Task.WhenAll(_shaders.Select(shader => shader.CompileTask));
 
+            BackgroundCompilationImpl();
+        }
+
+        private void BackgroundCompilation()
+        {
+            foreach (var shader in _shaders)
+            {
+                shader.CompileTask.Wait();
+            }
+
+            BackgroundCompilationImpl();
+        }
+
+        private void BackgroundCompilationImpl()
+        {
             if (Array.Exists(_shaders, shader => shader.CompileStatus == ProgramLinkStatus.Failure))
             {
                 LinkStatus = ProgramLinkStatus.Failure;
@@ -604,11 +621,11 @@ namespace Ryujinx.Graphics.Vulkan
                     }
                 }
 
-                if (!_compileTask.IsCompleted)
+                if (!_compileRequest.IsCompleted)
                 {
                     if (blocking)
                     {
-                        _compileTask.Wait();
+                        _compileRequest.Wait();
 
                         if (LinkStatus == ProgramLinkStatus.Failure)
                         {

src/Ryujinx.Graphics.Vulkan/ShaderCompilationQueue.cs

@@ -0,0 +1,131 @@
+using System;
+using System.Collections.Concurrent;
+using System.Threading;
+
+namespace Ryujinx.Graphics.Vulkan
+{
+    class ShaderCompilationQueue
+    {
+        private const int MaxParallelCompilations = 8;
+        private const int MaxThreadStackSize = 2 * 1024 * 1024; // MB
+
+        private struct Request
+        {
+            public readonly ulong Id;
+            public readonly Action Callback;
+
+            public Request(ulong id, Action callback)
+            {
+                Id = id;
+                Callback = callback;
+            }
+        }
+
+        private readonly Thread[] _workerThreads;
+        private readonly CancellationTokenSource _cts;
+        private readonly BlockingCollection<Request>[] _queues;
+        private readonly ulong[] _finishedIds;
+        private ulong _currentId;
+        private int _currentQueueIndex;
+
+        public ShaderCompilationQueue()
+        {
+            _workerThreads = new Thread[MaxParallelCompilations];
+            _queues = new BlockingCollection<Request>[MaxParallelCompilations];
+            _finishedIds = new ulong[MaxParallelCompilations];
+
+            _cts = new CancellationTokenSource();
+
+            for (int i = 0; i < MaxParallelCompilations; i++)
+            {
+                _queues[i] = new BlockingCollection<Request>();
+
+                Thread thread = new Thread(DoWork, MaxThreadStackSize) { Name = $"BackgroundShaderCompiler.{i}" };
+                thread.IsBackground = true;
+                thread.Start(i);
+
+                _workerThreads[i] = thread;
+            }
+        }
+
+        private void DoWork(object threadId)
+        {
+            int queueIndex = (int)threadId;
+
+            try
+            {
+                var queue = _queues[queueIndex];
+
+                foreach (var request in queue.GetConsumingEnumerable(_cts.Token))
+                {
+                    request.Callback();
+
+                    lock (queue)
+                    {
+                        _finishedIds[queueIndex] = request.Id;
+
+                        Monitor.PulseAll(queue);
+                    }
+                }
+            }
+            catch (OperationCanceledException)
+            {
+            }
+        }
+
+        public ShaderCompilationRequest Add(Action callback)
+        {
+            ulong newId = Interlocked.Increment(ref _currentId);
+
+            // Let's keep rotating between the queues to increase the chances
+            // that the selected queue thread is currently idle.
+            int queueIndex = Interlocked.Increment(ref _currentQueueIndex) % MaxParallelCompilations;
+
+            _queues[queueIndex].Add(new Request(newId, callback));
+
+            return new ShaderCompilationRequest(this, queueIndex, newId);
+        }
+
+        public void Wait(int queueIndex, ulong id)
+        {
+            var queue = _queues[queueIndex];
+
+            lock (queue)
+            {
+                while (_finishedIds[queueIndex] < id)
+                {
+                    Monitor.Wait(queue);
+                }
+            }
+        }
+
+        public bool IsCompleted(int queueIndex, ulong id)
+        {
+            var queue = _queues[queueIndex];
+
+            lock (queue)
+            {
+                return _finishedIds[queueIndex] >= id;
+            }
+        }
+
+        public void Dispose()
+        {
+            for (int i = 0; i < MaxParallelCompilations; i++)
+            {
+                _queues[i].CompleteAdding();
+            }
+
+            _cts.Cancel();
+
+            for (int i = 0; i < MaxParallelCompilations; i++)
+            {
+                _workerThreads[i].Join();
+
+                _queues[i].Dispose();
+            }
+
+            _cts.Dispose();
+        }
+    }
+}

src/Ryujinx.Graphics.Vulkan/ShaderCompilationRequest.cs

@@ -0,0 +1,55 @@
+using System.Threading.Tasks;
+
+namespace Ryujinx.Graphics.Vulkan
+{
+    struct ShaderCompilationRequest
+    {
+        private readonly Task _task;
+        private readonly ShaderCompilationQueue _queue;
+        private readonly int _queueIndex;
+        private readonly ulong _requestId;
+
+        public bool IsCompleted
+        {
+            get
+            {
+                if (_task != null)
+                {
+                    return _task.IsCompleted;
+                }
+                else
+                {
+                    return _queue.IsCompleted(_queueIndex, _requestId);
+                }
+            }
+        }
+
+        public ShaderCompilationRequest(Task task)
+        {
+            _task = task;
+            _queue = null;
+            _queueIndex = 0;
+            _requestId = 0;
+        }
+
+        public ShaderCompilationRequest(ShaderCompilationQueue queue, int queueIndex, ulong requestId)
+        {
+            _task = null;
+            _queue = queue;
+            _queueIndex = queueIndex;
+            _requestId = requestId;
+        }
+
+        public void Wait()
+        {
+            if (_task != null)
+            {
+                _task.Wait();
+            }
+            else
+            {
+                _queue.Wait(_queueIndex, _requestId);
+            }
+        }
+    }
+}

src/Ryujinx.Graphics.Vulkan/StagingBuffer.cs

@@ -266,7 +266,7 @@ namespace Ryujinx.Graphics.Vulkan
         public void FreeCompleted()
         {
             FenceHolder signalledFence = null;
-            while (_pendingCopies.TryPeek(out PendingCopy pc) && pc.Fence != null && (pc.Fence == signalledFence || pc.Fence.IsSignaled()))
+            while (_pendingCopies.TryPeek(out PendingCopy pc) && pc.Fence != null && (pc.Fence == signalledFence || pc.Fence.IsSignaledLazy()))
             {
                 signalledFence = pc.Fence; // Already checked - don't need to do it again.
                 PendingCopy dequeued = _pendingCopies.Dequeue();

src/Ryujinx.Graphics.Vulkan/VulkanRenderer.cs

@@ -55,6 +55,7 @@ namespace Ryujinx.Graphics.Vulkan
         internal CommandBufferPool CommandBufferPool { get; private set; }
         internal PipelineLayoutCache PipelineLayoutCache { get; private set; }
         internal BackgroundResources BackgroundResources { get; private set; }
+        internal ShaderCompilationQueue ShaderCompilationQueue { get; private set; }
         internal Action<Action> InterruptAction { get; private set; }
         internal SyncManager SyncManager { get; private set; }
 
