Move solution and projects to src

55 files changed, 15621 insertions, 0 deletions

diff --git a/src/Ryujinx.Graphics.OpenGL/BackgroundContextWorker.cs b/src/Ryujinx.Graphics.OpenGL/BackgroundContextWorker.cs
new file mode 100644
index 00000000..764ea715
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/BackgroundContextWorker.cs
@@ -0,0 +1,91 @@
+using Ryujinx.Common;
+using System;
+using System.Collections.Generic;
+using System.Threading;
+
+namespace Ryujinx.Graphics.OpenGL
+{
+    unsafe class BackgroundContextWorker : IDisposable
+    {
+        [ThreadStatic]
+        public static bool InBackground;
+        private Thread _thread;
+        private bool _running;
+
+        private AutoResetEvent _signal;
+        private Queue<Action> _work;
+        private ObjectPool<ManualResetEventSlim> _invokePool;
+        private readonly IOpenGLContext _backgroundContext;
+
+        public BackgroundContextWorker(IOpenGLContext backgroundContext)
+        {
+            _backgroundContext = backgroundContext;
+            _running = true;
+
+            _signal = new AutoResetEvent(false);
+            _work = new Queue<Action>();
+            _invokePool = new ObjectPool<ManualResetEventSlim>(() => new ManualResetEventSlim(), 10);
+
+            _thread = new Thread(Run);
+            _thread.Start();
+        }
+
+        private void Run()
+        {
+            InBackground = true;
+
+            _backgroundContext.MakeCurrent();
+
+            while (_running)
+            {
+                Action action;
+
+                lock (_work)
+                {
+                    _work.TryDequeue(out action);
+                }
+
+                if (action != null)
+                {
+                    action();
+                }
+                else
+                {
+                    _signal.WaitOne();
+                }
+            }
+
+            _backgroundContext.Dispose();
+        }
+
+        public void Invoke(Action action)
+        {
+            ManualResetEventSlim actionComplete = _invokePool.Allocate();
+
+            lock (_work)
+            {
+                _work.Enqueue(() =>
+                {
+                    action();
+                    actionComplete.Set();
+                });
+            }
+
+            _signal.Set();
+
+            actionComplete.Wait();
+            actionComplete.Reset();
+
+            _invokePool.Release(actionComplete);
+        }
+
+        public void Dispose()
+        {
+            _running = false;
+            _signal.Set();
+
+            _thread.Join();
+            _signal.Dispose();
+        }
+    }
+}
\ No newline at end of file
diff --git a/src/Ryujinx.Graphics.OpenGL/Buffer.cs b/src/Ryujinx.Graphics.OpenGL/Buffer.cs
new file mode 100644
index 00000000..af7d191a
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Buffer.cs
@@ -0,0 +1,103 @@
+using OpenTK.Graphics.OpenGL;
+using Ryujinx.Graphics.GAL;
+using System;
+
+namespace Ryujinx.Graphics.OpenGL
+{
+    static class Buffer
+    {
+        public static void Clear(BufferHandle destination, int offset, int size, uint value)
+        {
+            GL.BindBuffer(BufferTarget.CopyWriteBuffer, destination.ToInt32());
+
+            unsafe
+            {
+                uint* valueArr = stackalloc uint[1];
+
+                valueArr[0] = value;
+
+                GL.ClearBufferSubData(
+                    BufferTarget.CopyWriteBuffer,
+                    PixelInternalFormat.Rgba8ui,
+                    (IntPtr)offset,
+                    (IntPtr)size,
+                    PixelFormat.RgbaInteger,
+                    PixelType.UnsignedByte,
+                    (IntPtr)valueArr);
+            }
+        }
+
+        public static BufferHandle Create()
+        {
+            return Handle.FromInt32<BufferHandle>(GL.GenBuffer());
+        }
+
+        public static BufferHandle Create(int size)
+        {
+            int handle = GL.GenBuffer();
+
+            GL.BindBuffer(BufferTarget.CopyWriteBuffer, handle);
+            GL.BufferData(BufferTarget.CopyWriteBuffer, size, IntPtr.Zero, BufferUsageHint.DynamicDraw);
+
+            return Handle.FromInt32<BufferHandle>(handle);
+        }
+
+        public static void Copy(BufferHandle source, BufferHandle destination, int srcOffset, int dstOffset, int size)
+        {
+            GL.BindBuffer(BufferTarget.CopyReadBuffer, source.ToInt32());
+            GL.BindBuffer(BufferTarget.CopyWriteBuffer, destination.ToInt32());
+
+            GL.CopyBufferSubData(
+                BufferTarget.CopyReadBuffer,
+                BufferTarget.CopyWriteBuffer,
+                (IntPtr)srcOffset,
+                (IntPtr)dstOffset,
+                (IntPtr)size);
+        }
+
+        public static unsafe PinnedSpan<byte> GetData(OpenGLRenderer renderer, BufferHandle buffer, int offset, int size)
+        {
+            // Data in the persistent buffer and host array is guaranteed to be available
+            // until the next time the host thread requests data.
+
+            if (HwCapabilities.UsePersistentBufferForFlush)
+            {
+                return PinnedSpan<byte>.UnsafeFromSpan(renderer.PersistentBuffers.Default.GetBufferData(buffer, offset, size));
+            }
+            else
+            {
+                IntPtr target = renderer.PersistentBuffers.Default.GetHostArray(size);
+
+                GL.BindBuffer(BufferTarget.CopyReadBuffer, buffer.ToInt32());
+
+                GL.GetBufferSubData(BufferTarget.CopyReadBuffer, (IntPtr)offset, size, target);
+
+                return new PinnedSpan<byte>(target.ToPointer(), size);
+            }
+        }
+
+        public static void Resize(BufferHandle handle, int size)
+        {
+            GL.BindBuffer(BufferTarget.CopyWriteBuffer, handle.ToInt32());
+            GL.BufferData(BufferTarget.CopyWriteBuffer, size, IntPtr.Zero, BufferUsageHint.StreamCopy);
+        }
+
+        public static void SetData(BufferHandle buffer, int offset, ReadOnlySpan<byte> data)
+        {
+            GL.BindBuffer(BufferTarget.CopyWriteBuffer, buffer.ToInt32());
+
+            unsafe
+            {
+                fixed (byte* ptr = data)
+                {
+                    GL.BufferSubData(BufferTarget.CopyWriteBuffer, (IntPtr)offset, data.Length, (IntPtr)ptr);
+                }
+            }
+        }
+
+        public static void Delete(BufferHandle buffer)
+        {
+            GL.DeleteBuffer(buffer.ToInt32());
+        }
+    }
+}
diff --git a/src/Ryujinx.Graphics.OpenGL/Constants.cs b/src/Ryujinx.Graphics.OpenGL/Constants.cs
new file mode 100644
index 00000000..8817011a
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Constants.cs
@@ -0,0 +1,11 @@
+namespace Ryujinx.Graphics.OpenGL
+{
+    static class Constants
+    {
+        public const int MaxRenderTargets = 8;
+        public const int MaxViewports = 16;
+        public const int MaxVertexAttribs = 16;
+        public const int MaxVertexBuffers = 16;
+        public const int MaxTransformFeedbackBuffers = 4;
+    }
+}
diff --git a/src/Ryujinx.Graphics.OpenGL/Debugger.cs b/src/Ryujinx.Graphics.OpenGL/Debugger.cs
new file mode 100644
index 00000000..9f67cfc6
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Debugger.cs
@@ -0,0 +1,101 @@
+using OpenTK.Graphics.OpenGL;
+using Ryujinx.Common.Configuration;
+using Ryujinx.Common.Logging;
+using System;
+using System.Runtime.InteropServices;
+using System.Threading;
+
+namespace Ryujinx.Graphics.OpenGL
+{
+    public static class Debugger
+    {
+        private static DebugProc _debugCallback;
+
+        private static int _counter;
+
+        public static void Initialize(GraphicsDebugLevel logLevel)
+        {
+            // Disable everything
+            GL.DebugMessageControl(DebugSourceControl.DontCare, DebugTypeControl.DontCare, DebugSeverityControl.DontCare, 0, (int[])null, false);
+
+            if (logLevel == GraphicsDebugLevel.None)
+            {
+                GL.Disable(EnableCap.DebugOutputSynchronous);
+                GL.DebugMessageCallback(null, IntPtr.Zero);
+
+                return;
+            }
+
+            GL.Enable(EnableCap.DebugOutputSynchronous);
+
+            if (logLevel == GraphicsDebugLevel.Error)
+            {
+                GL.DebugMessageControl(DebugSourceControl.DontCare, DebugTypeControl.DebugTypeError, DebugSeverityControl.DontCare, 0, (int[])null, true);
+            }
+            else if (logLevel == GraphicsDebugLevel.Slowdowns)
+            {
+                GL.DebugMessageControl(DebugSourceControl.DontCare, DebugTypeControl.DebugTypeError, DebugSeverityControl.DontCare, 0, (int[])null, true);
+                GL.DebugMessageControl(DebugSourceControl.DontCare, DebugTypeControl.DebugTypePerformance, DebugSeverityControl.DontCare, 0, (int[])null, true);
+            }
+            else
+            {
+                GL.DebugMessageControl(DebugSourceControl.DontCare, DebugTypeControl.DontCare, DebugSeverityControl.DontCare, 0, (int[])null, true);
+            }
+
+            _counter = 0;
+            _debugCallback = GLDebugHandler;
+
+            GL.DebugMessageCallback(_debugCallback, IntPtr.Zero);
+
+            Logger.Warning?.Print(LogClass.Gpu, "OpenGL Debugging is enabled. Performance will be negatively impacted.");
+        }
+
+        private static void GLDebugHandler(
+            DebugSource source,
+            DebugType type,
+            int id,
+            DebugSeverity severity,
+            int length,
+            IntPtr message,
+            IntPtr userParam)
+        {
+            string msg = Marshal.PtrToStringUTF8(message).Replace('\n', ' ');
+
+            switch (type)
+            {
+                case DebugType.DebugTypeError      : Logger.Error?.Print(LogClass.Gpu, $"{severity}: {msg}\nCallStack={Environment.StackTrace}", "GLERROR"); break;
+                case DebugType.DebugTypePerformance: Logger.Warning?.Print(LogClass.Gpu, $"{severity}: {msg}", "GLPERF"); break;
+                case DebugType.DebugTypePushGroup  : Logger.Info?.Print(LogClass.Gpu, $"{{ ({id}) {severity}: {msg}", "GLINFO"); break;
+                case DebugType.DebugTypePopGroup   : Logger.Info?.Print(LogClass.Gpu, $"}} ({id}) {severity}: {msg}", "GLINFO"); break;
+                default:
+                    if (source == DebugSource.DebugSourceApplication)
+                    {
+                        Logger.Info?.Print(LogClass.Gpu, $"{type} {severity}: {msg}", "GLINFO");
+                    }
+                    else
+                    {
+                        Logger.Debug?.Print(LogClass.Gpu, $"{type} {severity}: {msg}", "GLDEBUG");
+                    }
+                    break;
+            }
+        }
+
+        // Useful debug helpers
+        public static void PushGroup(string dbgMsg)
+        {
+            int counter = Interlocked.Increment(ref _counter);
+
+            GL.PushDebugGroup(DebugSourceExternal.DebugSourceApplication, counter, dbgMsg.Length, dbgMsg);
+        }
+
+        public static void PopGroup()
+        {
+            GL.PopDebugGroup();
+        }
+
+        public static void Print(string dbgMsg, DebugType type = DebugType.DebugTypeMarker, DebugSeverity severity = DebugSeverity.DebugSeverityNotification, int id = 999999)
+        {
+            GL.DebugMessageInsert(DebugSourceExternal.DebugSourceApplication, type, id, severity, dbgMsg.Length, dbgMsg);
+        }
+    }
+}
diff --git a/src/Ryujinx.Graphics.OpenGL/DrawTextureEmulation.cs b/src/Ryujinx.Graphics.OpenGL/DrawTextureEmulation.cs
new file mode 100644
index 00000000..509e20fe
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/DrawTextureEmulation.cs
@@ -0,0 +1,138 @@
+using OpenTK.Graphics.OpenGL;
+using Ryujinx.Graphics.OpenGL.Image;
+using System;
+
+namespace Ryujinx.Graphics.OpenGL
+{
+    class DrawTextureEmulation
+    {
+        private const string VertexShader = @"#version 430 core
+
+uniform float srcX0;
+uniform float srcY0;
+uniform float srcX1;
+uniform float srcY1;
+
+layout (location = 0) out vec2 texcoord;
+
+void main()
+{
+    bool x1 = (gl_VertexID & 1) != 0;
+    bool y1 = (gl_VertexID & 2) != 0;
+    gl_Position = vec4(x1 ? 1 : -1, y1 ? -1 : 1, 0, 1);
+    texcoord = vec2(x1 ? srcX1 : srcX0, y1 ? srcY1 : srcY0);
+}";
+
+        private const string FragmentShader = @"#version 430 core
+
+layout (location = 0) uniform sampler2D tex;
+
+layout (location = 0) in vec2 texcoord;
+layout (location = 0) out vec4 colour;
+
+void main()
+{
+    colour = texture(tex, texcoord);
+}";
+
+        private int _vsHandle;
+        private int _fsHandle;
+        private int _programHandle;
+        private int _uniformSrcX0Location;
+        private int _uniformSrcY0Location;
+        private int _uniformSrcX1Location;
+        private int _uniformSrcY1Location;
+        private bool _initialized;
+
+        public void Draw(
+            TextureView texture,
+            Sampler sampler,
+            float x0,
+            float y0,
+            float x1,
+            float y1,
+            float s0,
+            float t0,
+            float s1,
+            float t1)
+        {
+            EnsureInitialized();
+
+            GL.UseProgram(_programHandle);
+
+            texture.Bind(0);
+            sampler.Bind(0);
+
+            if (x0 > x1)
+            {
+                float temp = s0;
+                s0 = s1;
+                s1 = temp;
+            }
+
+            if (y0 > y1)
+            {
+                float temp = t0;
+                t0 = t1;
+                t1 = temp;
+            }
+
+            GL.Uniform1(_uniformSrcX0Location, s0);
+            GL.Uniform1(_uniformSrcY0Location, t0);
+            GL.Uniform1(_uniformSrcX1Location, s1);
+            GL.Uniform1(_uniformSrcY1Location, t1);
+
+            GL.ViewportIndexed(0, MathF.Min(x0, x1), MathF.Min(y0, y1), MathF.Abs(x1 - x0), MathF.Abs(y1 - y0));
+
+            GL.DrawArrays(PrimitiveType.TriangleStrip, 0, 4);
+        }
+
+        private void EnsureInitialized()
+        {
+            if (_initialized)
+            {
+                return;
+            }
+
+            _initialized = true;
+
+            _vsHandle = GL.CreateShader(ShaderType.VertexShader);
+            _fsHandle = GL.CreateShader(ShaderType.FragmentShader);
+
+            GL.ShaderSource(_vsHandle, VertexShader);
+            GL.ShaderSource(_fsHandle, FragmentShader);
+
+            GL.CompileShader(_vsHandle);
+            GL.CompileShader(_fsHandle);
+
+            _programHandle = GL.CreateProgram();
+
+            GL.AttachShader(_programHandle, _vsHandle);
+            GL.AttachShader(_programHandle, _fsHandle);
+
+            GL.LinkProgram(_programHandle);
+
+            GL.DetachShader(_programHandle, _vsHandle);
+            GL.DetachShader(_programHandle, _fsHandle);
+
+            _uniformSrcX0Location = GL.GetUniformLocation(_programHandle, "srcX0");
+            _uniformSrcY0Location = GL.GetUniformLocation(_programHandle, "srcY0");
+            _uniformSrcX1Location = GL.GetUniformLocation(_programHandle, "srcX1");
+            _uniformSrcY1Location = GL.GetUniformLocation(_programHandle, "srcY1");
+        }
+
+        public void Dispose()
+        {
+            if (!_initialized)
+            {
+                return;
+            }
+
+            GL.DeleteShader(_vsHandle);
+            GL.DeleteShader(_fsHandle);