@@ -96,6 +97,7 @@ namespace Ryujinx.Graphics.Vulkan
         internal bool IsNvidiaPreTuring { get; private set; }
         internal bool IsIntelArc { get; private set; }
         internal bool IsQualcommProprietary { get; private set; }
+        internal bool IsTurnip { get; private set; }
         internal bool IsMoltenVk { get; private set; }
         internal bool SupportsMTL31 { get; private set; }
         internal bool IsTBDR { get; private set; }
@@ -127,6 +129,12 @@ namespace Ryujinx.Graphics.Vulkan
 
                 // Any device running on MacOS is using MoltenVK, even Intel and AMD vendors.
                 IsMoltenVk = true;
+
+                // The default thread stack size on MacOS is low, and can cause stack overflow
+                // on SPIR-V Cross during shader compilation.
+                // As a workaround, we use this custom queue which allows us to specify the stack
+                // size of the threads used for compilation.
+                ShaderCompilationQueue = new ShaderCompilationQueue();
             }
 
             SupportsMTL31 = OperatingSystem.IsMacOSVersionAtLeast(14);
@@ -396,6 +404,8 @@ namespace Ryujinx.Graphics.Vulkan
 
             IsFeedbackLoopDevice = IsAmdRdna3;
 
+            IsTurnip = GpuRenderer.StartsWith("Turnip");
+
             if (Vendor == Vendor.Nvidia)
             {
                 Match match = VendorUtils.NvidiaConsumerClassRegex().Match(GpuRenderer);
@@ -476,7 +486,7 @@ namespace Ryujinx.Graphics.Vulkan
             Api.TryGetDeviceExtension(_instance.Instance, _device, out ExtExternalMemoryHost hostMemoryApi);
             HostMemoryAllocator = new HostMemoryAllocator(MemoryAllocator, Api, hostMemoryApi, _device);
 
-            CommandBufferPool = new CommandBufferPool(Api, _device, Queue, QueueLock, queueFamilyIndex, IsQualcommProprietary);
+            CommandBufferPool = new CommandBufferPool(Api, _device, Queue, QueueLock, queueFamilyIndex, IsQualcommProprietary, IsTurnip);
 
             PipelineLayoutCache = new PipelineLayoutCache();
 
@@ -777,6 +787,7 @@ namespace Ryujinx.Graphics.Vulkan
                 supportsGeometryShader: Capabilities.SupportsGeometryShader,
                 supportsGeometryShaderPassthrough: Capabilities.SupportsGeometryShaderPassthrough,
                 supportsTransformFeedback: Capabilities.SupportsTransformFeedback,
+                supportsImageBufferPixelAlignment: false,
                 supportsImageLoadFormatted: features2.Features.ShaderStorageImageReadWithoutFormat,
                 supportsLayerVertexTessellation: featuresVk12.ShaderOutputLayer,
                 supportsMismatchingViewFormat: true,
@@ -790,6 +801,7 @@ namespace Ryujinx.Graphics.Vulkan
                 supportsShaderNonUniformIndexing:
                     featuresVk12.ShaderSampledImageArrayNonUniformIndexing &&
                     featuresVk12.ShaderStorageImageArrayNonUniformIndexing,
+                supportsTextureBufferPixelAlignment: false,
                 supportsTextureGatherOffsets: features2.Features.ShaderImageGatherExtended,
                 supportsTextureShadowLod: false,
                 supportsVertexStoreAndAtomics: features2.Features.VertexPipelineStoresAndAtomics,
@@ -1112,6 +1124,8 @@ namespace Ryujinx.Graphics.Vulkan
 
             SurfaceApi.DestroySurface(_instance.Instance, _surface, null);
 
+            ShaderCompilationQueue?.Dispose();
+
             Api.DestroyDevice(_device, null);
 
             _debugMessenger.Dispose();

src/Ryujinx.HLE/HOS/Applets/Error/ErrorApplet.cs

@@ -166,13 +166,15 @@ namespace Ryujinx.HLE.HOS.Applets.Error
 
             string[] buttons = GetButtonsText(module, description, "DlgBtn");
 
-            bool showDetails = _horizon.Device.UIHandler.DisplayErrorAppletDialog($"Error Code: {module}-{description:0000}", "\n" + message, buttons);
+            (uint Module, uint Description) errorCodeTuple = (module, uint.Parse(description.ToString("0000")));
+
+            bool showDetails = _horizon.Device.UIHandler.DisplayErrorAppletDialog($"Error Code: {module}-{description:0000}", "\n" + message, buttons, errorCodeTuple);
             if (showDetails)
             {
                 message = GetMessageText(module, description, "FlvMsg");
                 buttons = GetButtonsText(module, description, "FlvBtn");
 
-                _horizon.Device.UIHandler.DisplayErrorAppletDialog($"Details: {module}-{description:0000}", "\n" + message, buttons);
+                _horizon.Device.UIHandler.DisplayErrorAppletDialog($"Details: {module}-{description:0000}", "\n" + message, buttons, errorCodeTuple);
             }
         }
 

src/Ryujinx.HLE/HOS/Applets/PlayerSelect/PlayerSelectApplet.cs

@@ -27,9 +27,19 @@ namespace Ryujinx.HLE.HOS.Applets
             _normalSession = normalSession;
             _interactiveSession = interactiveSession;
 
-            // TODO(jduncanator): Parse PlayerSelectConfig from input data
-            _normalSession.Push(BuildResponse());
-
+            UserProfile selected = _system.Device.UIHandler.ShowPlayerSelectDialog();
+            if (selected == null)
+            {
+                _normalSession.Push(BuildResponse());
+            }
+            else if (selected.UserId == new UserId("00000000000000000000000000000080"))
+            {
+                _normalSession.Push(BuildGuestResponse());
+            }
+            else
+            {
+                _normalSession.Push(BuildResponse(selected));
+            }
             AppletStateChanged?.Invoke(this, null);
 
             _system.ReturnFocus();
@@ -37,16 +47,34 @@ namespace Ryujinx.HLE.HOS.Applets
             return ResultCode.Success;
         }
 
-        private byte[] BuildResponse()
+        private byte[] BuildResponse(UserProfile selectedUser)
         {
-            UserProfile currentUser = _system.AccountManager.LastOpenedUser;
-
             using MemoryStream stream = MemoryStreamManager.Shared.GetStream();
             using BinaryWriter writer = new(stream);
 
             writer.Write((ulong)PlayerSelectResult.Success);
 
-            currentUser.UserId.Write(writer);
+            selectedUser.UserId.Write(writer);
+
+            return stream.ToArray();
+        }
+
+        private byte[] BuildGuestResponse()
+        {
+            using MemoryStream stream = MemoryStreamManager.Shared.GetStream();
+            using BinaryWriter writer = new(stream);
+
+            writer.Write(new byte());
+
+            return stream.ToArray();
+        }
+        
+        private byte[] BuildResponse()
+        {
+            using MemoryStream stream = MemoryStreamManager.Shared.GetStream();
+            using BinaryWriter writer = new(stream);
+
+            writer.Write((ulong)PlayerSelectResult.Failure);
 
             return stream.ToArray();
         }

src/Ryujinx.HLE/Ryujinx.HLE.csproj

@@ -59,6 +59,7 @@
     <EmbeddedResource Include="HOS\Applets\SoftwareKeyboard\Resources\Icon_BtnB.png" />
     <EmbeddedResource Include="HOS\Applets\SoftwareKeyboard\Resources\Icon_KeyF6.png" />
     <EmbeddedResource Include="HOS\Services\Account\Acc\DefaultUserImage.jpg" />
+    <EmbeddedResource Include="HOS\Services\Account\Acc\GuestUserImage.jpg" />
   </ItemGroup>
 
 </Project>

src/Ryujinx.HLE/UI/IHostUIHandler.cs

@@ -1,4 +1,5 @@
 using Ryujinx.HLE.HOS.Applets;
+using Ryujinx.HLE.HOS.Services.Account.Acc;
 using Ryujinx.HLE.HOS.Services.Am.AppletOE.ApplicationProxyService.ApplicationProxy.Types;
 
 namespace Ryujinx.HLE.UI
@@ -48,7 +49,8 @@ namespace Ryujinx.HLE.UI
         /// Displays a Message Dialog box specific to Error Applet and blocks until it is closed.
         /// </summary>
         /// <returns>False when OK is pressed, True when another button (Details) is pressed.</returns>
-        bool DisplayErrorAppletDialog(string title, string message, string[] buttonsText);
+        // ReSharper disable once UnusedParameter.Global
+        bool DisplayErrorAppletDialog(string title, string message, string[] buttonsText, (uint Module, uint Description)? errorCode = null);
 
         /// <summary>
         /// Creates a handler to process keyboard inputs into text strings.
@@ -65,5 +67,10 @@ namespace Ryujinx.HLE.UI
         /// Takes a screenshot from the current renderer and saves it in the screenshots folder.
         /// </summary>
         void TakeScreenshot();
+
+        /// <summary>
+        /// Displays the player select dialog and returns the selected profile.
+        /// </summary>
+        UserProfile ShowPlayerSelectDialog();
     }
 }

src/Ryujinx.ShaderTools/Program.cs

@@ -116,7 +116,7 @@ namespace Ryujinx.ShaderTools
 
             if (options.VertexPassthrough)
             {
-                program = translatorContext.GenerateVertexPassthroughForCompute();
+                (program, _) = translatorContext.GenerateVertexPassthroughForCompute();
             }
             else
             {

src/Ryujinx.UI.Common/Helper/CommandLineState.cs

@@ -15,6 +15,9 @@ namespace Ryujinx.UI.Common.Helper
         public static string OverrideBackendThreading { get; private set; }
         public static string OverrideHideCursor { get; private set; }
         public static string BaseDirPathArg { get; private set; }
+
+        public static string RenderDocCaptureTitleFormat { get; private set; } =
+            "{EmuVersion}\n{GuestName} {GuestVersion} {GuestTitleId} {GuestArch}";
         public static FilePath FirmwareToInstallPathArg { get; set; }
         public static string Profile { get; private set; }
         public static string LaunchPathArg { get; private set; }
@@ -45,6 +48,20 @@ namespace Ryujinx.UI.Common.Helper
 
                         BaseDirPathArg = args[++i];
 
+                        arguments.Add(arg);
+                        arguments.Add(args[i]);
+                        break;
+                    case "-rdct":
+                    case "--rd-capture-title-format":
+                        if (i + 1 >= args.Length)
+                        {
+                            Logger.Error?.Print(LogClass.Application, $"Invalid option '{arg}'");
+
+                            continue;
+                        }
+
+                        RenderDocCaptureTitleFormat = args[++i];
+
                         arguments.Add(arg);
                         arguments.Add(args[i]);
                         break;

src/Ryujinx.UI.Common/Helper/TitleHelper.cs

@@ -1,3 +1,4 @@
+using Gommon;
 using Ryujinx.HLE.Loaders.Processes;
 using System;
 
@@ -26,5 +27,23 @@ namespace Ryujinx.UI.Common.Helper
 
             return appTitle;
         }
+
+        public static string FormatRenderDocCaptureTitle(ProcessResult activeProcess, string applicationVersion)
+        {
+            if (activeProcess == null)
+                return string.Empty;
+
+            string titleNameSection = string.IsNullOrWhiteSpace(activeProcess.Name) ? string.Empty : activeProcess.Name;
+            string titleVersionSection = string.IsNullOrWhiteSpace(activeProcess.DisplayVersion) ? string.Empty : $"v{activeProcess.DisplayVersion}";
+            string titleIdSection = $"({activeProcess.ProgramIdText.ToUpper()})";
+            string titleArchSection = activeProcess.Is64Bit ? "(64-bit)" : "(32-bit)";
+
+            return CommandLineState.RenderDocCaptureTitleFormat
+                .ReplaceIgnoreCase("{EmuVersion}", applicationVersion)
+                .ReplaceIgnoreCase("{GuestName}", titleNameSection)
+                .ReplaceIgnoreCase("{GuestVersion}", titleVersionSection)
+                .ReplaceIgnoreCase("{GuestTitleId}", titleIdSection)
+                .ReplaceIgnoreCase("{GuestArch}", titleArchSection);
+        }
     }
 }

src/Ryujinx/App.axaml.cs

@@ -14,6 +14,7 @@ using Ryujinx.Ava.UI.Helpers;
 using Ryujinx.Ava.UI.Windows;
 using Ryujinx.Common;
 using Ryujinx.Common.Logging;
+using Ryujinx.Graphics.RenderDocApi;
 using Ryujinx.UI.Common.Configuration;
 using Ryujinx.UI.Common.Helper;
 using System;