+            GL.DeleteProgram(_programHandle);
+
+            _initialized = false;
+        }
+    }
+}
diff --git a/src/Ryujinx.Graphics.OpenGL/Effects/FsrScalingFilter.cs b/src/Ryujinx.Graphics.OpenGL/Effects/FsrScalingFilter.cs
new file mode 100644
index 00000000..16678bb7
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Effects/FsrScalingFilter.cs
@@ -0,0 +1,177 @@
+using OpenTK.Graphics.OpenGL;
+using Ryujinx.Common;
+using Ryujinx.Graphics.GAL;
+using Ryujinx.Graphics.OpenGL.Image;
+using System;
+using static Ryujinx.Graphics.OpenGL.Effects.ShaderHelper;
+
+namespace Ryujinx.Graphics.OpenGL.Effects
+{
+    internal class FsrScalingFilter : IScalingFilter
+    {
+        private readonly OpenGLRenderer _renderer;
+        private int _inputUniform;
+        private int _outputUniform;
+        private int _sharpeningUniform;
+        private int _srcX0Uniform;
+        private int _srcX1Uniform;
+        private int _srcY0Uniform;
+        private int _scalingShaderProgram;
+        private int _sharpeningShaderProgram;
+        private float _scale = 1;
+        private int _srcY1Uniform;
+        private int _dstX0Uniform;
+        private int _dstX1Uniform;
+        private int _dstY0Uniform;
+        private int _dstY1Uniform;
+        private int _scaleXUniform;
+        private int _scaleYUniform;
+        private TextureStorage _intermediaryTexture;
+
+        public float Level
+        {
+            get => _scale;
+            set
+            {
+                _scale = MathF.Max(0.01f, value);
+            }
+        }
+
+        public FsrScalingFilter(OpenGLRenderer renderer, IPostProcessingEffect filter)
+        {
+            Initialize();
+
+            _renderer = renderer;
+        }
+
+        public void Dispose()
+        {
+            if (_scalingShaderProgram != 0)
+            {
+                GL.DeleteProgram(_scalingShaderProgram);
+                GL.DeleteProgram(_sharpeningShaderProgram);
+            }
+
+            _intermediaryTexture?.Dispose();
+        }
+
+        private void Initialize()
+        {
+            var scalingShader = EmbeddedResources.ReadAllText("Ryujinx.Graphics.OpenGL/Effects/Shaders/fsr_scaling.glsl");
+            var sharpeningShader = EmbeddedResources.ReadAllText("Ryujinx.Graphics.OpenGL/Effects/Shaders/fsr_sharpening.glsl");
+            var fsrA = EmbeddedResources.ReadAllText("Ryujinx.Graphics.OpenGL/Effects/Shaders/ffx_a.h");
+            var fsr1 = EmbeddedResources.ReadAllText("Ryujinx.Graphics.OpenGL/Effects/Shaders/ffx_fsr1.h");
+
+            scalingShader = scalingShader.Replace("#include \"ffx_a.h\"", fsrA);
+            scalingShader = scalingShader.Replace("#include \"ffx_fsr1.h\"", fsr1);
+            sharpeningShader = sharpeningShader.Replace("#include \"ffx_a.h\"", fsrA);
+            sharpeningShader = sharpeningShader.Replace("#include \"ffx_fsr1.h\"", fsr1);
+
+            _scalingShaderProgram = CompileProgram(scalingShader, ShaderType.ComputeShader);
+            _sharpeningShaderProgram = CompileProgram(sharpeningShader, ShaderType.ComputeShader);
+
+            _inputUniform = GL.GetUniformLocation(_scalingShaderProgram, "Source");
+            _outputUniform = GL.GetUniformLocation(_scalingShaderProgram, "imgOutput");
+            _sharpeningUniform = GL.GetUniformLocation(_sharpeningShaderProgram, "sharpening");
+
+            _srcX0Uniform = GL.GetUniformLocation(_scalingShaderProgram, "srcX0");
+            _srcX1Uniform = GL.GetUniformLocation(_scalingShaderProgram, "srcX1");
+            _srcY0Uniform = GL.GetUniformLocation(_scalingShaderProgram, "srcY0");
+            _srcY1Uniform = GL.GetUniformLocation(_scalingShaderProgram, "srcY1");
+            _dstX0Uniform = GL.GetUniformLocation(_scalingShaderProgram, "dstX0");
+            _dstX1Uniform = GL.GetUniformLocation(_scalingShaderProgram, "dstX1");
+            _dstY0Uniform = GL.GetUniformLocation(_scalingShaderProgram, "dstY0");
+            _dstY1Uniform = GL.GetUniformLocation(_scalingShaderProgram, "dstY1");
+            _scaleXUniform = GL.GetUniformLocation(_scalingShaderProgram, "scaleX");
+            _scaleYUniform = GL.GetUniformLocation(_scalingShaderProgram, "scaleY");
+        }
+
+        public void Run(
+            TextureView view,
+            TextureView destinationTexture,
+            int width,
+            int height,
+            Extents2D source,
+            Extents2D destination)
+        {
+            if (_intermediaryTexture == null || _intermediaryTexture.Info.Width != width || _intermediaryTexture.Info.Height != height)
+            {
+                _intermediaryTexture?.Dispose();
+                var originalInfo = view.Info;
+                var info = new TextureCreateInfo(width,
+                    height,
+                    originalInfo.Depth,
+                    originalInfo.Levels,
+                    originalInfo.Samples,
+                    originalInfo.BlockWidth,
+                    originalInfo.BlockHeight,
+                    originalInfo.BytesPerPixel,
+                    originalInfo.Format,
+                    originalInfo.DepthStencilMode,
+                    originalInfo.Target,
+                    originalInfo.SwizzleR,
+                    originalInfo.SwizzleG,
+                    originalInfo.SwizzleB,
+                    originalInfo.SwizzleA);
+
+                _intermediaryTexture = new TextureStorage(_renderer, info, view.ScaleFactor);
+                _intermediaryTexture.CreateDefaultView();
+            }
+
+            var textureView = _intermediaryTexture.CreateView(_intermediaryTexture.Info, 0, 0) as TextureView;
+
+            int previousProgram = GL.GetInteger(GetPName.CurrentProgram);
+            int previousUnit = GL.GetInteger(GetPName.ActiveTexture);
+            GL.ActiveTexture(TextureUnit.Texture0);
+            int previousTextureBinding = GL.GetInteger(GetPName.TextureBinding2D);
+
+            GL.BindImageTexture(0, textureView.Handle, 0, false, 0, TextureAccess.ReadWrite, SizedInternalFormat.Rgba8);
+
+            int threadGroupWorkRegionDim = 16;
+            int dispatchX = (width + (threadGroupWorkRegionDim - 1)) / threadGroupWorkRegionDim;
+            int dispatchY = (height + (threadGroupWorkRegionDim - 1)) / threadGroupWorkRegionDim;
+
+            // Scaling pass
+            float srcWidth = Math.Abs(source.X2 - source.X1);
+            float srcHeight = Math.Abs(source.Y2 - source.Y1);
+            float scaleX = srcWidth / view.Width;
+            float scaleY = srcHeight / view.Height;
+            GL.UseProgram(_scalingShaderProgram);
+            view.Bind(0);
+            GL.Uniform1(_inputUniform, 0);
+            GL.Uniform1(_outputUniform, 0);
+            GL.Uniform1(_srcX0Uniform, (float)source.X1);
+            GL.Uniform1(_srcX1Uniform, (float)source.X2);
+            GL.Uniform1(_srcY0Uniform, (float)source.Y1);
+            GL.Uniform1(_srcY1Uniform, (float)source.Y2);
+            GL.Uniform1(_dstX0Uniform, (float)destination.X1);
+            GL.Uniform1(_dstX1Uniform, (float)destination.X2);
+            GL.Uniform1(_dstY0Uniform, (float)destination.Y1);
+            GL.Uniform1(_dstY1Uniform, (float)destination.Y2);
+            GL.Uniform1(_scaleXUniform, scaleX);
+            GL.Uniform1(_scaleYUniform, scaleY);
+            GL.DispatchCompute(dispatchX, dispatchY, 1);
+
+            GL.MemoryBarrier(MemoryBarrierFlags.ShaderImageAccessBarrierBit);
+
+            // Sharpening Pass
+            GL.UseProgram(_sharpeningShaderProgram);
+            GL.BindImageTexture(0, destinationTexture.Handle, 0, false, 0, TextureAccess.ReadWrite, SizedInternalFormat.Rgba8);
+            textureView.Bind(0);
+            GL.Uniform1(_inputUniform, 0);
+            GL.Uniform1(_outputUniform, 0);
+            GL.Uniform1(_sharpeningUniform, 1.5f - (Level * 0.01f * 1.5f));
+            GL.DispatchCompute(dispatchX, dispatchY, 1);
+
+            GL.UseProgram(previousProgram);
+            GL.MemoryBarrier(MemoryBarrierFlags.ShaderImageAccessBarrierBit);
+
+            (_renderer.Pipeline as Pipeline).RestoreImages1And2();
+
+            GL.ActiveTexture(TextureUnit.Texture0);
+            GL.BindTexture(TextureTarget.Texture2D, previousTextureBinding);
+
+            GL.ActiveTexture((TextureUnit)previousUnit);
+        }
+    }
+}
\ No newline at end of file
diff --git a/src/Ryujinx.Graphics.OpenGL/Effects/FxaaPostProcessingEffect.cs b/src/Ryujinx.Graphics.OpenGL/Effects/FxaaPostProcessingEffect.cs
new file mode 100644
index 00000000..3a2d685b
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Effects/FxaaPostProcessingEffect.cs
@@ -0,0 +1,81 @@
+using OpenTK.Graphics.OpenGL;
+using Ryujinx.Common;
+using Ryujinx.Graphics.OpenGL.Image;
+
+namespace Ryujinx.Graphics.OpenGL.Effects
+{
+    internal class FxaaPostProcessingEffect : IPostProcessingEffect
+    {
+        private readonly OpenGLRenderer _renderer;
+        private int _resolutionUniform;
+        private int _inputUniform;
+        private int _outputUniform;
+        private int _shaderProgram;
+        private TextureStorage _textureStorage;
+
+        public FxaaPostProcessingEffect(OpenGLRenderer renderer)
+        {
+            Initialize();
+
+            _renderer = renderer;
+        }
+
+        public void Dispose()
+        {
+            if (_shaderProgram != 0)
+            {
+                GL.DeleteProgram(_shaderProgram);
+                _textureStorage?.Dispose();
+            }
+        }
+
+        private void Initialize()
+        {
+            _shaderProgram = ShaderHelper.CompileProgram(EmbeddedResources.ReadAllText("Ryujinx.Graphics.OpenGL/Effects/Shaders/fxaa.glsl"), ShaderType.ComputeShader);
+
+            _resolutionUniform = GL.GetUniformLocation(_shaderProgram, "invResolution");
+            _inputUniform = GL.GetUniformLocation(_shaderProgram, "inputTexture");
+            _outputUniform = GL.GetUniformLocation(_shaderProgram, "imgOutput");
+        }
+
+        public TextureView Run(TextureView view, int width, int height)
+        {
+            if (_textureStorage == null || _textureStorage.Info.Width != view.Width || _textureStorage.Info.Height != view.Height)
+            {
+                _textureStorage?.Dispose();
+                _textureStorage = new TextureStorage(_renderer, view.Info, view.ScaleFactor);
+                _textureStorage.CreateDefaultView();
+            }
+
+            var textureView = _textureStorage.CreateView(view.Info, 0, 0) as TextureView;
+
+            int previousProgram = GL.GetInteger(GetPName.CurrentProgram);
+            int previousUnit = GL.GetInteger(GetPName.ActiveTexture);
+            GL.ActiveTexture(TextureUnit.Texture0);
+            int previousTextureBinding = GL.GetInteger(GetPName.TextureBinding2D);
+
+            GL.BindImageTexture(0, textureView.Handle, 0, false, 0, TextureAccess.ReadWrite, SizedInternalFormat.Rgba8);
+            GL.UseProgram(_shaderProgram);
+
+            var dispatchX = BitUtils.DivRoundUp(view.Width, IPostProcessingEffect.LocalGroupSize);
+            var dispatchY = BitUtils.DivRoundUp(view.Height, IPostProcessingEffect.LocalGroupSize);
+
+            view.Bind(0);
+            GL.Uniform1(_inputUniform, 0);
+            GL.Uniform1(_outputUniform, 0);
+            GL.Uniform2(_resolutionUniform, (float)view.Width, (float)view.Height);
+            GL.DispatchCompute(dispatchX, dispatchY, 1);
+            GL.UseProgram(previousProgram);
+            GL.MemoryBarrier(MemoryBarrierFlags.ShaderImageAccessBarrierBit);
+
+            (_renderer.Pipeline as Pipeline).RestoreImages1And2();
+
+            GL.ActiveTexture(TextureUnit.Texture0);
+            GL.BindTexture(TextureTarget.Texture2D, previousTextureBinding);
+
+            GL.ActiveTexture((TextureUnit)previousUnit);
+
+            return textureView;
+        }
+    }
+}
\ No newline at end of file
diff --git a/src/Ryujinx.Graphics.OpenGL/Effects/IPostProcessingEffect.cs b/src/Ryujinx.Graphics.OpenGL/Effects/IPostProcessingEffect.cs
new file mode 100644
index 00000000..7a045a02
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Effects/IPostProcessingEffect.cs
@@ -0,0 +1,11 @@
+using Ryujinx.Graphics.OpenGL.Image;
+using System;
+
+namespace Ryujinx.Graphics.OpenGL.Effects
+{
+    internal interface IPostProcessingEffect :  IDisposable
+    {
+        const int LocalGroupSize = 64;
+        TextureView Run(TextureView view, int width, int height);
+    }
+}
\ No newline at end of file
diff --git a/src/Ryujinx.Graphics.OpenGL/Effects/IScalingFilter.cs b/src/Ryujinx.Graphics.OpenGL/Effects/IScalingFilter.cs
new file mode 100644
index 00000000..e1e1b2c1
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Effects/IScalingFilter.cs
@@ -0,0 +1,18 @@
+using Ryujinx.Graphics.GAL;
+using Ryujinx.Graphics.OpenGL.Image;
+using System;
+
+namespace Ryujinx.Graphics.OpenGL.Effects
+{
+    internal interface IScalingFilter : IDisposable
+    {
+        float Level { get; set; }
+        void Run(
+            TextureView view,
+            TextureView destinationTexture,
+            int width,
+            int height,
+            Extents2D source,
+            Extents2D destination);
+    }
+}
\ No newline at end of file
diff --git a/src/Ryujinx.Graphics.OpenGL/Effects/ShaderHelper.cs b/src/Ryujinx.Graphics.OpenGL/Effects/ShaderHelper.cs
new file mode 100644
index 00000000..72c5a98f
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Effects/ShaderHelper.cs
@@ -0,0 +1,40 @@
+using OpenTK.Graphics.OpenGL;
+using System;
+
+namespace Ryujinx.Graphics.OpenGL.Effects
+{
+    internal static class ShaderHelper
+    {
+        public static int CompileProgram(string shaderCode, ShaderType shaderType)
+        {
+            var shader = GL.CreateShader(shaderType);
+            GL.ShaderSource(shader, shaderCode);
+            GL.CompileShader(shader);
+
+            var program = GL.CreateProgram();
+            GL.AttachShader(program, shader);
+            GL.LinkProgram(program);
+
+            GL.DetachShader(program, shader);
+            GL.DeleteShader(shader);
+
+            return program;
+        }
+
+        public static int CompileProgram(string[] shaders, ShaderType shaderType)
+        {
+            var shader = GL.CreateShader(shaderType);
+            GL.ShaderSource(shader, shaders.Length, shaders, (int[])null);
+            GL.CompileShader(shader);
+
+            var program = GL.CreateProgram();
+            GL.AttachShader(program, shader);
+            GL.LinkProgram(program);
+
+            GL.DetachShader(program, shader);
+            GL.DeleteShader(shader);
+
+            return program;
+        }
+    }
+}
diff --git a/src/Ryujinx.Graphics.OpenGL/Effects/Shaders/ffx_a.h b/src/Ryujinx.Graphics.OpenGL/Effects/Shaders/ffx_a.h
new file mode 100644
index 00000000..d04bff55
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Effects/Shaders/ffx_a.h
@@ -0,0 +1,2656 @@
+//==============================================================================================================================
+//
+//                                               [A] SHADER PORTABILITY 1.20210629
+//
+//==============================================================================================================================
+// FidelityFX Super Resolution Sample
+//
+// Copyright (c) 2021 Advanced Micro Devices, Inc. All rights reserved.