src/Ryujinx/Assets/Locales/en_US.json

@@ -934,5 +934,9 @@
   "GameListContextMenuExtractDataAocRomFSToolTip": "Extract the RomFS from a selected DLC file",
   "ExtractAocListHeader": "Select a DLC to Extract",
   "SettingsTabSystemSkipUserProfilesManager": "Skip Dialog 'Manage User Profiles'",
-  "SkipUserProfilesTooltip": "This option skips the 'Manage User Profiles' dialog during gameplay, using a pre-selected profile.\n\nProfile switching is found in 'Settings' - 'Manager User Profiles'. Select the desired profile before loading the game."
+  "SkipUserProfilesTooltip": "This option skips the 'Manage User Profiles' dialog during gameplay, using a pre-selected profile.\n\nProfile switching is found in 'Settings' - 'Manager User Profiles'. Select the desired profile before loading the game.",
+  "MenuBarActions_StartCapture": "Start RenderDoc Frame Capture",
+  "MenuBarActions_EndCapture": "End RenderDoc Frame Capture",
+  "MenuBarActions_DiscardCapture": "Discard RenderDoc Frame Capture",
+  "MenuBarActions_DiscardCapture_ToolTip": "Ends the currently active RenderDoc Frame Capture, immediately discarding its result."
 }

src/Ryujinx/Assets/Locales/es_ES.json

@@ -807,5 +807,9 @@
   "MultiplayerModeDisabled": "Deshabilitar",
   "MultiplayerModeLdnMitm": "ldn_mitm",
   "SettingsTabSystemSkipUserProfilesManager": "Omitir el Diálogo 'Gestionar Perfiles de Usuario'",
-  "SkipUserProfilesTooltip": "Esta opción omite el diálogo de 'Gestionar perfiles de usuario' durante el juego, utilizando un perfil preseleccionado.\n\nEl cambio de perfil se encuentra en 'Configuración' - 'Gestionar perfiles de usuario'. Seleccione el perfil deseado antes de cargar el juego."
+  "SkipUserProfilesTooltip": "Esta opción omite el diálogo de 'Gestionar perfiles de usuario' durante el juego, utilizando un perfil preseleccionado.\n\nEl cambio de perfil se encuentra en 'Configuración' - 'Gestionar perfiles de usuario'. Seleccione el perfil deseado antes de cargar el juego.",
+  "MenuBarActions_StartCapture": "Iniciar una captura de fotograma de RenderDoc",
+  "MenuBarActions_EndCapture": "Detener la captura de fotograma de RenderDoc",
+  "MenuBarActions_DiscardCapture": "Descartar la captura de fotograma de RenderDoc",
+  "MenuBarActions_DiscardCapture_ToolTip": "Finaliza la captura de fotograma de RenderDoc actualmente activa y descarta inmediatamente su resultado."
 }

src/Ryujinx/Assets/Locales/fr_FR.json

@@ -828,5 +828,9 @@
   "GameListContextMenuExtractDataAocRomFSToolTip": "Extraire les RomFS d'un fichier DLC choisi",
   "ExtractAocListHeader": "Choisissez un DLC à extraire",
   "SettingsTabSystemSkipUserProfilesManager": "Ignorer la Boîte de Dialogue « Gérer les Profils d'Utilisateurs »",
-  "SkipUserProfilesTooltip": "Cette option permet d'éviter le dialogue du 'Gérer les profils d'utilisateurs' pendant le jeu, en utilisant un profil pré-sélectionné.\n\nLa sélection du profil se trouve dans 'Paramètres' - 'Gérer les profils d'utilisateurs'. Sélectionnez le profil souhaité avant de charger la partie."
+  "SkipUserProfilesTooltip": "Cette option permet d'éviter le dialogue du 'Gérer les profils d'utilisateurs' pendant le jeu, en utilisant un profil pré-sélectionné.\n\nLa sélection du profil se trouve dans 'Paramètres' - 'Gérer les profils d'utilisateurs'. Sélectionnez le profil souhaité avant de charger la partie.",
+  "MenuBarActions_StartCapture": "Démarrer une capture de trame RenderDoc",
+  "MenuBarActions_EndCapture": "Arrêter la capture de trame RenderDoc",
+  "MenuBarActions_DiscardCapture": "Supprimer la capture de trame RenderDoc",
+  "MenuBarActions_DiscardCapture_ToolTip": "Met fin à la capture de trame RenderDoc en cours, en supprimant immédiatement son résultat."
 }

src/Ryujinx/Headless/WindowBase.cs

@@ -10,6 +10,7 @@ using Ryujinx.Graphics.GAL.Multithreading;
 using Ryujinx.Graphics.Gpu;
 using Ryujinx.Graphics.OpenGL;
 using Ryujinx.HLE.HOS.Applets;
+using Ryujinx.HLE.HOS.Services.Account.Acc;
 using Ryujinx.HLE.HOS.Services.Am.AppletOE.ApplicationProxyService.ApplicationProxy.Types;
 using Ryujinx.HLE.Loaders.Processes;
 using Ryujinx.HLE.UI;
@@ -28,6 +29,7 @@ using static SDL.SDL3;
 using AntiAliasing = Ryujinx.Common.Configuration.AntiAliasing;
 using ScalingFilter = Ryujinx.Common.Configuration.ScalingFilter;
 using Switch = Ryujinx.HLE.Switch;
+using UserProfile = Ryujinx.HLE.HOS.Services.Account.Acc.UserProfile;
 
 namespace Ryujinx.Headless
 {
@@ -531,7 +533,7 @@ namespace Ryujinx.Headless
             Exit();
         }
 
-        public bool DisplayErrorAppletDialog(string title, string message, string[] buttonsText)
+        public bool DisplayErrorAppletDialog(string title, string message, string[] buttonsText, (uint Module, uint Description)? errorCode = null)
         {
             SDL_MessageBoxButtonData[] buttons = new SDL_MessageBoxButtonData[buttonsText.Length];
 
@@ -590,5 +592,10 @@ namespace Ryujinx.Headless
         {
             throw new NotImplementedException();
         }
+        
+        public UserProfile ShowPlayerSelectDialog()
+        {
+            return AccountSaveDataManager.GetLastUsedUser();
+        }
     }
 }

src/Ryujinx/Program.cs

@@ -8,6 +8,7 @@ using Ryujinx.Common.GraphicsDriver;
 using Ryujinx.Common.Logging;
 using Ryujinx.Common.SystemInterop;
 using Ryujinx.Common.Utilities;
+using Ryujinx.Graphics.RenderDocApi;
 using Ryujinx.Graphics.Vulkan.MoltenVK;
 using Ryujinx.Headless;
 using Ryujinx.Modules;

src/Ryujinx/Ryujinx.csproj

@@ -56,6 +56,7 @@
   </ItemGroup>
 
   <ItemGroup>
+    <ProjectReference Include="..\Ryujinx.Graphics.RenderDocApi\Ryujinx.Graphics.RenderDocApi.csproj" />
     <ProjectReference Include="..\Ryujinx.Audio.Backends.SDL3\Ryujinx.Audio.Backends.SDL3.csproj" />
     <ProjectReference Include="..\Ryujinx.Graphics.Vulkan\Ryujinx.Graphics.Vulkan.csproj" />
     <ProjectReference Include="..\Ryujinx.Graphics.OpenGL\Ryujinx.Graphics.OpenGL.csproj" />
@@ -168,4 +169,10 @@
   <ItemGroup>
     <TrimmerRootDescriptor Include="TrimmerRootDescriptor.xml" />
   </ItemGroup>
+  <ItemGroup>
+    <Compile Update="UI\Applet\UserSelectorDialog.axaml.cs">
+      <DependentUpon>UserSelectorDialog.axaml</DependentUpon>
+      <SubType>Code</SubType>
+    </Compile>
+  </ItemGroup>
 </Project>