+// Permission is hereby granted, free of charge, to any person obtaining a copy
+// of this software and associated documentation files(the "Software"), to deal
+// in the Software without restriction, including without limitation the rights
+// to use, copy, modify, merge, publish, distribute, sublicense, and / or sell
+// copies of the Software, and to permit persons to whom the Software is
+// furnished to do so, subject to the following conditions :
+// The above copyright notice and this permission notice shall be included in
+// all copies or substantial portions of the Software.
+// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.IN NO EVENT SHALL THE
+// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+// THE SOFTWARE.
+//------------------------------------------------------------------------------------------------------------------------------
+// MIT LICENSE
+// ===========
+// Copyright (c) 2014 Michal Drobot (for concepts used in "FLOAT APPROXIMATIONS").
+// -----------
+// Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation
+// files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy,
+// modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the
+// Software is furnished to do so, subject to the following conditions:
+// -----------
+// The above copyright notice and this permission notice shall be included in all copies or substantial portions of the
+// Software.
+// -----------
+// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE
+// WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL THE AUTHORS OR
+// COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+// ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+//------------------------------------------------------------------------------------------------------------------------------
+// ABOUT
+// =====
+// Common central point for high-level shading language and C portability for various shader headers.
+//------------------------------------------------------------------------------------------------------------------------------
+// DEFINES
+// =======
+// A_CPU ..... Include the CPU related code.
+// A_GPU ..... Include the GPU related code.
+// A_GLSL .... Using GLSL.
+// A_HLSL .... Using HLSL.
+// A_HLSL_6_2  Using HLSL 6.2 with new 'uint16_t' and related types (requires '-enable-16bit-types').
+// A_NO_16_BIT_CAST Don't use instructions that are not availabe in SPIR-V (needed for running A_HLSL_6_2 on Vulkan)
+// A_GCC ..... Using a GCC compatible compiler (else assume MSVC compatible compiler by default).
+// =======
+// A_BYTE .... Support 8-bit integer.
+// A_HALF .... Support 16-bit integer and floating point.
+// A_LONG .... Support 64-bit integer.
+// A_DUBL .... Support 64-bit floating point.
+// =======
+// A_WAVE .... Support wave-wide operations.
+//------------------------------------------------------------------------------------------------------------------------------
+// To get #include "ffx_a.h" working in GLSL use '#extension GL_GOOGLE_include_directive:require'.
+//------------------------------------------------------------------------------------------------------------------------------
+// SIMPLIFIED TYPE SYSTEM
+// ======================
+//  - All ints will be unsigned with exception of when signed is required.
+//  - Type naming simplified and shortened "A<type><#components>",
+//     - H = 16-bit float (half)
+//     - F = 32-bit float (float)
+//     - D = 64-bit float (double)
+//     - P = 1-bit integer (predicate, not using bool because 'B' is used for byte)
+//     - B = 8-bit integer (byte)
+//     - W = 16-bit integer (word)
+//     - U = 32-bit integer (unsigned)
+//     - L = 64-bit integer (long)
+//  - Using "AS<type><#components>" for signed when required.
+//------------------------------------------------------------------------------------------------------------------------------
+// TODO
+// ====
+//  - Make sure 'ALerp*(a,b,m)' does 'b*m+(-a*m+a)' (2 ops).
+//------------------------------------------------------------------------------------------------------------------------------
+// CHANGE LOG
+// ==========
+// 20200914 - Expanded wave ops and prx code.
+// 20200713 - Added [ZOL] section, fixed serious bugs in sRGB and Rec.709 color conversion code, etc.
+//==============================================================================================================================
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                           COMMON
+//==============================================================================================================================
+#define A_2PI 6.28318530718
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//
+//
+//                                                             CPU
+//
+//
+//==============================================================================================================================
+#ifdef A_CPU
+ // Supporting user defined overrides.
+ #ifndef A_RESTRICT
+  #define A_RESTRICT __restrict
+ #endif
+//------------------------------------------------------------------------------------------------------------------------------
+ #ifndef A_STATIC
+  #define A_STATIC static
+ #endif
+//------------------------------------------------------------------------------------------------------------------------------
+ // Same types across CPU and GPU.
+ // Predicate uses 32-bit integer (C friendly bool).
+ typedef uint32_t AP1;
+ typedef float AF1;
+ typedef double AD1;
+ typedef uint8_t AB1;
+ typedef uint16_t AW1;
+ typedef uint32_t AU1;
+ typedef uint64_t AL1;
+ typedef int8_t ASB1;
+ typedef int16_t ASW1;
+ typedef int32_t ASU1;
+ typedef int64_t ASL1;
+//------------------------------------------------------------------------------------------------------------------------------
+ #define AD1_(a) ((AD1)(a))
+ #define AF1_(a) ((AF1)(a))
+ #define AL1_(a) ((AL1)(a))
+ #define AU1_(a) ((AU1)(a))
+//------------------------------------------------------------------------------------------------------------------------------
+ #define ASL1_(a) ((ASL1)(a))
+ #define ASU1_(a) ((ASU1)(a))
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC AU1 AU1_AF1(AF1 a){union{AF1 f;AU1 u;}bits;bits.f=a;return bits.u;}
+//------------------------------------------------------------------------------------------------------------------------------
+ #define A_TRUE 1
+ #define A_FALSE 0
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//
+//                                                       CPU/GPU PORTING
+//
+//------------------------------------------------------------------------------------------------------------------------------
+// Get CPU and GPU to share all setup code, without duplicate code paths.
+// This uses a lower-case prefix for special vector constructs.
+//  - In C restrict pointers are used.
+//  - In the shading language, in/inout/out arguments are used.
+// This depends on the ability to access a vector value in both languages via array syntax (aka color[2]).
+//==============================================================================================================================
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                     VECTOR ARGUMENT/RETURN/INITIALIZATION PORTABILITY
+//==============================================================================================================================
+ #define retAD2 AD1 *A_RESTRICT
+ #define retAD3 AD1 *A_RESTRICT
+ #define retAD4 AD1 *A_RESTRICT
+ #define retAF2 AF1 *A_RESTRICT
+ #define retAF3 AF1 *A_RESTRICT
+ #define retAF4 AF1 *A_RESTRICT
+ #define retAL2 AL1 *A_RESTRICT
+ #define retAL3 AL1 *A_RESTRICT
+ #define retAL4 AL1 *A_RESTRICT
+ #define retAU2 AU1 *A_RESTRICT
+ #define retAU3 AU1 *A_RESTRICT
+ #define retAU4 AU1 *A_RESTRICT
+//------------------------------------------------------------------------------------------------------------------------------
+ #define inAD2 AD1 *A_RESTRICT
+ #define inAD3 AD1 *A_RESTRICT
+ #define inAD4 AD1 *A_RESTRICT
+ #define inAF2 AF1 *A_RESTRICT
+ #define inAF3 AF1 *A_RESTRICT
+ #define inAF4 AF1 *A_RESTRICT
+ #define inAL2 AL1 *A_RESTRICT
+ #define inAL3 AL1 *A_RESTRICT
+ #define inAL4 AL1 *A_RESTRICT
+ #define inAU2 AU1 *A_RESTRICT
+ #define inAU3 AU1 *A_RESTRICT
+ #define inAU4 AU1 *A_RESTRICT
+//------------------------------------------------------------------------------------------------------------------------------
+ #define inoutAD2 AD1 *A_RESTRICT
+ #define inoutAD3 AD1 *A_RESTRICT
+ #define inoutAD4 AD1 *A_RESTRICT
+ #define inoutAF2 AF1 *A_RESTRICT
+ #define inoutAF3 AF1 *A_RESTRICT
+ #define inoutAF4 AF1 *A_RESTRICT
+ #define inoutAL2 AL1 *A_RESTRICT
+ #define inoutAL3 AL1 *A_RESTRICT
+ #define inoutAL4 AL1 *A_RESTRICT
+ #define inoutAU2 AU1 *A_RESTRICT
+ #define inoutAU3 AU1 *A_RESTRICT
+ #define inoutAU4 AU1 *A_RESTRICT
+//------------------------------------------------------------------------------------------------------------------------------
+ #define outAD2 AD1 *A_RESTRICT
+ #define outAD3 AD1 *A_RESTRICT
+ #define outAD4 AD1 *A_RESTRICT
+ #define outAF2 AF1 *A_RESTRICT
+ #define outAF3 AF1 *A_RESTRICT
+ #define outAF4 AF1 *A_RESTRICT
+ #define outAL2 AL1 *A_RESTRICT
+ #define outAL3 AL1 *A_RESTRICT
+ #define outAL4 AL1 *A_RESTRICT
+ #define outAU2 AU1 *A_RESTRICT
+ #define outAU3 AU1 *A_RESTRICT
+ #define outAU4 AU1 *A_RESTRICT
+//------------------------------------------------------------------------------------------------------------------------------
+ #define varAD2(x) AD1 x[2]
+ #define varAD3(x) AD1 x[3]
+ #define varAD4(x) AD1 x[4]
+ #define varAF2(x) AF1 x[2]
+ #define varAF3(x) AF1 x[3]
+ #define varAF4(x) AF1 x[4]
+ #define varAL2(x) AL1 x[2]
+ #define varAL3(x) AL1 x[3]
+ #define varAL4(x) AL1 x[4]
+ #define varAU2(x) AU1 x[2]
+ #define varAU3(x) AU1 x[3]
+ #define varAU4(x) AU1 x[4]
+//------------------------------------------------------------------------------------------------------------------------------
+ #define initAD2(x,y) {x,y}
+ #define initAD3(x,y,z) {x,y,z}
+ #define initAD4(x,y,z,w) {x,y,z,w}
+ #define initAF2(x,y) {x,y}
+ #define initAF3(x,y,z) {x,y,z}
+ #define initAF4(x,y,z,w) {x,y,z,w}
+ #define initAL2(x,y) {x,y}
+ #define initAL3(x,y,z) {x,y,z}
+ #define initAL4(x,y,z,w) {x,y,z,w}
+ #define initAU2(x,y) {x,y}
+ #define initAU3(x,y,z) {x,y,z}
+ #define initAU4(x,y,z,w) {x,y,z,w}
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                     SCALAR RETURN OPS
+//------------------------------------------------------------------------------------------------------------------------------
+// TODO
+// ====
+//  - Replace transcendentals with manual versions. 
+//==============================================================================================================================
+ #ifdef A_GCC
+  A_STATIC AD1 AAbsD1(AD1 a){return __builtin_fabs(a);}
+  A_STATIC AF1 AAbsF1(AF1 a){return __builtin_fabsf(a);}
+  A_STATIC AU1 AAbsSU1(AU1 a){return AU1_(__builtin_abs(ASU1_(a)));}
+  A_STATIC AL1 AAbsSL1(AL1 a){return AL1_(__builtin_llabs(ASL1_(a)));}
+ #else
+  A_STATIC AD1 AAbsD1(AD1 a){return fabs(a);}
+  A_STATIC AF1 AAbsF1(AF1 a){return fabsf(a);}
+  A_STATIC AU1 AAbsSU1(AU1 a){return AU1_(abs(ASU1_(a)));}
+  A_STATIC AL1 AAbsSL1(AL1 a){return AL1_(labs((long)ASL1_(a)));}
+ #endif
+//------------------------------------------------------------------------------------------------------------------------------
+ #ifdef A_GCC
+  A_STATIC AD1 ACosD1(AD1 a){return __builtin_cos(a);}
+  A_STATIC AF1 ACosF1(AF1 a){return __builtin_cosf(a);}
+ #else
+  A_STATIC AD1 ACosD1(AD1 a){return cos(a);}
+  A_STATIC AF1 ACosF1(AF1 a){return cosf(a);}
+ #endif
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC AD1 ADotD2(inAD2 a,inAD2 b){return a[0]*b[0]+a[1]*b[1];}
+ A_STATIC AD1 ADotD3(inAD3 a,inAD3 b){return a[0]*b[0]+a[1]*b[1]+a[2]*b[2];}
+ A_STATIC AD1 ADotD4(inAD4 a,inAD4 b){return a[0]*b[0]+a[1]*b[1]+a[2]*b[2]+a[3]*b[3];}
+ A_STATIC AF1 ADotF2(inAF2 a,inAF2 b){return a[0]*b[0]+a[1]*b[1];}
+ A_STATIC AF1 ADotF3(inAF3 a,inAF3 b){return a[0]*b[0]+a[1]*b[1]+a[2]*b[2];}
+ A_STATIC AF1 ADotF4(inAF4 a,inAF4 b){return a[0]*b[0]+a[1]*b[1]+a[2]*b[2]+a[3]*b[3];}
+//------------------------------------------------------------------------------------------------------------------------------
+ #ifdef A_GCC
+  A_STATIC AD1 AExp2D1(AD1 a){return __builtin_exp2(a);}
+  A_STATIC AF1 AExp2F1(AF1 a){return __builtin_exp2f(a);}
+ #else
+  A_STATIC AD1 AExp2D1(AD1 a){return exp2(a);}
+  A_STATIC AF1 AExp2F1(AF1 a){return exp2f(a);}
+ #endif
+//------------------------------------------------------------------------------------------------------------------------------
+ #ifdef A_GCC
+  A_STATIC AD1 AFloorD1(AD1 a){return __builtin_floor(a);}
+  A_STATIC AF1 AFloorF1(AF1 a){return __builtin_floorf(a);}
+ #else
+  A_STATIC AD1 AFloorD1(AD1 a){return floor(a);}
+  A_STATIC AF1 AFloorF1(AF1 a){return floorf(a);}
+ #endif
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC AD1 ALerpD1(AD1 a,AD1 b,AD1 c){return b*c+(-a*c+a);}
+ A_STATIC AF1 ALerpF1(AF1 a,AF1 b,AF1 c){return b*c+(-a*c+a);}
+//------------------------------------------------------------------------------------------------------------------------------
+ #ifdef A_GCC
+  A_STATIC AD1 ALog2D1(AD1 a){return __builtin_log2(a);}
+  A_STATIC AF1 ALog2F1(AF1 a){return __builtin_log2f(a);}
+ #else
+  A_STATIC AD1 ALog2D1(AD1 a){return log2(a);}
+  A_STATIC AF1 ALog2F1(AF1 a){return log2f(a);}
+ #endif
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC AD1 AMaxD1(AD1 a,AD1 b){return a>b?a:b;}
+ A_STATIC AF1 AMaxF1(AF1 a,AF1 b){return a>b?a:b;}
+ A_STATIC AL1 AMaxL1(AL1 a,AL1 b){return a>b?a:b;}
+ A_STATIC AU1 AMaxU1(AU1 a,AU1 b){return a>b?a:b;}
+//------------------------------------------------------------------------------------------------------------------------------
+ // These follow the convention that A integer types don't have signage, until they are operated on. 
+ A_STATIC AL1 AMaxSL1(AL1 a,AL1 b){return (ASL1_(a)>ASL1_(b))?a:b;}
+ A_STATIC AU1 AMaxSU1(AU1 a,AU1 b){return (ASU1_(a)>ASU1_(b))?a:b;}
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC AD1 AMinD1(AD1 a,AD1 b){return a<b?a:b;}
+ A_STATIC AF1 AMinF1(AF1 a,AF1 b){return a<b?a:b;}
+ A_STATIC AL1 AMinL1(AL1 a,AL1 b){return a<b?a:b;}
+ A_STATIC AU1 AMinU1(AU1 a,AU1 b){return a<b?a:b;}
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC AL1 AMinSL1(AL1 a,AL1 b){return (ASL1_(a)<ASL1_(b))?a:b;}
+ A_STATIC AU1 AMinSU1(AU1 a,AU1 b){return (ASU1_(a)<ASU1_(b))?a:b;}
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC AD1 ARcpD1(AD1 a){return 1.0/a;}
+ A_STATIC AF1 ARcpF1(AF1 a){return 1.0f/a;}
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC AL1 AShrSL1(AL1 a,AL1 b){return AL1_(ASL1_(a)>>ASL1_(b));}
+ A_STATIC AU1 AShrSU1(AU1 a,AU1 b){return AU1_(ASU1_(a)>>ASU1_(b));}
+//------------------------------------------------------------------------------------------------------------------------------
+ #ifdef A_GCC
+  A_STATIC AD1 ASinD1(AD1 a){return __builtin_sin(a);}
+  A_STATIC AF1 ASinF1(AF1 a){return __builtin_sinf(a);}
+ #else
+  A_STATIC AD1 ASinD1(AD1 a){return sin(a);}
+  A_STATIC AF1 ASinF1(AF1 a){return sinf(a);}
+ #endif
+//------------------------------------------------------------------------------------------------------------------------------
+ #ifdef A_GCC
+  A_STATIC AD1 ASqrtD1(AD1 a){return __builtin_sqrt(a);}
+  A_STATIC AF1 ASqrtF1(AF1 a){return __builtin_sqrtf(a);}
+ #else
+  A_STATIC AD1 ASqrtD1(AD1 a){return sqrt(a);}
+  A_STATIC AF1 ASqrtF1(AF1 a){return sqrtf(a);}
+ #endif
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                               SCALAR RETURN OPS - DEPENDENT
+//==============================================================================================================================
+ A_STATIC AD1 AClampD1(AD1 x,AD1 n,AD1 m){return AMaxD1(n,AMinD1(x,m));}
+ A_STATIC AF1 AClampF1(AF1 x,AF1 n,AF1 m){return AMaxF1(n,AMinF1(x,m));}
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC AD1 AFractD1(AD1 a){return a-AFloorD1(a);}
+ A_STATIC AF1 AFractF1(AF1 a){return a-AFloorF1(a);}
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC AD1 APowD1(AD1 a,AD1 b){return AExp2D1(b*ALog2D1(a));}
+ A_STATIC AF1 APowF1(AF1 a,AF1 b){return AExp2F1(b*ALog2F1(a));}
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC AD1 ARsqD1(AD1 a){return ARcpD1(ASqrtD1(a));}
+ A_STATIC AF1 ARsqF1(AF1 a){return ARcpF1(ASqrtF1(a));}
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC AD1 ASatD1(AD1 a){return AMinD1(1.0,AMaxD1(0.0,a));}
+ A_STATIC AF1 ASatF1(AF1 a){return AMinF1(1.0f,AMaxF1(0.0f,a));}
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                         VECTOR OPS
+//------------------------------------------------------------------------------------------------------------------------------
+// These are added as needed for production or prototyping, so not necessarily a complete set.
+// They follow a convention of taking in a destination and also returning the destination value to increase utility.
+//==============================================================================================================================
+ A_STATIC retAD2 opAAbsD2(outAD2 d,inAD2 a){d[0]=AAbsD1(a[0]);d[1]=AAbsD1(a[1]);return d;}
+ A_STATIC retAD3 opAAbsD3(outAD3 d,inAD3 a){d[0]=AAbsD1(a[0]);d[1]=AAbsD1(a[1]);d[2]=AAbsD1(a[2]);return d;}
+ A_STATIC retAD4 opAAbsD4(outAD4 d,inAD4 a){d[0]=AAbsD1(a[0]);d[1]=AAbsD1(a[1]);d[2]=AAbsD1(a[2]);d[3]=AAbsD1(a[3]);return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC retAF2 opAAbsF2(outAF2 d,inAF2 a){d[0]=AAbsF1(a[0]);d[1]=AAbsF1(a[1]);return d;}
+ A_STATIC retAF3 opAAbsF3(outAF3 d,inAF3 a){d[0]=AAbsF1(a[0]);d[1]=AAbsF1(a[1]);d[2]=AAbsF1(a[2]);return d;}
+ A_STATIC retAF4 opAAbsF4(outAF4 d,inAF4 a){d[0]=AAbsF1(a[0]);d[1]=AAbsF1(a[1]);d[2]=AAbsF1(a[2]);d[3]=AAbsF1(a[3]);return d;}
+//==============================================================================================================================
+ A_STATIC retAD2 opAAddD2(outAD2 d,inAD2 a,inAD2 b){d[0]=a[0]+b[0];d[1]=a[1]+b[1];return d;}
+ A_STATIC retAD3 opAAddD3(outAD3 d,inAD3 a,inAD3 b){d[0]=a[0]+b[0];d[1]=a[1]+b[1];d[2]=a[2]+b[2];return d;}
+ A_STATIC retAD4 opAAddD4(outAD4 d,inAD4 a,inAD4 b){d[0]=a[0]+b[0];d[1]=a[1]+b[1];d[2]=a[2]+b[2];d[3]=a[3]+b[3];return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC retAF2 opAAddF2(outAF2 d,inAF2 a,inAF2 b){d[0]=a[0]+b[0];d[1]=a[1]+b[1];return d;}
+ A_STATIC retAF3 opAAddF3(outAF3 d,inAF3 a,inAF3 b){d[0]=a[0]+b[0];d[1]=a[1]+b[1];d[2]=a[2]+b[2];return d;}
+ A_STATIC retAF4 opAAddF4(outAF4 d,inAF4 a,inAF4 b){d[0]=a[0]+b[0];d[1]=a[1]+b[1];d[2]=a[2]+b[2];d[3]=a[3]+b[3];return d;}
+//==============================================================================================================================
+ A_STATIC retAD2 opAAddOneD2(outAD2 d,inAD2 a,AD1 b){d[0]=a[0]+b;d[1]=a[1]+b;return d;}
+ A_STATIC retAD3 opAAddOneD3(outAD3 d,inAD3 a,AD1 b){d[0]=a[0]+b;d[1]=a[1]+b;d[2]=a[2]+b;return d;}
+ A_STATIC retAD4 opAAddOneD4(outAD4 d,inAD4 a,AD1 b){d[0]=a[0]+b;d[1]=a[1]+b;d[2]=a[2]+b;d[3]=a[3]+b;return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC retAF2 opAAddOneF2(outAF2 d,inAF2 a,AF1 b){d[0]=a[0]+b;d[1]=a[1]+b;return d;}
+ A_STATIC retAF3 opAAddOneF3(outAF3 d,inAF3 a,AF1 b){d[0]=a[0]+b;d[1]=a[1]+b;d[2]=a[2]+b;return d;}
+ A_STATIC retAF4 opAAddOneF4(outAF4 d,inAF4 a,AF1 b){d[0]=a[0]+b;d[1]=a[1]+b;d[2]=a[2]+b;d[3]=a[3]+b;return d;}
+//==============================================================================================================================
+ A_STATIC retAD2 opACpyD2(outAD2 d,inAD2 a){d[0]=a[0];d[1]=a[1];return d;}
+ A_STATIC retAD3 opACpyD3(outAD3 d,inAD3 a){d[0]=a[0];d[1]=a[1];d[2]=a[2];return d;}
+ A_STATIC retAD4 opACpyD4(outAD4 d,inAD4 a){d[0]=a[0];d[1]=a[1];d[2]=a[2];d[3]=a[3];return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC retAF2 opACpyF2(outAF2 d,inAF2 a){d[0]=a[0];d[1]=a[1];return d;}
+ A_STATIC retAF3 opACpyF3(outAF3 d,inAF3 a){d[0]=a[0];d[1]=a[1];d[2]=a[2];return d;}
+ A_STATIC retAF4 opACpyF4(outAF4 d,inAF4 a){d[0]=a[0];d[1]=a[1];d[2]=a[2];d[3]=a[3];return d;}
+//==============================================================================================================================
+ A_STATIC retAD2 opALerpD2(outAD2 d,inAD2 a,inAD2 b,inAD2 c){d[0]=ALerpD1(a[0],b[0],c[0]);d[1]=ALerpD1(a[1],b[1],c[1]);return d;}
+ A_STATIC retAD3 opALerpD3(outAD3 d,inAD3 a,inAD3 b,inAD3 c){d[0]=ALerpD1(a[0],b[0],c[0]);d[1]=ALerpD1(a[1],b[1],c[1]);d[2]=ALerpD1(a[2],b[2],c[2]);return d;}
+ A_STATIC retAD4 opALerpD4(outAD4 d,inAD4 a,inAD4 b,inAD4 c){d[0]=ALerpD1(a[0],b[0],c[0]);d[1]=ALerpD1(a[1],b[1],c[1]);d[2]=ALerpD1(a[2],b[2],c[2]);d[3]=ALerpD1(a[3],b[3],c[3]);return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC retAF2 opALerpF2(outAF2 d,inAF2 a,inAF2 b,inAF2 c){d[0]=ALerpF1(a[0],b[0],c[0]);d[1]=ALerpF1(a[1],b[1],c[1]);return d;}
+ A_STATIC retAF3 opALerpF3(outAF3 d,inAF3 a,inAF3 b,inAF3 c){d[0]=ALerpF1(a[0],b[0],c[0]);d[1]=ALerpF1(a[1],b[1],c[1]);d[2]=ALerpF1(a[2],b[2],c[2]);return d;}
+ A_STATIC retAF4 opALerpF4(outAF4 d,inAF4 a,inAF4 b,inAF4 c){d[0]=ALerpF1(a[0],b[0],c[0]);d[1]=ALerpF1(a[1],b[1],c[1]);d[2]=ALerpF1(a[2],b[2],c[2]);d[3]=ALerpF1(a[3],b[3],c[3]);return d;}
+//==============================================================================================================================
+ A_STATIC retAD2 opALerpOneD2(outAD2 d,inAD2 a,inAD2 b,AD1 c){d[0]=ALerpD1(a[0],b[0],c);d[1]=ALerpD1(a[1],b[1],c);return d;}
+ A_STATIC retAD3 opALerpOneD3(outAD3 d,inAD3 a,inAD3 b,AD1 c){d[0]=ALerpD1(a[0],b[0],c);d[1]=ALerpD1(a[1],b[1],c);d[2]=ALerpD1(a[2],b[2],c);return d;}
+ A_STATIC retAD4 opALerpOneD4(outAD4 d,inAD4 a,inAD4 b,AD1 c){d[0]=ALerpD1(a[0],b[0],c);d[1]=ALerpD1(a[1],b[1],c);d[2]=ALerpD1(a[2],b[2],c);d[3]=ALerpD1(a[3],b[3],c);return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC retAF2 opALerpOneF2(outAF2 d,inAF2 a,inAF2 b,AF1 c){d[0]=ALerpF1(a[0],b[0],c);d[1]=ALerpF1(a[1],b[1],c);return d;}
+ A_STATIC retAF3 opALerpOneF3(outAF3 d,inAF3 a,inAF3 b,AF1 c){d[0]=ALerpF1(a[0],b[0],c);d[1]=ALerpF1(a[1],b[1],c);d[2]=ALerpF1(a[2],b[2],c);return d;}
+ A_STATIC retAF4 opALerpOneF4(outAF4 d,inAF4 a,inAF4 b,AF1 c){d[0]=ALerpF1(a[0],b[0],c);d[1]=ALerpF1(a[1],b[1],c);d[2]=ALerpF1(a[2],b[2],c);d[3]=ALerpF1(a[3],b[3],c);return d;}
+//==============================================================================================================================
+ A_STATIC retAD2 opAMaxD2(outAD2 d,inAD2 a,inAD2 b){d[0]=AMaxD1(a[0],b[0]);d[1]=AMaxD1(a[1],b[1]);return d;}
+ A_STATIC retAD3 opAMaxD3(outAD3 d,inAD3 a,inAD3 b){d[0]=AMaxD1(a[0],b[0]);d[1]=AMaxD1(a[1],b[1]);d[2]=AMaxD1(a[2],b[2]);return d;}
+ A_STATIC retAD4 opAMaxD4(outAD4 d,inAD4 a,inAD4 b){d[0]=AMaxD1(a[0],b[0]);d[1]=AMaxD1(a[1],b[1]);d[2]=AMaxD1(a[2],b[2]);d[3]=AMaxD1(a[3],b[3]);return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC retAF2 opAMaxF2(outAF2 d,inAF2 a,inAF2 b){d[0]=AMaxF1(a[0],b[0]);d[1]=AMaxF1(a[1],b[1]);return d;}
+ A_STATIC retAF3 opAMaxF3(outAF3 d,inAF3 a,inAF3 b){d[0]=AMaxF1(a[0],b[0]);d[1]=AMaxF1(a[1],b[1]);d[2]=AMaxF1(a[2],b[2]);return d;}
+ A_STATIC retAF4 opAMaxF4(outAF4 d,inAF4 a,inAF4 b){d[0]=AMaxF1(a[0],b[0]);d[1]=AMaxF1(a[1],b[1]);d[2]=AMaxF1(a[2],b[2]);d[3]=AMaxF1(a[3],b[3]);return d;}
+//==============================================================================================================================
+ A_STATIC retAD2 opAMinD2(outAD2 d,inAD2 a,inAD2 b){d[0]=AMinD1(a[0],b[0]);d[1]=AMinD1(a[1],b[1]);return d;}
+ A_STATIC retAD3 opAMinD3(outAD3 d,inAD3 a,inAD3 b){d[0]=AMinD1(a[0],b[0]);d[1]=AMinD1(a[1],b[1]);d[2]=AMinD1(a[2],b[2]);return d;}
+ A_STATIC retAD4 opAMinD4(outAD4 d,inAD4 a,inAD4 b){d[0]=AMinD1(a[0],b[0]);d[1]=AMinD1(a[1],b[1]);d[2]=AMinD1(a[2],b[2]);d[3]=AMinD1(a[3],b[3]);return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC retAF2 opAMinF2(outAF2 d,inAF2 a,inAF2 b){d[0]=AMinF1(a[0],b[0]);d[1]=AMinF1(a[1],b[1]);return d;}
+ A_STATIC retAF3 opAMinF3(outAF3 d,inAF3 a,inAF3 b){d[0]=AMinF1(a[0],b[0]);d[1]=AMinF1(a[1],b[1]);d[2]=AMinF1(a[2],b[2]);return d;}
+ A_STATIC retAF4 opAMinF4(outAF4 d,inAF4 a,inAF4 b){d[0]=AMinF1(a[0],b[0]);d[1]=AMinF1(a[1],b[1]);d[2]=AMinF1(a[2],b[2]);d[3]=AMinF1(a[3],b[3]);return d;}
+//==============================================================================================================================
+ A_STATIC retAD2 opAMulD2(outAD2 d,inAD2 a,inAD2 b){d[0]=a[0]*b[0];d[1]=a[1]*b[1];return d;}
+ A_STATIC retAD3 opAMulD3(outAD3 d,inAD3 a,inAD3 b){d[0]=a[0]*b[0];d[1]=a[1]*b[1];d[2]=a[2]*b[2];return d;}
+ A_STATIC retAD4 opAMulD4(outAD4 d,inAD4 a,inAD4 b){d[0]=a[0]*b[0];d[1]=a[1]*b[1];d[2]=a[2]*b[2];d[3]=a[3]*b[3];return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC retAF2 opAMulF2(outAF2 d,inAF2 a,inAF2 b){d[0]=a[0]*b[0];d[1]=a[1]*b[1];return d;}
+ A_STATIC retAF3 opAMulF3(outAF3 d,inAF3 a,inAF3 b){d[0]=a[0]*b[0];d[1]=a[1]*b[1];d[2]=a[2]*b[2];return d;}
+ A_STATIC retAF4 opAMulF4(outAF4 d,inAF4 a,inAF4 b){d[0]=a[0]*b[0];d[1]=a[1]*b[1];d[2]=a[2]*b[2];d[3]=a[3]*b[3];return d;}
+//==============================================================================================================================
+ A_STATIC retAD2 opAMulOneD2(outAD2 d,inAD2 a,AD1 b){d[0]=a[0]*b;d[1]=a[1]*b;return d;}
+ A_STATIC retAD3 opAMulOneD3(outAD3 d,inAD3 a,AD1 b){d[0]=a[0]*b;d[1]=a[1]*b;d[2]=a[2]*b;return d;}
+ A_STATIC retAD4 opAMulOneD4(outAD4 d,inAD4 a,AD1 b){d[0]=a[0]*b;d[1]=a[1]*b;d[2]=a[2]*b;d[3]=a[3]*b;return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC retAF2 opAMulOneF2(outAF2 d,inAF2 a,AF1 b){d[0]=a[0]*b;d[1]=a[1]*b;return d;}
+ A_STATIC retAF3 opAMulOneF3(outAF3 d,inAF3 a,AF1 b){d[0]=a[0]*b;d[1]=a[1]*b;d[2]=a[2]*b;return d;}
+ A_STATIC retAF4 opAMulOneF4(outAF4 d,inAF4 a,AF1 b){d[0]=a[0]*b;d[1]=a[1]*b;d[2]=a[2]*b;d[3]=a[3]*b;return d;}
+//==============================================================================================================================
+ A_STATIC retAD2 opANegD2(outAD2 d,inAD2 a){d[0]=-a[0];d[1]=-a[1];return d;}
+ A_STATIC retAD3 opANegD3(outAD3 d,inAD3 a){d[0]=-a[0];d[1]=-a[1];d[2]=-a[2];return d;}
+ A_STATIC retAD4 opANegD4(outAD4 d,inAD4 a){d[0]=-a[0];d[1]=-a[1];d[2]=-a[2];d[3]=-a[3];return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC retAF2 opANegF2(outAF2 d,inAF2 a){d[0]=-a[0];d[1]=-a[1];return d;}
+ A_STATIC retAF3 opANegF3(outAF3 d,inAF3 a){d[0]=-a[0];d[1]=-a[1];d[2]=-a[2];return d;}
+ A_STATIC retAF4 opANegF4(outAF4 d,inAF4 a){d[0]=-a[0];d[1]=-a[1];d[2]=-a[2];d[3]=-a[3];return d;}
+//==============================================================================================================================
+ A_STATIC retAD2 opARcpD2(outAD2 d,inAD2 a){d[0]=ARcpD1(a[0]);d[1]=ARcpD1(a[1]);return d;}
+ A_STATIC retAD3 opARcpD3(outAD3 d,inAD3 a){d[0]=ARcpD1(a[0]);d[1]=ARcpD1(a[1]);d[2]=ARcpD1(a[2]);return d;}
+ A_STATIC retAD4 opARcpD4(outAD4 d,inAD4 a){d[0]=ARcpD1(a[0]);d[1]=ARcpD1(a[1]);d[2]=ARcpD1(a[2]);d[3]=ARcpD1(a[3]);return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC retAF2 opARcpF2(outAF2 d,inAF2 a){d[0]=ARcpF1(a[0]);d[1]=ARcpF1(a[1]);return d;}
+ A_STATIC retAF3 opARcpF3(outAF3 d,inAF3 a){d[0]=ARcpF1(a[0]);d[1]=ARcpF1(a[1]);d[2]=ARcpF1(a[2]);return d;}
+ A_STATIC retAF4 opARcpF4(outAF4 d,inAF4 a){d[0]=ARcpF1(a[0]);d[1]=ARcpF1(a[1]);d[2]=ARcpF1(a[2]);d[3]=ARcpF1(a[3]);return d;}
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                     HALF FLOAT PACKING
+//==============================================================================================================================
+ // Convert float to half (in lower 16-bits of output).