src/Ryujinx/UI/Applet/AvaHostUIHandler.cs

@@ -1,17 +1,23 @@
 using Avalonia.Controls;
 using Avalonia.Threading;
 using FluentAvalonia.UI.Controls;
+using Gommon;
 using Ryujinx.Ava.Common.Locale;
 using Ryujinx.Ava.UI.Controls;
 using Ryujinx.Ava.UI.Helpers;
+using Ryujinx.Ava.UI.ViewModels;
 using Ryujinx.Ava.UI.Windows;
+using Ryujinx.Common;
 using Ryujinx.HLE;
 using Ryujinx.HLE.HOS.Applets;
 using Ryujinx.HLE.HOS.Applets.SoftwareKeyboard;
+using Ryujinx.HLE.HOS.Services.Account.Acc;
 using Ryujinx.HLE.HOS.Services.Am.AppletOE.ApplicationProxyService.ApplicationProxy.Types;
 using Ryujinx.HLE.UI;
 using Ryujinx.UI.Common.Configuration;
 using System;
+using System.Collections.ObjectModel;
+using System.Linq;
 using System.Threading;
 
 namespace Ryujinx.Ava.UI.Applet
@@ -215,7 +221,7 @@ namespace Ryujinx.Ava.UI.Applet
             _parent.ViewModel.AppHost?.Stop();
         }
 
-        public bool DisplayErrorAppletDialog(string title, string message, string[] buttons)
+        public bool DisplayErrorAppletDialog(string title, string message, string[] buttons, (uint Module, uint Description)? errorCode = null)
         {
             ManualResetEvent dialogCloseEvent = new(false);
 
@@ -256,11 +262,61 @@ namespace Ryujinx.Ava.UI.Applet
             return showDetails;
         }
 
-        public IDynamicTextInputHandler CreateDynamicTextInputHandler()
-        {
-            return new AvaloniaDynamicTextInputHandler(_parent);
-        }
+        public IDynamicTextInputHandler CreateDynamicTextInputHandler() => new AvaloniaDynamicTextInputHandler(_parent);
         
+        public UserProfile ShowPlayerSelectDialog()
+        {
+            UserId selected = UserId.Null;
+            byte[] defaultGuestImage = EmbeddedResources.Read("Ryujinx.HLE/HOS/Services/Account/Acc/GuestUserImage.jpg");
+            UserProfile guest = new UserProfile(new UserId("00000000000000000000000000000080"), "Guest", defaultGuestImage);
+    
+            ManualResetEvent dialogCloseEvent = new(false);
+    
+            Dispatcher.UIThread.InvokeAsync(async () =>
+            {
+                ObservableCollection<BaseModel> profiles = [];
+                NavigationDialogHost nav = new();
+                
+                _parent.AccountManager.GetAllUsers()
+                    .OrderBy(x => x.Name)
+                    .ForEach(profile => profiles.Add(new Models.UserProfile(profile, nav)));
+                
+                profiles.Add(new Models.UserProfile(guest, nav));
+                ProfileSelectorDialogViewModel viewModel = new()
+                {
+                    Profiles = profiles,
+                    SelectedUserId = _parent.AccountManager.LastOpenedUser.UserId
+                };
+                (selected, _) = await ProfileSelectorDialog.ShowInputDialog(viewModel);
+        
+                dialogCloseEvent.Set();
+            });
+    
+            dialogCloseEvent.WaitOne();
+            
+            UserProfile profile = _parent.AccountManager.LastOpenedUser;
+            if (selected == guest.UserId)
+            {
+                profile = guest;
+            }
+            else if (selected == UserId.Null)
+            {
+                profile = null;
+            }
+            else
+            {
+                foreach (UserProfile p in _parent.AccountManager.GetAllUsers())
+                {
+                    if (p.UserId == selected)
+                    {
+                        profile = p;
+                        break;
+                    }
+                }
+            }
+            return profile;
+        }
+
         public void TakeScreenshot()
         {
             _parent.ViewModel.AppHost.ScreenshotRequested = true;