+ // Same fast technique as documented here: ftp://ftp.fox-toolkit.org/pub/fasthalffloatconversion.pdf
+ // Supports denormals.
+ // Conversion rules are to make computations possibly "safer" on the GPU,
+ //  -INF & -NaN -> -65504
+ //  +INF & +NaN -> +65504
+ A_STATIC AU1 AU1_AH1_AF1(AF1 f){
+  static AW1 base[512]={
+   0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,
+   0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,
+   0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,
+   0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,
+   0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,
+   0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,
+   0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0001,0x0002,0x0004,0x0008,0x0010,0x0020,0x0040,0x0080,0x0100,
+   0x0200,0x0400,0x0800,0x0c00,0x1000,0x1400,0x1800,0x1c00,0x2000,0x2400,0x2800,0x2c00,0x3000,0x3400,0x3800,0x3c00,
+   0x4000,0x4400,0x4800,0x4c00,0x5000,0x5400,0x5800,0x5c00,0x6000,0x6400,0x6800,0x6c00,0x7000,0x7400,0x7800,0x7bff,
+   0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,
+   0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,
+   0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,
+   0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,
+   0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,
+   0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,
+   0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,
+   0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,
+   0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,
+   0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,
+   0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,
+   0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,
+   0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,
+   0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8001,0x8002,0x8004,0x8008,0x8010,0x8020,0x8040,0x8080,0x8100,
+   0x8200,0x8400,0x8800,0x8c00,0x9000,0x9400,0x9800,0x9c00,0xa000,0xa400,0xa800,0xac00,0xb000,0xb400,0xb800,0xbc00,
+   0xc000,0xc400,0xc800,0xcc00,0xd000,0xd400,0xd800,0xdc00,0xe000,0xe400,0xe800,0xec00,0xf000,0xf400,0xf800,0xfbff,
+   0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,
+   0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,
+   0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,
+   0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,
+   0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,
+   0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,
+   0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff};
+  static AB1 shift[512]={
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x17,0x16,0x15,0x14,0x13,0x12,0x11,0x10,0x0f,
+   0x0e,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,
+   0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x17,0x16,0x15,0x14,0x13,0x12,0x11,0x10,0x0f,
+   0x0e,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,
+   0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18};
+  union{AF1 f;AU1 u;}bits;bits.f=f;AU1 u=bits.u;AU1 i=u>>23;return (AU1)(base[i])+((u&0x7fffff)>>shift[i]);}
+//------------------------------------------------------------------------------------------------------------------------------
+ // Used to output packed constant.
+ A_STATIC AU1 AU1_AH2_AF2(inAF2 a){return AU1_AH1_AF1(a[0])+(AU1_AH1_AF1(a[1])<<16);}
+#endif
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//
+//
+//                                                            GLSL
+//
+//
+//==============================================================================================================================
+#if defined(A_GLSL) && defined(A_GPU)
+ #ifndef A_SKIP_EXT
+  #ifdef A_HALF
+   #extension GL_EXT_shader_16bit_storage:require
+   #extension GL_EXT_shader_explicit_arithmetic_types:require 
+  #endif
+//------------------------------------------------------------------------------------------------------------------------------
+  #ifdef A_LONG
+   #extension GL_ARB_gpu_shader_int64:require
+   #extension GL_NV_shader_atomic_int64:require
+  #endif
+//------------------------------------------------------------------------------------------------------------------------------
+  #ifdef A_WAVE
+   #extension GL_KHR_shader_subgroup_arithmetic:require
+   #extension GL_KHR_shader_subgroup_ballot:require
+   #extension GL_KHR_shader_subgroup_quad:require
+   #extension GL_KHR_shader_subgroup_shuffle:require
+  #endif
+ #endif
+//==============================================================================================================================
+ #define AP1 bool
+ #define AP2 bvec2
+ #define AP3 bvec3
+ #define AP4 bvec4
+//------------------------------------------------------------------------------------------------------------------------------
+ #define AF1 float
+ #define AF2 vec2
+ #define AF3 vec3
+ #define AF4 vec4
+//------------------------------------------------------------------------------------------------------------------------------
+ #define AU1 uint
+ #define AU2 uvec2
+ #define AU3 uvec3
+ #define AU4 uvec4
+//------------------------------------------------------------------------------------------------------------------------------
+ #define ASU1 int
+ #define ASU2 ivec2
+ #define ASU3 ivec3
+ #define ASU4 ivec4
+//==============================================================================================================================
+ #define AF1_AU1(x) uintBitsToFloat(AU1(x))
+ #define AF2_AU2(x) uintBitsToFloat(AU2(x))
+ #define AF3_AU3(x) uintBitsToFloat(AU3(x))
+ #define AF4_AU4(x) uintBitsToFloat(AU4(x))
+//------------------------------------------------------------------------------------------------------------------------------
+ #define AU1_AF1(x) floatBitsToUint(AF1(x))
+ #define AU2_AF2(x) floatBitsToUint(AF2(x))
+ #define AU3_AF3(x) floatBitsToUint(AF3(x))
+ #define AU4_AF4(x) floatBitsToUint(AF4(x))
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 AU1_AH1_AF1_x(AF1 a){return packHalf2x16(AF2(a,0.0));}
+ #define AU1_AH1_AF1(a) AU1_AH1_AF1_x(AF1(a))
+//------------------------------------------------------------------------------------------------------------------------------
+ #define AU1_AH2_AF2 packHalf2x16
+ #define AU1_AW2Unorm_AF2 packUnorm2x16
+ #define AU1_AB4Unorm_AF4 packUnorm4x8
+//------------------------------------------------------------------------------------------------------------------------------
+ #define AF2_AH2_AU1 unpackHalf2x16
+ #define AF2_AW2Unorm_AU1 unpackUnorm2x16
+ #define AF4_AB4Unorm_AU1 unpackUnorm4x8
+//==============================================================================================================================
+ AF1 AF1_x(AF1 a){return AF1(a);}
+ AF2 AF2_x(AF1 a){return AF2(a,a);}
+ AF3 AF3_x(AF1 a){return AF3(a,a,a);}
+ AF4 AF4_x(AF1 a){return AF4(a,a,a,a);}
+ #define AF1_(a) AF1_x(AF1(a))
+ #define AF2_(a) AF2_x(AF1(a))
+ #define AF3_(a) AF3_x(AF1(a))
+ #define AF4_(a) AF4_x(AF1(a))
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 AU1_x(AU1 a){return AU1(a);}
+ AU2 AU2_x(AU1 a){return AU2(a,a);}
+ AU3 AU3_x(AU1 a){return AU3(a,a,a);}
+ AU4 AU4_x(AU1 a){return AU4(a,a,a,a);}
+ #define AU1_(a) AU1_x(AU1(a))
+ #define AU2_(a) AU2_x(AU1(a))
+ #define AU3_(a) AU3_x(AU1(a))
+ #define AU4_(a) AU4_x(AU1(a))
+//==============================================================================================================================
+ AU1 AAbsSU1(AU1 a){return AU1(abs(ASU1(a)));}
+ AU2 AAbsSU2(AU2 a){return AU2(abs(ASU2(a)));}
+ AU3 AAbsSU3(AU3 a){return AU3(abs(ASU3(a)));}
+ AU4 AAbsSU4(AU4 a){return AU4(abs(ASU4(a)));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 ABfe(AU1 src,AU1 off,AU1 bits){return bitfieldExtract(src,ASU1(off),ASU1(bits));}
+ AU1 ABfi(AU1 src,AU1 ins,AU1 mask){return (ins&mask)|(src&(~mask));}
+ // Proxy for V_BFI_B32 where the 'mask' is set as 'bits', 'mask=(1<<bits)-1', and 'bits' needs to be an immediate.
+ AU1 ABfiM(AU1 src,AU1 ins,AU1 bits){return bitfieldInsert(src,ins,0,ASU1(bits));}
+//------------------------------------------------------------------------------------------------------------------------------
+ // V_MED3_F32.
+ AF1 AClampF1(AF1 x,AF1 n,AF1 m){return clamp(x,n,m);}
+ AF2 AClampF2(AF2 x,AF2 n,AF2 m){return clamp(x,n,m);}
+ AF3 AClampF3(AF3 x,AF3 n,AF3 m){return clamp(x,n,m);}
+ AF4 AClampF4(AF4 x,AF4 n,AF4 m){return clamp(x,n,m);}
+//------------------------------------------------------------------------------------------------------------------------------
+ // V_FRACT_F32 (note DX frac() is different).
+ AF1 AFractF1(AF1 x){return fract(x);}
+ AF2 AFractF2(AF2 x){return fract(x);}
+ AF3 AFractF3(AF3 x){return fract(x);}
+ AF4 AFractF4(AF4 x){return fract(x);}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF1 ALerpF1(AF1 x,AF1 y,AF1 a){return mix(x,y,a);}
+ AF2 ALerpF2(AF2 x,AF2 y,AF2 a){return mix(x,y,a);}
+ AF3 ALerpF3(AF3 x,AF3 y,AF3 a){return mix(x,y,a);}
+ AF4 ALerpF4(AF4 x,AF4 y,AF4 a){return mix(x,y,a);}
+//------------------------------------------------------------------------------------------------------------------------------
+ // V_MAX3_F32.
+ AF1 AMax3F1(AF1 x,AF1 y,AF1 z){return max(x,max(y,z));}
+ AF2 AMax3F2(AF2 x,AF2 y,AF2 z){return max(x,max(y,z));}
+ AF3 AMax3F3(AF3 x,AF3 y,AF3 z){return max(x,max(y,z));}
+ AF4 AMax3F4(AF4 x,AF4 y,AF4 z){return max(x,max(y,z));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 AMax3SU1(AU1 x,AU1 y,AU1 z){return AU1(max(ASU1(x),max(ASU1(y),ASU1(z))));}
+ AU2 AMax3SU2(AU2 x,AU2 y,AU2 z){return AU2(max(ASU2(x),max(ASU2(y),ASU2(z))));}
+ AU3 AMax3SU3(AU3 x,AU3 y,AU3 z){return AU3(max(ASU3(x),max(ASU3(y),ASU3(z))));}
+ AU4 AMax3SU4(AU4 x,AU4 y,AU4 z){return AU4(max(ASU4(x),max(ASU4(y),ASU4(z))));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 AMax3U1(AU1 x,AU1 y,AU1 z){return max(x,max(y,z));}
+ AU2 AMax3U2(AU2 x,AU2 y,AU2 z){return max(x,max(y,z));}
+ AU3 AMax3U3(AU3 x,AU3 y,AU3 z){return max(x,max(y,z));}
+ AU4 AMax3U4(AU4 x,AU4 y,AU4 z){return max(x,max(y,z));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 AMaxSU1(AU1 a,AU1 b){return AU1(max(ASU1(a),ASU1(b)));}
+ AU2 AMaxSU2(AU2 a,AU2 b){return AU2(max(ASU2(a),ASU2(b)));}
+ AU3 AMaxSU3(AU3 a,AU3 b){return AU3(max(ASU3(a),ASU3(b)));}
+ AU4 AMaxSU4(AU4 a,AU4 b){return AU4(max(ASU4(a),ASU4(b)));}
+//------------------------------------------------------------------------------------------------------------------------------
+ // Clamp has an easier pattern match for med3 when some ordering is known.
+ // V_MED3_F32.