src/Ryujinx/UI/Applet/ProfileSelectorDialog.axaml

@@ -0,0 +1,121 @@
+<UserControl
+    x:Class="Ryujinx.Ava.UI.Applet.ProfileSelectorDialog"
+    xmlns="https://github.com/avaloniaui"
+    xmlns:x="http://schemas.microsoft.com/winfx/2006/xaml"
+    xmlns:d="http://schemas.microsoft.com/expression/blend/2008"
+    xmlns:mc="http://schemas.openxmlformats.org/markup-compatibility/2006"
+    xmlns:helpers="clr-namespace:Ryujinx.Ava.UI.Helpers"
+    xmlns:models="clr-namespace:Ryujinx.Ava.UI.Models"
+    xmlns:viewModels="clr-namespace:Ryujinx.Ava.UI.ViewModels"
+    d:DesignHeight="450"
+    MinWidth="500"
+    d:DesignWidth="800"
+    mc:Ignorable="d"
+    Focusable="True"
+    x:DataType="viewModels:ProfileSelectorDialogViewModel">
+    
+    <UserControl.Resources>
+        <helpers:BitmapArrayValueConverter x:Key="ByteImage" />
+    </UserControl.Resources>
+    
+    <Design.DataContext>
+        <viewModels:ProfileSelectorDialogViewModel />
+    </Design.DataContext>
+    
+    <Grid HorizontalAlignment="Stretch" VerticalAlignment="Stretch">
+        <Grid.RowDefinitions>
+            <RowDefinition />
+            <RowDefinition Height="Auto" />
+        </Grid.RowDefinitions>
+        
+        <Border
+            CornerRadius="5"
+            BorderBrush="{DynamicResource AppListHoverBackgroundColor}"
+            BorderThickness="1">
+            
+            <ListBox
+                MaxHeight="300"
+                HorizontalAlignment="Stretch"
+                VerticalAlignment="Center"
+                Background="Transparent"
+                ItemsSource="{Binding Profiles}"
+                SelectionChanged="ProfilesList_SelectionChanged">
+                
+                <ListBox.ItemsPanel>
+                    <ItemsPanelTemplate>
+                        <WrapPanel
+                            HorizontalAlignment="Left"
+                            VerticalAlignment="Center"
+                            Orientation="Horizontal" />
+                    </ItemsPanelTemplate>
+                </ListBox.ItemsPanel>
+                
+                <ListBox.Styles>
+                    <Style Selector="ListBoxItem">
+                        <Setter Property="Margin" Value="5 5 0 5" />
+                        <Setter Property="CornerRadius" Value="5" />
+                    </Style>
+                    <Style Selector="Rectangle#SelectionIndicator">
+                        <Setter Property="Opacity" Value="0" />
+                    </Style>
+                </ListBox.Styles>
+                
+                <ListBox.DataTemplates>
+                    <DataTemplate
+                        DataType="models:UserProfile">
+                        <Grid
+                            PointerEntered="Grid_PointerEntered"
+                            PointerExited="Grid_OnPointerExited">
+                            <Border
+                                HorizontalAlignment="Stretch"
+                                VerticalAlignment="Stretch"
+                                ClipToBounds="True"
+                                CornerRadius="5"
+                                Background="{Binding BackgroundColor}">
+                                <StackPanel
+                                    HorizontalAlignment="Stretch"
+                                    VerticalAlignment="Stretch">
+                                    <Image
+                                        Width="96"
+                                        Height="96"
+                                        HorizontalAlignment="Stretch"
+                                        VerticalAlignment="Top"
+                                        Source="{Binding Image, Converter={StaticResource ByteImage}}" />
+                                    <TextBlock
+                                        HorizontalAlignment="Stretch"
+                                        MaxWidth="90"
+                                        Text="{Binding Name}"
+                                        TextAlignment="Center"
+                                        TextWrapping="Wrap"
+                                        TextTrimming="CharacterEllipsis"
+                                        MaxLines="2"
+                                        Margin="5" />
+                                </StackPanel>
+                            </Border>
+                        </Grid>
+                    </DataTemplate>
+                    <DataTemplate
+                        DataType="viewModels:BaseModel">
+                        <Panel
+                            Height="118"
+                            Width="96">
+                            <Panel.Styles>
+                                <Style Selector="Panel">
+                                    <Setter Property="Background" Value="{DynamicResource ListBoxBackground}" />
+                                </Style>
+                            </Panel.Styles>
+                        </Panel>
+                    </DataTemplate>
+                </ListBox.DataTemplates>
+            </ListBox>
+        </Border>
+        
+        <StackPanel
+            Grid.Row="1"
+            Margin="0 24 0 0"
+            HorizontalAlignment="Left"
+            Orientation="Horizontal"
+            Spacing="10">
+        </StackPanel>
+    </Grid>
+</UserControl>

src/Ryujinx/UI/Applet/ProfileSelectorDialog.axaml.cs

@@ -0,0 +1,125 @@
+using Avalonia;
+using Avalonia.Controls;
+using Avalonia.Input;
+using FluentAvalonia.UI.Controls;
+using Ryujinx.Ava.Common.Locale;
+using Ryujinx.Ava.UI.Controls;
+using Ryujinx.Ava.UI.Helpers;
+using Ryujinx.Ava.UI.ViewModels;
+using Ryujinx.Common.Logging;
+using Ryujinx.HLE.HOS.Services.Account.Acc;
+using Ryujinx.UI.Common.Configuration;
+using System.Collections.ObjectModel;
+using System.Threading.Tasks;
+using UserProfile = Ryujinx.Ava.UI.Models.UserProfile;
+using UserProfileSft = Ryujinx.HLE.HOS.Services.Account.Acc.UserProfile;
+
+namespace Ryujinx.Ava.UI.Applet
+{
+    public partial class ProfileSelectorDialog : UserControl
+    {
+        public ProfileSelectorDialogViewModel ViewModel { get; set; }
+
+        public ProfileSelectorDialog(ProfileSelectorDialogViewModel viewModel)
+        {
+            DataContext = ViewModel = viewModel;
+            
+            InitializeComponent();
+        }
+        
+        private void Grid_PointerEntered(object sender, PointerEventArgs e)
+        {
+            if (sender is Grid { DataContext: UserProfile profile })
+            {
+                profile.IsPointerOver = true;
+            }
+        }
+
+        private void Grid_OnPointerExited(object sender, PointerEventArgs e)
+        {
+            if (sender is Grid { DataContext: UserProfile profile })
+            {
+                profile.IsPointerOver = false;
+            }
+        }
+
+        private void ProfilesList_SelectionChanged(object sender, SelectionChangedEventArgs e)
+        {
+            if (sender is ListBox listBox)
+            {
+                int selectedIndex = listBox.SelectedIndex;
+
+                if (selectedIndex >= 0 && selectedIndex < ViewModel.Profiles.Count)
+                {
+                    if (ViewModel.Profiles[selectedIndex] is UserProfile userProfile)
+                    {
+                        ViewModel.SelectedUserId = userProfile.UserId;
+                        Logger.Info?.Print(LogClass.UI, $"Selected: {userProfile.UserId}", "ProfileSelector");
+
+                        ObservableCollection<BaseModel> newProfiles = [];
+
+                        foreach (BaseModel item in ViewModel.Profiles)
+                        {
+                            if (item is UserProfile originalItem)
+                            {
+                                UserProfileSft profile = new(originalItem.UserId, originalItem.Name, originalItem.Image);
+                                
+                                if (profile.UserId == ViewModel.SelectedUserId)
+                                {
+                                    profile.AccountState = AccountState.Open;
+                                }
+
+                                newProfiles.Add(new UserProfile(profile, new NavigationDialogHost()));
+                            }
+                        }
+
+                        ViewModel.Profiles = newProfiles;
+                    }
+                }
+            }
+        }
+
+        public static async Task<(UserId Id, bool Result)> ShowInputDialog(ProfileSelectorDialogViewModel viewModel)
+        {
+
+            if (ConfigurationState.Instance.System.SkipUserProfilesManager)
+            {
+                UserId defaultId = viewModel.SelectedUserId;
+                return (defaultId, true);
+            }
+
+            FAContentDialog contentDialog = new()
+            {
+                Title = LocaleManager.Instance[LocaleKeys.UserProfileWindowTitle],
+                PrimaryButtonText = LocaleManager.Instance[LocaleKeys.Continue],
+                SecondaryButtonText = string.Empty,
+                CloseButtonText = LocaleManager.Instance[LocaleKeys.Cancel],
+                Content = new ProfileSelectorDialog(viewModel),
+                Padding = new Thickness(0)
+            };
+
+            UserId result = UserId.Null;
+            bool input = false;
+            
+            contentDialog.Closed += Handler;
+
+            await ContentDialogHelper.ShowAsync(contentDialog);
+
+            return (result, input);
+            
+            void Handler(FAContentDialog sender, FAContentDialogClosedEventArgs eventArgs)
+            {
+                if (eventArgs.Result == FAContentDialogResult.Primary)
+                {
+                    result = viewModel.SelectedUserId;
+                    input = true;
+                }
+                else
+                {
+                    result = UserId.Null;
+                    input = false;
+                }
+            }
+        }
+    }
+}

src/Ryujinx/UI/Renderer/EmbeddedWindow.cs

@@ -4,6 +4,9 @@ using Avalonia.Platform;
 using Ryujinx.Common.Configuration;
 using Ryujinx.UI.Common.Configuration;
 using Ryujinx.UI.Common.Helper;
+using Ryujinx.Common.Logging;
+using Ryujinx.Graphics.RenderDocApi;
+using Ryujinx.HLE;
 using SPB.Graphics;
 using SPB.Platform;
 using SPB.Platform.GLX;
@@ -30,6 +33,7 @@ namespace Ryujinx.Ava.UI.Renderer
         protected nint MetalLayer { get; set; }
 
         public delegate void UpdateBoundsCallbackDelegate(Rect rect);
+
         private UpdateBoundsCallbackDelegate _updateBoundsCallback;
 
         public event EventHandler<nint> WindowCreated;
@@ -46,6 +50,55 @@ namespace Ryujinx.Ava.UI.Renderer
 
         protected virtual void OnWindowDestroyed() { }
 
+        public bool ToggleRenderDocCapture(Switch device)
+        {
+            if (!RenderDoc.IsAvailable) return false;
+
+            if (RenderDoc.IsFrameCapturing)
+            {
+                if (EndRenderDocCapture())
+                {
+                    Logger.Info?.Print(LogClass.Application, "Ended RenderDoc capture.");
+                    return true;
+                }
+            } 
+            else if (StartRenderDocCapture(device))
+            {
+                Logger.Info?.Print(LogClass.Application, "Starting RenderDoc capture.");
+                return true;
+            }
+
+            return false;
+        }
+
+        public bool StartRenderDocCapture(Switch device)
+        {
+            if (!RenderDoc.IsAvailable) return false;
+
+            if (RenderDoc.IsFrameCapturing) return false;
+
+            RenderDoc.StartFrameCapture(nint.Zero, WindowHandle);
+            RenderDoc.SetCaptureTitle(TitleHelper.FormatRenderDocCaptureTitle(device.Processes.ActiveApplication, Program.Version));
+
+            return true;
+        }
+
+        public bool EndRenderDocCapture()
+        {
+            if (!RenderDoc.IsAvailable) return false;
+            if (!RenderDoc.IsFrameCapturing) return false;
+
+            return RenderDoc.IsFrameCapturing && RenderDoc.EndFrameCapture(nint.Zero, WindowHandle);
+        }
+
+        public bool DiscardRenderDocCapture()
+        {
+            if (!RenderDoc.IsAvailable) return false;
+            if (!RenderDoc.IsFrameCapturing) return false;
+
+            return RenderDoc.IsFrameCapturing && RenderDoc.DiscardFrameCapture(nint.Zero, WindowHandle);
+        }
+
         protected virtual void OnWindowDestroying()
         {
             WindowHandle = nint.Zero;
@@ -124,7 +177,9 @@ namespace Ryujinx.Ava.UI.Renderer
             }
             else
             {
-                X11Window = PlatformHelper.CreateOpenGLWindow(new FramebufferFormat(new ColorFormat(8, 8, 8, 0), 16, 0, ColorFormat.Zero, 0, 2, false), 0, 0, 100, 100) as GLXWindow;
+                X11Window = PlatformHelper.CreateOpenGLWindow(
+                    new FramebufferFormat(new ColorFormat(8, 8, 8, 0), 16, 0, ColorFormat.Zero, 0, 2, false), 0, 0, 100,
+                    100) as GLXWindow;
             }
 
             if (X11Window != null)
@@ -141,7 +196,7 @@ namespace Ryujinx.Ava.UI.Renderer
         {
             _className = "NativeWindow-" + Guid.NewGuid();
 
-            _wndProcDelegate = delegate (nint hWnd, WindowsMessages msg, nint wParam, nint lParam)
+            _wndProcDelegate = delegate(nint hWnd, WindowsMessages msg, nint wParam, nint lParam)
             {
                 switch (msg)
                 {
@@ -164,7 +219,8 @@ namespace Ryujinx.Ava.UI.Renderer
 
             RegisterClassEx(ref wndClassEx);
 
-            WindowHandle = CreateWindowEx(0, _className, "NativeWindow", WindowStyles.WsChild, 0, 0, 640, 480, control.Handle, nint.Zero, nint.Zero, nint.Zero);
+            WindowHandle = CreateWindowEx(0, _className, "NativeWindow", WindowStyles.WsChild, 0, 0, 640, 480,
+                control.Handle, nint.Zero, nint.Zero, nint.Zero);
 
             SetWindowLongPtrW(control.Handle, GWLP_WNDPROC, wndClassEx.lpfnWndProc);
 

src/Ryujinx/UI/ViewModels/MainWindowViewModel.cs

@@ -7,6 +7,7 @@ using Avalonia.Platform.Storage;
 using Avalonia.Threading;
 using System.Runtime.Versioning;
 using CommunityToolkit.Mvvm.ComponentModel;
+using CommunityToolkit.Mvvm.Input;
 using DynamicData;
 using DynamicData.Binding;
 using FluentAvalonia.UI.Controls;
@@ -25,6 +26,7 @@ using Ryujinx.Common.Configuration;
 using Ryujinx.Common.Logging;
 using Ryujinx.Common.Utilities;
 using Ryujinx.Cpu;
+using Ryujinx.Graphics.RenderDocApi;
 using Ryujinx.HLE;
 using Ryujinx.HLE.FileSystem;
 using Ryujinx.HLE.HOS;
@@ -1833,6 +1835,31 @@ namespace Ryujinx.Ava.UI.ViewModels
         }
 
 
+        public void ReloadRenderDocApi()
+        {
+            RenderDoc.ReloadApi(ignoreAlreadyLoaded: true);
+
+            OnPropertyChanged(nameof(ShowStartCaptureButton));
+            OnPropertyChanged(nameof(ShowEndCaptureButton));
+            OnPropertyChanged(nameof(RenderDocIsAvailable));
+
+            if (RenderDoc.IsAvailable)
+                RenderDocIsCapturing = RenderDoc.IsFrameCapturing;
+
+            NotificationHelper.ShowInformation(
+                "RenderDoc API reloaded",
+                RenderDoc.IsAvailable ? "RenderDoc is now available." : "RenderDoc is no longer available."
+            );
+        }
+
+        public void ToggleCapture()
+        {
+            if (ShowLoadProgress) return;
+
+            AppHost.RendererHost.EmbeddedWindow.ToggleRenderDocCapture(AppHost.Device);
+            RenderDocIsCapturing = RenderDoc.IsFrameCapturing;
+        }
+
         public void ToggleFullscreen()
         {
             if (Environment.TickCount64 - LastFullscreenToggle < HotKeyPressDelayMs)
@@ -2112,5 +2139,25 @@ namespace Ryujinx.Ava.UI.ViewModels
         }
 