+ AF1 AMed3F1(AF1 x,AF1 y,AF1 z){return max(min(x,y),min(max(x,y),z));}
+ AF2 AMed3F2(AF2 x,AF2 y,AF2 z){return max(min(x,y),min(max(x,y),z));}
+ AF3 AMed3F3(AF3 x,AF3 y,AF3 z){return max(min(x,y),min(max(x,y),z));}
+ AF4 AMed3F4(AF4 x,AF4 y,AF4 z){return max(min(x,y),min(max(x,y),z));}
+//------------------------------------------------------------------------------------------------------------------------------
+ // V_MIN3_F32.
+ AF1 AMin3F1(AF1 x,AF1 y,AF1 z){return min(x,min(y,z));}
+ AF2 AMin3F2(AF2 x,AF2 y,AF2 z){return min(x,min(y,z));}
+ AF3 AMin3F3(AF3 x,AF3 y,AF3 z){return min(x,min(y,z));}
+ AF4 AMin3F4(AF4 x,AF4 y,AF4 z){return min(x,min(y,z));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 AMin3SU1(AU1 x,AU1 y,AU1 z){return AU1(min(ASU1(x),min(ASU1(y),ASU1(z))));}
+ AU2 AMin3SU2(AU2 x,AU2 y,AU2 z){return AU2(min(ASU2(x),min(ASU2(y),ASU2(z))));}
+ AU3 AMin3SU3(AU3 x,AU3 y,AU3 z){return AU3(min(ASU3(x),min(ASU3(y),ASU3(z))));}
+ AU4 AMin3SU4(AU4 x,AU4 y,AU4 z){return AU4(min(ASU4(x),min(ASU4(y),ASU4(z))));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 AMin3U1(AU1 x,AU1 y,AU1 z){return min(x,min(y,z));}
+ AU2 AMin3U2(AU2 x,AU2 y,AU2 z){return min(x,min(y,z));}
+ AU3 AMin3U3(AU3 x,AU3 y,AU3 z){return min(x,min(y,z));}
+ AU4 AMin3U4(AU4 x,AU4 y,AU4 z){return min(x,min(y,z));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 AMinSU1(AU1 a,AU1 b){return AU1(min(ASU1(a),ASU1(b)));}
+ AU2 AMinSU2(AU2 a,AU2 b){return AU2(min(ASU2(a),ASU2(b)));}
+ AU3 AMinSU3(AU3 a,AU3 b){return AU3(min(ASU3(a),ASU3(b)));}
+ AU4 AMinSU4(AU4 a,AU4 b){return AU4(min(ASU4(a),ASU4(b)));}
+//------------------------------------------------------------------------------------------------------------------------------
+ // Normalized trig. Valid input domain is {-256 to +256}. No GLSL compiler intrinsic exists to map to this currently.
+ // V_COS_F32.
+ AF1 ANCosF1(AF1 x){return cos(x*AF1_(A_2PI));}
+ AF2 ANCosF2(AF2 x){return cos(x*AF2_(A_2PI));}
+ AF3 ANCosF3(AF3 x){return cos(x*AF3_(A_2PI));}
+ AF4 ANCosF4(AF4 x){return cos(x*AF4_(A_2PI));}
+//------------------------------------------------------------------------------------------------------------------------------
+ // Normalized trig. Valid input domain is {-256 to +256}. No GLSL compiler intrinsic exists to map to this currently.
+ // V_SIN_F32.
+ AF1 ANSinF1(AF1 x){return sin(x*AF1_(A_2PI));}
+ AF2 ANSinF2(AF2 x){return sin(x*AF2_(A_2PI));}
+ AF3 ANSinF3(AF3 x){return sin(x*AF3_(A_2PI));}
+ AF4 ANSinF4(AF4 x){return sin(x*AF4_(A_2PI));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF1 ARcpF1(AF1 x){return AF1_(1.0)/x;}
+ AF2 ARcpF2(AF2 x){return AF2_(1.0)/x;}
+ AF3 ARcpF3(AF3 x){return AF3_(1.0)/x;}
+ AF4 ARcpF4(AF4 x){return AF4_(1.0)/x;}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF1 ARsqF1(AF1 x){return AF1_(1.0)/sqrt(x);}
+ AF2 ARsqF2(AF2 x){return AF2_(1.0)/sqrt(x);}
+ AF3 ARsqF3(AF3 x){return AF3_(1.0)/sqrt(x);}
+ AF4 ARsqF4(AF4 x){return AF4_(1.0)/sqrt(x);}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF1 ASatF1(AF1 x){return clamp(x,AF1_(0.0),AF1_(1.0));}
+ AF2 ASatF2(AF2 x){return clamp(x,AF2_(0.0),AF2_(1.0));}
+ AF3 ASatF3(AF3 x){return clamp(x,AF3_(0.0),AF3_(1.0));}
+ AF4 ASatF4(AF4 x){return clamp(x,AF4_(0.0),AF4_(1.0));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 AShrSU1(AU1 a,AU1 b){return AU1(ASU1(a)>>ASU1(b));}
+ AU2 AShrSU2(AU2 a,AU2 b){return AU2(ASU2(a)>>ASU2(b));}
+ AU3 AShrSU3(AU3 a,AU3 b){return AU3(ASU3(a)>>ASU3(b));}
+ AU4 AShrSU4(AU4 a,AU4 b){return AU4(ASU4(a)>>ASU4(b));}
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                          GLSL BYTE
+//==============================================================================================================================
+ #ifdef A_BYTE
+  #define AB1 uint8_t
+  #define AB2 u8vec2
+  #define AB3 u8vec3
+  #define AB4 u8vec4
+//------------------------------------------------------------------------------------------------------------------------------
+  #define ASB1 int8_t
+  #define ASB2 i8vec2
+  #define ASB3 i8vec3
+  #define ASB4 i8vec4
+//------------------------------------------------------------------------------------------------------------------------------
+  AB1 AB1_x(AB1 a){return AB1(a);}
+  AB2 AB2_x(AB1 a){return AB2(a,a);}
+  AB3 AB3_x(AB1 a){return AB3(a,a,a);}
+  AB4 AB4_x(AB1 a){return AB4(a,a,a,a);}
+  #define AB1_(a) AB1_x(AB1(a))
+  #define AB2_(a) AB2_x(AB1(a))
+  #define AB3_(a) AB3_x(AB1(a))
+  #define AB4_(a) AB4_x(AB1(a))
+ #endif
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                          GLSL HALF
+//==============================================================================================================================
+ #ifdef A_HALF
+  #define AH1 float16_t
+  #define AH2 f16vec2
+  #define AH3 f16vec3
+  #define AH4 f16vec4
+//------------------------------------------------------------------------------------------------------------------------------
+  #define AW1 uint16_t
+  #define AW2 u16vec2
+  #define AW3 u16vec3
+  #define AW4 u16vec4
+//------------------------------------------------------------------------------------------------------------------------------
+  #define ASW1 int16_t
+  #define ASW2 i16vec2
+  #define ASW3 i16vec3
+  #define ASW4 i16vec4
+//==============================================================================================================================
+  #define AH2_AU1(x) unpackFloat2x16(AU1(x))
+  AH4 AH4_AU2_x(AU2 x){return AH4(unpackFloat2x16(x.x),unpackFloat2x16(x.y));}
+  #define AH4_AU2(x) AH4_AU2_x(AU2(x))
+  #define AW2_AU1(x) unpackUint2x16(AU1(x))
+  #define AW4_AU2(x) unpackUint4x16(pack64(AU2(x)))
+//------------------------------------------------------------------------------------------------------------------------------
+  #define AU1_AH2(x) packFloat2x16(AH2(x))
+  AU2 AU2_AH4_x(AH4 x){return AU2(packFloat2x16(x.xy),packFloat2x16(x.zw));}
+  #define AU2_AH4(x) AU2_AH4_x(AH4(x))
+  #define AU1_AW2(x) packUint2x16(AW2(x))
+  #define AU2_AW4(x) unpack32(packUint4x16(AW4(x)))
+//==============================================================================================================================
+  #define AW1_AH1(x) halfBitsToUint16(AH1(x))
+  #define AW2_AH2(x) halfBitsToUint16(AH2(x))
+  #define AW3_AH3(x) halfBitsToUint16(AH3(x))
+  #define AW4_AH4(x) halfBitsToUint16(AH4(x))
+//------------------------------------------------------------------------------------------------------------------------------
+  #define AH1_AW1(x) uint16BitsToHalf(AW1(x))
+  #define AH2_AW2(x) uint16BitsToHalf(AW2(x))
+  #define AH3_AW3(x) uint16BitsToHalf(AW3(x))
+  #define AH4_AW4(x) uint16BitsToHalf(AW4(x))
+//==============================================================================================================================
+  AH1 AH1_x(AH1 a){return AH1(a);}
+  AH2 AH2_x(AH1 a){return AH2(a,a);}
+  AH3 AH3_x(AH1 a){return AH3(a,a,a);}
+  AH4 AH4_x(AH1 a){return AH4(a,a,a,a);}
+  #define AH1_(a) AH1_x(AH1(a))
+  #define AH2_(a) AH2_x(AH1(a))
+  #define AH3_(a) AH3_x(AH1(a))
+  #define AH4_(a) AH4_x(AH1(a))
+//------------------------------------------------------------------------------------------------------------------------------
+  AW1 AW1_x(AW1 a){return AW1(a);}
+  AW2 AW2_x(AW1 a){return AW2(a,a);}
+  AW3 AW3_x(AW1 a){return AW3(a,a,a);}
+  AW4 AW4_x(AW1 a){return AW4(a,a,a,a);}
+  #define AW1_(a) AW1_x(AW1(a))
+  #define AW2_(a) AW2_x(AW1(a))
+  #define AW3_(a) AW3_x(AW1(a))
+  #define AW4_(a) AW4_x(AW1(a))
+//==============================================================================================================================
+  AW1 AAbsSW1(AW1 a){return AW1(abs(ASW1(a)));}
+  AW2 AAbsSW2(AW2 a){return AW2(abs(ASW2(a)));}
+  AW3 AAbsSW3(AW3 a){return AW3(abs(ASW3(a)));}
+  AW4 AAbsSW4(AW4 a){return AW4(abs(ASW4(a)));}
+//------------------------------------------------------------------------------------------------------------------------------
+  AH1 AClampH1(AH1 x,AH1 n,AH1 m){return clamp(x,n,m);}
+  AH2 AClampH2(AH2 x,AH2 n,AH2 m){return clamp(x,n,m);}
+  AH3 AClampH3(AH3 x,AH3 n,AH3 m){return clamp(x,n,m);}
+  AH4 AClampH4(AH4 x,AH4 n,AH4 m){return clamp(x,n,m);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AH1 AFractH1(AH1 x){return fract(x);}
+  AH2 AFractH2(AH2 x){return fract(x);}
+  AH3 AFractH3(AH3 x){return fract(x);}
+  AH4 AFractH4(AH4 x){return fract(x);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AH1 ALerpH1(AH1 x,AH1 y,AH1 a){return mix(x,y,a);}
+  AH2 ALerpH2(AH2 x,AH2 y,AH2 a){return mix(x,y,a);}
+  AH3 ALerpH3(AH3 x,AH3 y,AH3 a){return mix(x,y,a);}
+  AH4 ALerpH4(AH4 x,AH4 y,AH4 a){return mix(x,y,a);}
+//------------------------------------------------------------------------------------------------------------------------------
+  // No packed version of max3.
+  AH1 AMax3H1(AH1 x,AH1 y,AH1 z){return max(x,max(y,z));}
+  AH2 AMax3H2(AH2 x,AH2 y,AH2 z){return max(x,max(y,z));}
+  AH3 AMax3H3(AH3 x,AH3 y,AH3 z){return max(x,max(y,z));}
+  AH4 AMax3H4(AH4 x,AH4 y,AH4 z){return max(x,max(y,z));}
+//------------------------------------------------------------------------------------------------------------------------------
+  AW1 AMaxSW1(AW1 a,AW1 b){return AW1(max(ASU1(a),ASU1(b)));}
+  AW2 AMaxSW2(AW2 a,AW2 b){return AW2(max(ASU2(a),ASU2(b)));}
+  AW3 AMaxSW3(AW3 a,AW3 b){return AW3(max(ASU3(a),ASU3(b)));}
+  AW4 AMaxSW4(AW4 a,AW4 b){return AW4(max(ASU4(a),ASU4(b)));}
+//------------------------------------------------------------------------------------------------------------------------------
+  // No packed version of min3.