         #endregion
+
+        #region Context Menu commands
+
+        public bool ShowStartCaptureButton => !RenderDocIsCapturing && RenderDoc.IsAvailable;
+        public bool ShowEndCaptureButton => RenderDocIsCapturing && RenderDoc.IsAvailable;
+        public static bool RenderDocIsAvailable => RenderDoc.IsAvailable;
+
+        public bool RenderDocIsCapturing
+        {
+            get;
+            set
+            {
+                field = value;
+                OnPropertyChanged();
+                OnPropertyChanged(nameof(ShowStartCaptureButton));
+                OnPropertyChanged(nameof(ShowEndCaptureButton));
+            }
+        }
+
+        #endregion
     }
 }

src/Ryujinx/UI/ViewModels/ProfileSelectorDialogViewModel.cs

@@ -0,0 +1,14 @@
+using CommunityToolkit.Mvvm.ComponentModel;
+using Ryujinx.HLE.HOS.Services.Account.Acc;
+using System.Collections.ObjectModel;
+
+namespace Ryujinx.Ava.UI.ViewModels
+{
+    public partial class ProfileSelectorDialogViewModel : BaseModel
+    {
+
+        [ObservableProperty] private UserId _selectedUserId;
+
+        [ObservableProperty] private ObservableCollection<BaseModel> _profiles = [];
+    }
+}

src/Ryujinx/UI/Views/Main/MainMenuBarView.axaml

@@ -227,6 +227,26 @@
                     Click="OpenCheatManagerForCurrentApp"
                     Header="{locale:Locale GameListContextMenuManageCheat}"
                     IsEnabled="{Binding IsGameRunning}" />
+                <Separator IsVisible="{Binding RenderDocIsAvailable}" />
+                <MenuItem
+                    Click="StartRenderDocCapture_Click"
+                    IsVisible="{Binding ShowStartCaptureButton}"
+                    CommandParameter="{Binding}"
+                    Header="{locale:Locale MenuBarActions_StartCapture}"
+                    IsEnabled="{Binding IsGameRunning}" />
+                <MenuItem
+                    Click="EndRenderDocCapture_Click"
+                    IsVisible="{Binding ShowEndCaptureButton}"
+                    CommandParameter="{Binding}"
+                    Header="{locale:Locale MenuBarActions_EndCapture}"
+                    IsEnabled="{Binding IsGameRunning}" />
+                <MenuItem
+                    Click="DiscardRenderDocCapture_Click"
+                    IsVisible="{Binding ShowEndCaptureButton}"
+                    CommandParameter="{Binding}"
+                    Header="{locale:Locale MenuBarActions_DiscardCapture}"
+                    ToolTip.Tip="{locale:Locale MenuBarActions_DiscardCapture_ToolTip}"
+                    IsEnabled="{Binding IsGameRunning}" />
             </MenuItem>
             <MenuItem VerticalAlignment="Center" Header="{locale:Locale MenuBarTools}">
                 <MenuItem Header="{locale:Locale MenuBarToolsInstallKeys}" IsEnabled="{Binding EnableNonGameRunningControls}">

src/Ryujinx/UI/Views/Main/MainMenuBarView.axaml.cs

@@ -9,7 +9,9 @@ using Ryujinx.Ava.UI.Helpers;
 using Ryujinx.Ava.UI.ViewModels;
 using Ryujinx.Ava.UI.Windows;
 using Ryujinx.Common;
+using Ryujinx.Common.Logging;
 using Ryujinx.Common.Utilities;
+using Ryujinx.Graphics.RenderDocApi;
 using Ryujinx.HLE.HOS.Services.Nfc.AmiiboDecryption;
 using Ryujinx.Modules;
 using Ryujinx.UI.App.Common;
@@ -266,6 +268,52 @@ namespace Ryujinx.Ava.UI.Views.Main
             }
         }
 
+        public void StartRenderDocCapture_Click(object sender, RoutedEventArgs args)
+        {
+            MainWindowViewModel viewModel = (sender as MenuItem)?.DataContext as MainWindowViewModel;
+
+            if (!RenderDoc.IsFrameCapturing && RenderDoc.IsAvailable && viewModel is not { ShowLoadProgress: true })
+            {
+                if (viewModel != null && viewModel.AppHost.RendererHost
+                        .EmbeddedWindow.StartRenderDocCapture(viewModel.AppHost.Device))
+                {
+                    Logger.Info?.Print(LogClass.Application, "Starting RenderDoc capture.");
+                }
+            }
+
+            viewModel?.RenderDocIsCapturing = RenderDoc.IsFrameCapturing;
+        }
+
+        public void EndRenderDocCapture_Click(object sender, RoutedEventArgs args)
+        {
+            MainWindowViewModel viewModel = (sender as MenuItem)?.DataContext as MainWindowViewModel;
+
+            if (RenderDoc.IsFrameCapturing && RenderDoc.IsAvailable && viewModel is not { ShowLoadProgress: true })
+            {
+                if (viewModel != null && viewModel.AppHost.RendererHost.EmbeddedWindow.EndRenderDocCapture())
+                {
+                    Logger.Info?.Print(LogClass.Application, "Ended RenderDoc capture.");
+                }
+            }
+
+            viewModel?.RenderDocIsCapturing = RenderDoc.IsFrameCapturing;
+        }
+
+        public void DiscardRenderDocCapture_Click(object sender, RoutedEventArgs args)
+        {
+            MainWindowViewModel viewModel = (sender as MenuItem)?.DataContext as MainWindowViewModel;
+
+            if (RenderDoc.IsFrameCapturing && RenderDoc.IsAvailable && viewModel is not { ShowLoadProgress: true })
+            {
+                if (viewModel != null && viewModel.AppHost.RendererHost.EmbeddedWindow.DiscardRenderDocCapture())
+                {
+                    Logger.Info?.Print(LogClass.Application, "Discarded RenderDoc capture.");
+                }
+            }
+
+            viewModel?.RenderDocIsCapturing = RenderDoc.IsFrameCapturing;
+        }
+
         public async void CheckForUpdates(object sender, RoutedEventArgs e)
         {
             if (Updater.CanUpdate(true))

src/Ryujinx/UI/Windows/MainWindow.axaml

@@ -43,6 +43,8 @@
         <KeyBinding Gesture="Ctrl+B" Command="{Binding OpenBinFile}" />
         <KeyBinding Gesture="Ctrl+," Command="{Binding OpenSettings}" />
         <KeyBinding Gesture="Ctrl+R" Command="{Binding RestartEmulation}" />
+        <KeyBinding Gesture="Ctrl+Shift+R" Command="{Binding ReloadRenderDocApi}" />
+        <KeyBinding Gesture="Ctrl+Shift+C" Command="{Binding ToggleCapture}" />
     </Window.KeyBindings>
     <Grid HorizontalAlignment="Stretch" VerticalAlignment="Stretch">
         <Grid.RowDefinitions>