+  AH1 AMin3H1(AH1 x,AH1 y,AH1 z){return min(x,min(y,z));}
+  AH2 AMin3H2(AH2 x,AH2 y,AH2 z){return min(x,min(y,z));}
+  AH3 AMin3H3(AH3 x,AH3 y,AH3 z){return min(x,min(y,z));}
+  AH4 AMin3H4(AH4 x,AH4 y,AH4 z){return min(x,min(y,z));}
+//------------------------------------------------------------------------------------------------------------------------------
+  AW1 AMinSW1(AW1 a,AW1 b){return AW1(min(ASU1(a),ASU1(b)));}
+  AW2 AMinSW2(AW2 a,AW2 b){return AW2(min(ASU2(a),ASU2(b)));}
+  AW3 AMinSW3(AW3 a,AW3 b){return AW3(min(ASU3(a),ASU3(b)));}
+  AW4 AMinSW4(AW4 a,AW4 b){return AW4(min(ASU4(a),ASU4(b)));}
+//------------------------------------------------------------------------------------------------------------------------------
+  AH1 ARcpH1(AH1 x){return AH1_(1.0)/x;}
+  AH2 ARcpH2(AH2 x){return AH2_(1.0)/x;}
+  AH3 ARcpH3(AH3 x){return AH3_(1.0)/x;}
+  AH4 ARcpH4(AH4 x){return AH4_(1.0)/x;}
+//------------------------------------------------------------------------------------------------------------------------------
+  AH1 ARsqH1(AH1 x){return AH1_(1.0)/sqrt(x);}
+  AH2 ARsqH2(AH2 x){return AH2_(1.0)/sqrt(x);}
+  AH3 ARsqH3(AH3 x){return AH3_(1.0)/sqrt(x);}
+  AH4 ARsqH4(AH4 x){return AH4_(1.0)/sqrt(x);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AH1 ASatH1(AH1 x){return clamp(x,AH1_(0.0),AH1_(1.0));}
+  AH2 ASatH2(AH2 x){return clamp(x,AH2_(0.0),AH2_(1.0));}
+  AH3 ASatH3(AH3 x){return clamp(x,AH3_(0.0),AH3_(1.0));}
+  AH4 ASatH4(AH4 x){return clamp(x,AH4_(0.0),AH4_(1.0));}
+//------------------------------------------------------------------------------------------------------------------------------
+  AW1 AShrSW1(AW1 a,AW1 b){return AW1(ASW1(a)>>ASW1(b));}
+  AW2 AShrSW2(AW2 a,AW2 b){return AW2(ASW2(a)>>ASW2(b));}
+  AW3 AShrSW3(AW3 a,AW3 b){return AW3(ASW3(a)>>ASW3(b));}
+  AW4 AShrSW4(AW4 a,AW4 b){return AW4(ASW4(a)>>ASW4(b));}
+ #endif
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                         GLSL DOUBLE
+//==============================================================================================================================
+ #ifdef A_DUBL
+  #define AD1 double
+  #define AD2 dvec2
+  #define AD3 dvec3
+  #define AD4 dvec4
+//------------------------------------------------------------------------------------------------------------------------------
+  AD1 AD1_x(AD1 a){return AD1(a);}
+  AD2 AD2_x(AD1 a){return AD2(a,a);}
+  AD3 AD3_x(AD1 a){return AD3(a,a,a);}
+  AD4 AD4_x(AD1 a){return AD4(a,a,a,a);}
+  #define AD1_(a) AD1_x(AD1(a))
+  #define AD2_(a) AD2_x(AD1(a))
+  #define AD3_(a) AD3_x(AD1(a))
+  #define AD4_(a) AD4_x(AD1(a))
+//==============================================================================================================================
+  AD1 AFractD1(AD1 x){return fract(x);}
+  AD2 AFractD2(AD2 x){return fract(x);}
+  AD3 AFractD3(AD3 x){return fract(x);}
+  AD4 AFractD4(AD4 x){return fract(x);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AD1 ALerpD1(AD1 x,AD1 y,AD1 a){return mix(x,y,a);}
+  AD2 ALerpD2(AD2 x,AD2 y,AD2 a){return mix(x,y,a);}
+  AD3 ALerpD3(AD3 x,AD3 y,AD3 a){return mix(x,y,a);}
+  AD4 ALerpD4(AD4 x,AD4 y,AD4 a){return mix(x,y,a);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AD1 ARcpD1(AD1 x){return AD1_(1.0)/x;}
+  AD2 ARcpD2(AD2 x){return AD2_(1.0)/x;}
+  AD3 ARcpD3(AD3 x){return AD3_(1.0)/x;}
+  AD4 ARcpD4(AD4 x){return AD4_(1.0)/x;}
+//------------------------------------------------------------------------------------------------------------------------------
+  AD1 ARsqD1(AD1 x){return AD1_(1.0)/sqrt(x);}
+  AD2 ARsqD2(AD2 x){return AD2_(1.0)/sqrt(x);}
+  AD3 ARsqD3(AD3 x){return AD3_(1.0)/sqrt(x);}
+  AD4 ARsqD4(AD4 x){return AD4_(1.0)/sqrt(x);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AD1 ASatD1(AD1 x){return clamp(x,AD1_(0.0),AD1_(1.0));}
+  AD2 ASatD2(AD2 x){return clamp(x,AD2_(0.0),AD2_(1.0));}
+  AD3 ASatD3(AD3 x){return clamp(x,AD3_(0.0),AD3_(1.0));}
+  AD4 ASatD4(AD4 x){return clamp(x,AD4_(0.0),AD4_(1.0));}
+ #endif
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                         GLSL LONG
+//==============================================================================================================================
+ #ifdef A_LONG
+  #define AL1 uint64_t
+  #define AL2 u64vec2
+  #define AL3 u64vec3
+  #define AL4 u64vec4
+//------------------------------------------------------------------------------------------------------------------------------
+  #define ASL1 int64_t
+  #define ASL2 i64vec2
+  #define ASL3 i64vec3
+  #define ASL4 i64vec4
+//------------------------------------------------------------------------------------------------------------------------------
+  #define AL1_AU2(x) packUint2x32(AU2(x))
+  #define AU2_AL1(x) unpackUint2x32(AL1(x))
+//------------------------------------------------------------------------------------------------------------------------------
+  AL1 AL1_x(AL1 a){return AL1(a);}
+  AL2 AL2_x(AL1 a){return AL2(a,a);}
+  AL3 AL3_x(AL1 a){return AL3(a,a,a);}
+  AL4 AL4_x(AL1 a){return AL4(a,a,a,a);}
+  #define AL1_(a) AL1_x(AL1(a))
+  #define AL2_(a) AL2_x(AL1(a))
+  #define AL3_(a) AL3_x(AL1(a))
+  #define AL4_(a) AL4_x(AL1(a))
+//==============================================================================================================================
+  AL1 AAbsSL1(AL1 a){return AL1(abs(ASL1(a)));}
+  AL2 AAbsSL2(AL2 a){return AL2(abs(ASL2(a)));}
+  AL3 AAbsSL3(AL3 a){return AL3(abs(ASL3(a)));}
+  AL4 AAbsSL4(AL4 a){return AL4(abs(ASL4(a)));}
+//------------------------------------------------------------------------------------------------------------------------------
+  AL1 AMaxSL1(AL1 a,AL1 b){return AL1(max(ASU1(a),ASU1(b)));}
+  AL2 AMaxSL2(AL2 a,AL2 b){return AL2(max(ASU2(a),ASU2(b)));}
+  AL3 AMaxSL3(AL3 a,AL3 b){return AL3(max(ASU3(a),ASU3(b)));}
+  AL4 AMaxSL4(AL4 a,AL4 b){return AL4(max(ASU4(a),ASU4(b)));}
+//------------------------------------------------------------------------------------------------------------------------------
+  AL1 AMinSL1(AL1 a,AL1 b){return AL1(min(ASU1(a),ASU1(b)));}
+  AL2 AMinSL2(AL2 a,AL2 b){return AL2(min(ASU2(a),ASU2(b)));}
+  AL3 AMinSL3(AL3 a,AL3 b){return AL3(min(ASU3(a),ASU3(b)));}
+  AL4 AMinSL4(AL4 a,AL4 b){return AL4(min(ASU4(a),ASU4(b)));}
+ #endif
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                      WAVE OPERATIONS
+//==============================================================================================================================
+ #ifdef A_WAVE
+  // Where 'x' must be a compile time literal.
+  AF1 AWaveXorF1(AF1 v,AU1 x){return subgroupShuffleXor(v,x);}
+  AF2 AWaveXorF2(AF2 v,AU1 x){return subgroupShuffleXor(v,x);}
+  AF3 AWaveXorF3(AF3 v,AU1 x){return subgroupShuffleXor(v,x);}
+  AF4 AWaveXorF4(AF4 v,AU1 x){return subgroupShuffleXor(v,x);}
+  AU1 AWaveXorU1(AU1 v,AU1 x){return subgroupShuffleXor(v,x);}
+  AU2 AWaveXorU2(AU2 v,AU1 x){return subgroupShuffleXor(v,x);}
+  AU3 AWaveXorU3(AU3 v,AU1 x){return subgroupShuffleXor(v,x);}
+  AU4 AWaveXorU4(AU4 v,AU1 x){return subgroupShuffleXor(v,x);}
+//------------------------------------------------------------------------------------------------------------------------------
+  #ifdef A_HALF
+   AH2 AWaveXorH2(AH2 v,AU1 x){return AH2_AU1(subgroupShuffleXor(AU1_AH2(v),x));}
+   AH4 AWaveXorH4(AH4 v,AU1 x){return AH4_AU2(subgroupShuffleXor(AU2_AH4(v),x));}
+   AW2 AWaveXorW2(AW2 v,AU1 x){return AW2_AU1(subgroupShuffleXor(AU1_AW2(v),x));}
+   AW4 AWaveXorW4(AW4 v,AU1 x){return AW4_AU2(subgroupShuffleXor(AU2_AW4(v),x));}
+  #endif
+ #endif
+//==============================================================================================================================
+#endif
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//
+//
+//                                                            HLSL
+//
+//
+//==============================================================================================================================
+#if defined(A_HLSL) && defined(A_GPU)
+ #ifdef A_HLSL_6_2
+  #define AP1 bool
+  #define AP2 bool2
+  #define AP3 bool3
+  #define AP4 bool4
+//------------------------------------------------------------------------------------------------------------------------------
+  #define AF1 float32_t
+  #define AF2 float32_t2
+  #define AF3 float32_t3
+  #define AF4 float32_t4
+//------------------------------------------------------------------------------------------------------------------------------
+  #define AU1 uint32_t
+  #define AU2 uint32_t2
+  #define AU3 uint32_t3
+  #define AU4 uint32_t4
+//------------------------------------------------------------------------------------------------------------------------------
+  #define ASU1 int32_t
+  #define ASU2 int32_t2
+  #define ASU3 int32_t3
+  #define ASU4 int32_t4
+ #else
+  #define AP1 bool
+  #define AP2 bool2
+  #define AP3 bool3
+  #define AP4 bool4
+//------------------------------------------------------------------------------------------------------------------------------
+  #define AF1 float
+  #define AF2 float2
+  #define AF3 float3
+  #define AF4 float4
+//------------------------------------------------------------------------------------------------------------------------------
+  #define AU1 uint
+  #define AU2 uint2
+  #define AU3 uint3
+  #define AU4 uint4
+//------------------------------------------------------------------------------------------------------------------------------
+  #define ASU1 int
+  #define ASU2 int2
+  #define ASU3 int3
+  #define ASU4 int4
+ #endif
+//==============================================================================================================================
+ #define AF1_AU1(x) asfloat(AU1(x))
+ #define AF2_AU2(x) asfloat(AU2(x))
+ #define AF3_AU3(x) asfloat(AU3(x))
+ #define AF4_AU4(x) asfloat(AU4(x))
+//------------------------------------------------------------------------------------------------------------------------------
+ #define AU1_AF1(x) asuint(AF1(x))
+ #define AU2_AF2(x) asuint(AF2(x))
+ #define AU3_AF3(x) asuint(AF3(x))
+ #define AU4_AF4(x) asuint(AF4(x))
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 AU1_AH1_AF1_x(AF1 a){return f32tof16(a);}
+ #define AU1_AH1_AF1(a) AU1_AH1_AF1_x(AF1(a))
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 AU1_AH2_AF2_x(AF2 a){return f32tof16(a.x)|(f32tof16(a.y)<<16);}
+ #define AU1_AH2_AF2(a) AU1_AH2_AF2_x(AF2(a)) 
+ #define AU1_AB4Unorm_AF4(x) D3DCOLORtoUBYTE4(AF4(x))
+//------------------------------------------------------------------------------------------------------------------------------
+ AF2 AF2_AH2_AU1_x(AU1 x){return AF2(f16tof32(x&0xFFFF),f16tof32(x>>16));}
+ #define AF2_AH2_AU1(x) AF2_AH2_AU1_x(AU1(x))
+//==============================================================================================================================
+ AF1 AF1_x(AF1 a){return AF1(a);}
+ AF2 AF2_x(AF1 a){return AF2(a,a);}
+ AF3 AF3_x(AF1 a){return AF3(a,a,a);}
+ AF4 AF4_x(AF1 a){return AF4(a,a,a,a);}
+ #define AF1_(a) AF1_x(AF1(a))
+ #define AF2_(a) AF2_x(AF1(a))
+ #define AF3_(a) AF3_x(AF1(a))
+ #define AF4_(a) AF4_x(AF1(a))
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 AU1_x(AU1 a){return AU1(a);}
+ AU2 AU2_x(AU1 a){return AU2(a,a);}
+ AU3 AU3_x(AU1 a){return AU3(a,a,a);}
+ AU4 AU4_x(AU1 a){return AU4(a,a,a,a);}
+ #define AU1_(a) AU1_x(AU1(a))
+ #define AU2_(a) AU2_x(AU1(a))
+ #define AU3_(a) AU3_x(AU1(a))
+ #define AU4_(a) AU4_x(AU1(a))
+//==============================================================================================================================
+ AU1 AAbsSU1(AU1 a){return AU1(abs(ASU1(a)));}
+ AU2 AAbsSU2(AU2 a){return AU2(abs(ASU2(a)));}
+ AU3 AAbsSU3(AU3 a){return AU3(abs(ASU3(a)));}
+ AU4 AAbsSU4(AU4 a){return AU4(abs(ASU4(a)));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 ABfe(AU1 src,AU1 off,AU1 bits){AU1 mask=(1u<<bits)-1;return (src>>off)&mask;}
+ AU1 ABfi(AU1 src,AU1 ins,AU1 mask){return (ins&mask)|(src&(~mask));}
+ AU1 ABfiM(AU1 src,AU1 ins,AU1 bits){AU1 mask=(1u<<bits)-1;return (ins&mask)|(src&(~mask));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF1 AClampF1(AF1 x,AF1 n,AF1 m){return max(n,min(x,m));}
+ AF2 AClampF2(AF2 x,AF2 n,AF2 m){return max(n,min(x,m));}
+ AF3 AClampF3(AF3 x,AF3 n,AF3 m){return max(n,min(x,m));}
+ AF4 AClampF4(AF4 x,AF4 n,AF4 m){return max(n,min(x,m));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF1 AFractF1(AF1 x){return x-floor(x);}
+ AF2 AFractF2(AF2 x){return x-floor(x);}
+ AF3 AFractF3(AF3 x){return x-floor(x);}
+ AF4 AFractF4(AF4 x){return x-floor(x);}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF1 ALerpF1(AF1 x,AF1 y,AF1 a){return lerp(x,y,a);}
+ AF2 ALerpF2(AF2 x,AF2 y,AF2 a){return lerp(x,y,a);}
+ AF3 ALerpF3(AF3 x,AF3 y,AF3 a){return lerp(x,y,a);}
+ AF4 ALerpF4(AF4 x,AF4 y,AF4 a){return lerp(x,y,a);}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF1 AMax3F1(AF1 x,AF1 y,AF1 z){return max(x,max(y,z));}
+ AF2 AMax3F2(AF2 x,AF2 y,AF2 z){return max(x,max(y,z));}
+ AF3 AMax3F3(AF3 x,AF3 y,AF3 z){return max(x,max(y,z));}
+ AF4 AMax3F4(AF4 x,AF4 y,AF4 z){return max(x,max(y,z));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 AMax3SU1(AU1 x,AU1 y,AU1 z){return AU1(max(ASU1(x),max(ASU1(y),ASU1(z))));}
+ AU2 AMax3SU2(AU2 x,AU2 y,AU2 z){return AU2(max(ASU2(x),max(ASU2(y),ASU2(z))));}
+ AU3 AMax3SU3(AU3 x,AU3 y,AU3 z){return AU3(max(ASU3(x),max(ASU3(y),ASU3(z))));}
+ AU4 AMax3SU4(AU4 x,AU4 y,AU4 z){return AU4(max(ASU4(x),max(ASU4(y),ASU4(z))));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 AMax3U1(AU1 x,AU1 y,AU1 z){return max(x,max(y,z));}
+ AU2 AMax3U2(AU2 x,AU2 y,AU2 z){return max(x,max(y,z));}
+ AU3 AMax3U3(AU3 x,AU3 y,AU3 z){return max(x,max(y,z));}
+ AU4 AMax3U4(AU4 x,AU4 y,AU4 z){return max(x,max(y,z));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 AMaxSU1(AU1 a,AU1 b){return AU1(max(ASU1(a),ASU1(b)));}
+ AU2 AMaxSU2(AU2 a,AU2 b){return AU2(max(ASU2(a),ASU2(b)));}
+ AU3 AMaxSU3(AU3 a,AU3 b){return AU3(max(ASU3(a),ASU3(b)));}
+ AU4 AMaxSU4(AU4 a,AU4 b){return AU4(max(ASU4(a),ASU4(b)));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF1 AMed3F1(AF1 x,AF1 y,AF1 z){return max(min(x,y),min(max(x,y),z));}
+ AF2 AMed3F2(AF2 x,AF2 y,AF2 z){return max(min(x,y),min(max(x,y),z));}
+ AF3 AMed3F3(AF3 x,AF3 y,AF3 z){return max(min(x,y),min(max(x,y),z));}
+ AF4 AMed3F4(AF4 x,AF4 y,AF4 z){return max(min(x,y),min(max(x,y),z));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF1 AMin3F1(AF1 x,AF1 y,AF1 z){return min(x,min(y,z));}
+ AF2 AMin3F2(AF2 x,AF2 y,AF2 z){return min(x,min(y,z));}
+ AF3 AMin3F3(AF3 x,AF3 y,AF3 z){return min(x,min(y,z));}
+ AF4 AMin3F4(AF4 x,AF4 y,AF4 z){return min(x,min(y,z));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 AMin3SU1(AU1 x,AU1 y,AU1 z){return AU1(min(ASU1(x),min(ASU1(y),ASU1(z))));}
+ AU2 AMin3SU2(AU2 x,AU2 y,AU2 z){return AU2(min(ASU2(x),min(ASU2(y),ASU2(z))));}
+ AU3 AMin3SU3(AU3 x,AU3 y,AU3 z){return AU3(min(ASU3(x),min(ASU3(y),ASU3(z))));}
+ AU4 AMin3SU4(AU4 x,AU4 y,AU4 z){return AU4(min(ASU4(x),min(ASU4(y),ASU4(z))));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 AMin3U1(AU1 x,AU1 y,AU1 z){return min(x,min(y,z));}
+ AU2 AMin3U2(AU2 x,AU2 y,AU2 z){return min(x,min(y,z));}
+ AU3 AMin3U3(AU3 x,AU3 y,AU3 z){return min(x,min(y,z));}
+ AU4 AMin3U4(AU4 x,AU4 y,AU4 z){return min(x,min(y,z));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 AMinSU1(AU1 a,AU1 b){return AU1(min(ASU1(a),ASU1(b)));}
+ AU2 AMinSU2(AU2 a,AU2 b){return AU2(min(ASU2(a),ASU2(b)));}
+ AU3 AMinSU3(AU3 a,AU3 b){return AU3(min(ASU3(a),ASU3(b)));}
+ AU4 AMinSU4(AU4 a,AU4 b){return AU4(min(ASU4(a),ASU4(b)));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF1 ANCosF1(AF1 x){return cos(x*AF1_(A_2PI));}
+ AF2 ANCosF2(AF2 x){return cos(x*AF2_(A_2PI));}
+ AF3 ANCosF3(AF3 x){return cos(x*AF3_(A_2PI));}
+ AF4 ANCosF4(AF4 x){return cos(x*AF4_(A_2PI));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF1 ANSinF1(AF1 x){return sin(x*AF1_(A_2PI));}
+ AF2 ANSinF2(AF2 x){return sin(x*AF2_(A_2PI));}
+ AF3 ANSinF3(AF3 x){return sin(x*AF3_(A_2PI));}
+ AF4 ANSinF4(AF4 x){return sin(x*AF4_(A_2PI));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF1 ARcpF1(AF1 x){return rcp(x);}
+ AF2 ARcpF2(AF2 x){return rcp(x);}
+ AF3 ARcpF3(AF3 x){return rcp(x);}
+ AF4 ARcpF4(AF4 x){return rcp(x);}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF1 ARsqF1(AF1 x){return rsqrt(x);}
+ AF2 ARsqF2(AF2 x){return rsqrt(x);}
+ AF3 ARsqF3(AF3 x){return rsqrt(x);}
+ AF4 ARsqF4(AF4 x){return rsqrt(x);}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF1 ASatF1(AF1 x){return saturate(x);}
+ AF2 ASatF2(AF2 x){return saturate(x);}
+ AF3 ASatF3(AF3 x){return saturate(x);}
+ AF4 ASatF4(AF4 x){return saturate(x);}
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 AShrSU1(AU1 a,AU1 b){return AU1(ASU1(a)>>ASU1(b));}
+ AU2 AShrSU2(AU2 a,AU2 b){return AU2(ASU2(a)>>ASU2(b));}
+ AU3 AShrSU3(AU3 a,AU3 b){return AU3(ASU3(a)>>ASU3(b));}
+ AU4 AShrSU4(AU4 a,AU4 b){return AU4(ASU4(a)>>ASU4(b));}
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                          HLSL BYTE
+//==============================================================================================================================
+ #ifdef A_BYTE
+ #endif
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                          HLSL HALF
+//==============================================================================================================================
+ #ifdef A_HALF
+  #ifdef A_HLSL_6_2
+   #define AH1 float16_t
+   #define AH2 float16_t2
+   #define AH3 float16_t3
+   #define AH4 float16_t4
+//------------------------------------------------------------------------------------------------------------------------------
+   #define AW1 uint16_t
+   #define AW2 uint16_t2
+   #define AW3 uint16_t3
+   #define AW4 uint16_t4
+//------------------------------------------------------------------------------------------------------------------------------
+   #define ASW1 int16_t
+   #define ASW2 int16_t2
+   #define ASW3 int16_t3
+   #define ASW4 int16_t4
+  #else
+   #define AH1 min16float
+   #define AH2 min16float2
+   #define AH3 min16float3
+   #define AH4 min16float4
+//------------------------------------------------------------------------------------------------------------------------------
+   #define AW1 min16uint
+   #define AW2 min16uint2
+   #define AW3 min16uint3
+   #define AW4 min16uint4
+//------------------------------------------------------------------------------------------------------------------------------
+   #define ASW1 min16int
+   #define ASW2 min16int2
+   #define ASW3 min16int3
+   #define ASW4 min16int4
+  #endif
+//==============================================================================================================================
+  // Need to use manual unpack to get optimal execution (don't use packed types in buffers directly).
+  // Unpack requires this pattern: https://gpuopen.com/first-steps-implementing-fp16/
+  AH2 AH2_AU1_x(AU1 x){AF2 t=f16tof32(AU2(x&0xFFFF,x>>16));return AH2(t);}
+  AH4 AH4_AU2_x(AU2 x){return AH4(AH2_AU1_x(x.x),AH2_AU1_x(x.y));}
+  AW2 AW2_AU1_x(AU1 x){AU2 t=AU2(x&0xFFFF,x>>16);return AW2(t);}
+  AW4 AW4_AU2_x(AU2 x){return AW4(AW2_AU1_x(x.x),AW2_AU1_x(x.y));}
+  #define AH2_AU1(x) AH2_AU1_x(AU1(x))
+  #define AH4_AU2(x) AH4_AU2_x(AU2(x))
+  #define AW2_AU1(x) AW2_AU1_x(AU1(x))
+  #define AW4_AU2(x) AW4_AU2_x(AU2(x))
+//------------------------------------------------------------------------------------------------------------------------------
+  AU1 AU1_AH2_x(AH2 x){return f32tof16(x.x)+(f32tof16(x.y)<<16);}
+  AU2 AU2_AH4_x(AH4 x){return AU2(AU1_AH2_x(x.xy),AU1_AH2_x(x.zw));}
+  AU1 AU1_AW2_x(AW2 x){return AU1(x.x)+(AU1(x.y)<<16);}
+  AU2 AU2_AW4_x(AW4 x){return AU2(AU1_AW2_x(x.xy),AU1_AW2_x(x.zw));}
+  #define AU1_AH2(x) AU1_AH2_x(AH2(x))
+  #define AU2_AH4(x) AU2_AH4_x(AH4(x))
+  #define AU1_AW2(x) AU1_AW2_x(AW2(x))
+  #define AU2_AW4(x) AU2_AW4_x(AW4(x))
+//==============================================================================================================================
+  #if defined(A_HLSL_6_2) && !defined(A_NO_16_BIT_CAST)
+   #define AW1_AH1(x) asuint16(x)
+   #define AW2_AH2(x) asuint16(x)
+   #define AW3_AH3(x) asuint16(x)
+   #define AW4_AH4(x) asuint16(x)
+  #else
+   #define AW1_AH1(a) AW1(f32tof16(AF1(a)))
+   #define AW2_AH2(a) AW2(AW1_AH1((a).x),AW1_AH1((a).y))
+   #define AW3_AH3(a) AW3(AW1_AH1((a).x),AW1_AH1((a).y),AW1_AH1((a).z))
+   #define AW4_AH4(a) AW4(AW1_AH1((a).x),AW1_AH1((a).y),AW1_AH1((a).z),AW1_AH1((a).w))
+  #endif
+//------------------------------------------------------------------------------------------------------------------------------
+  #if defined(A_HLSL_6_2) && !defined(A_NO_16_BIT_CAST)
+   #define AH1_AW1(x) asfloat16(x)
+   #define AH2_AW2(x) asfloat16(x)
+   #define AH3_AW3(x) asfloat16(x)
+   #define AH4_AW4(x) asfloat16(x)
+  #else
+   #define AH1_AW1(a) AH1(f16tof32(AU1(a)))
+   #define AH2_AW2(a) AH2(AH1_AW1((a).x),AH1_AW1((a).y))
+   #define AH3_AW3(a) AH3(AH1_AW1((a).x),AH1_AW1((a).y),AH1_AW1((a).z))
+   #define AH4_AW4(a) AH4(AH1_AW1((a).x),AH1_AW1((a).y),AH1_AW1((a).z),AH1_AW1((a).w))
+  #endif
+//==============================================================================================================================
+  AH1 AH1_x(AH1 a){return AH1(a);}
+  AH2 AH2_x(AH1 a){return AH2(a,a);}
+  AH3 AH3_x(AH1 a){return AH3(a,a,a);}
+  AH4 AH4_x(AH1 a){return AH4(a,a,a,a);}
+  #define AH1_(a) AH1_x(AH1(a))
+  #define AH2_(a) AH2_x(AH1(a))
+  #define AH3_(a) AH3_x(AH1(a))
+  #define AH4_(a) AH4_x(AH1(a))
+//------------------------------------------------------------------------------------------------------------------------------
+  AW1 AW1_x(AW1 a){return AW1(a);}
+  AW2 AW2_x(AW1 a){return AW2(a,a);}
+  AW3 AW3_x(AW1 a){return AW3(a,a,a);}
+  AW4 AW4_x(AW1 a){return AW4(a,a,a,a);}
+  #define AW1_(a) AW1_x(AW1(a))
+  #define AW2_(a) AW2_x(AW1(a))
+  #define AW3_(a) AW3_x(AW1(a))
+  #define AW4_(a) AW4_x(AW1(a))
+//==============================================================================================================================
+  AW1 AAbsSW1(AW1 a){return AW1(abs(ASW1(a)));}
+  AW2 AAbsSW2(AW2 a){return AW2(abs(ASW2(a)));}
+  AW3 AAbsSW3(AW3 a){return AW3(abs(ASW3(a)));}
+  AW4 AAbsSW4(AW4 a){return AW4(abs(ASW4(a)));}
+//------------------------------------------------------------------------------------------------------------------------------
+  AH1 AClampH1(AH1 x,AH1 n,AH1 m){return max(n,min(x,m));}
+  AH2 AClampH2(AH2 x,AH2 n,AH2 m){return max(n,min(x,m));}
+  AH3 AClampH3(AH3 x,AH3 n,AH3 m){return max(n,min(x,m));}
+  AH4 AClampH4(AH4 x,AH4 n,AH4 m){return max(n,min(x,m));}
+//------------------------------------------------------------------------------------------------------------------------------
+ // V_FRACT_F16 (note DX frac() is different).
+  AH1 AFractH1(AH1 x){return x-floor(x);}
+  AH2 AFractH2(AH2 x){return x-floor(x);}
+  AH3 AFractH3(AH3 x){return x-floor(x);}
+  AH4 AFractH4(AH4 x){return x-floor(x);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AH1 ALerpH1(AH1 x,AH1 y,AH1 a){return lerp(x,y,a);}
+  AH2 ALerpH2(AH2 x,AH2 y,AH2 a){return lerp(x,y,a);}
+  AH3 ALerpH3(AH3 x,AH3 y,AH3 a){return lerp(x,y,a);}
+  AH4 ALerpH4(AH4 x,AH4 y,AH4 a){return lerp(x,y,a);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AH1 AMax3H1(AH1 x,AH1 y,AH1 z){return max(x,max(y,z));}
+  AH2 AMax3H2(AH2 x,AH2 y,AH2 z){return max(x,max(y,z));}
+  AH3 AMax3H3(AH3 x,AH3 y,AH3 z){return max(x,max(y,z));}
+  AH4 AMax3H4(AH4 x,AH4 y,AH4 z){return max(x,max(y,z));}
+//------------------------------------------------------------------------------------------------------------------------------
+  AW1 AMaxSW1(AW1 a,AW1 b){return AW1(max(ASU1(a),ASU1(b)));}
+  AW2 AMaxSW2(AW2 a,AW2 b){return AW2(max(ASU2(a),ASU2(b)));}
+  AW3 AMaxSW3(AW3 a,AW3 b){return AW3(max(ASU3(a),ASU3(b)));}
+  AW4 AMaxSW4(AW4 a,AW4 b){return AW4(max(ASU4(a),ASU4(b)));}
+//------------------------------------------------------------------------------------------------------------------------------
+  AH1 AMin3H1(AH1 x,AH1 y,AH1 z){return min(x,min(y,z));}
+  AH2 AMin3H2(AH2 x,AH2 y,AH2 z){return min(x,min(y,z));}
+  AH3 AMin3H3(AH3 x,AH3 y,AH3 z){return min(x,min(y,z));}
+  AH4 AMin3H4(AH4 x,AH4 y,AH4 z){return min(x,min(y,z));}
+//------------------------------------------------------------------------------------------------------------------------------
+  AW1 AMinSW1(AW1 a,AW1 b){return AW1(min(ASU1(a),ASU1(b)));}
+  AW2 AMinSW2(AW2 a,AW2 b){return AW2(min(ASU2(a),ASU2(b)));}
+  AW3 AMinSW3(AW3 a,AW3 b){return AW3(min(ASU3(a),ASU3(b)));}
+  AW4 AMinSW4(AW4 a,AW4 b){return AW4(min(ASU4(a),ASU4(b)));}
+//------------------------------------------------------------------------------------------------------------------------------
+  AH1 ARcpH1(AH1 x){return rcp(x);}
+  AH2 ARcpH2(AH2 x){return rcp(x);}
+  AH3 ARcpH3(AH3 x){return rcp(x);}
+  AH4 ARcpH4(AH4 x){return rcp(x);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AH1 ARsqH1(AH1 x){return rsqrt(x);}
+  AH2 ARsqH2(AH2 x){return rsqrt(x);}
+  AH3 ARsqH3(AH3 x){return rsqrt(x);}
+  AH4 ARsqH4(AH4 x){return rsqrt(x);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AH1 ASatH1(AH1 x){return saturate(x);}
+  AH2 ASatH2(AH2 x){return saturate(x);}
+  AH3 ASatH3(AH3 x){return saturate(x);}
+  AH4 ASatH4(AH4 x){return saturate(x);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AW1 AShrSW1(AW1 a,AW1 b){return AW1(ASW1(a)>>ASW1(b));}
+  AW2 AShrSW2(AW2 a,AW2 b){return AW2(ASW2(a)>>ASW2(b));}
+  AW3 AShrSW3(AW3 a,AW3 b){return AW3(ASW3(a)>>ASW3(b));}
+  AW4 AShrSW4(AW4 a,AW4 b){return AW4(ASW4(a)>>ASW4(b));}
+ #endif
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                         HLSL DOUBLE
+//==============================================================================================================================
+ #ifdef A_DUBL
+  #ifdef A_HLSL_6_2
+   #define AD1 float64_t
+   #define AD2 float64_t2
+   #define AD3 float64_t3
+   #define AD4 float64_t4
+  #else
+   #define AD1 double
+   #define AD2 double2
+   #define AD3 double3
+   #define AD4 double4
+  #endif
+//------------------------------------------------------------------------------------------------------------------------------
+  AD1 AD1_x(AD1 a){return AD1(a);}
+  AD2 AD2_x(AD1 a){return AD2(a,a);}
+  AD3 AD3_x(AD1 a){return AD3(a,a,a);}
+  AD4 AD4_x(AD1 a){return AD4(a,a,a,a);}
+  #define AD1_(a) AD1_x(AD1(a))
+  #define AD2_(a) AD2_x(AD1(a))
+  #define AD3_(a) AD3_x(AD1(a))
+  #define AD4_(a) AD4_x(AD1(a))
+//==============================================================================================================================
+  AD1 AFractD1(AD1 a){return a-floor(a);}
+  AD2 AFractD2(AD2 a){return a-floor(a);}
+  AD3 AFractD3(AD3 a){return a-floor(a);}
+  AD4 AFractD4(AD4 a){return a-floor(a);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AD1 ALerpD1(AD1 x,AD1 y,AD1 a){return lerp(x,y,a);}
+  AD2 ALerpD2(AD2 x,AD2 y,AD2 a){return lerp(x,y,a);}
+  AD3 ALerpD3(AD3 x,AD3 y,AD3 a){return lerp(x,y,a);}
+  AD4 ALerpD4(AD4 x,AD4 y,AD4 a){return lerp(x,y,a);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AD1 ARcpD1(AD1 x){return rcp(x);}
+  AD2 ARcpD2(AD2 x){return rcp(x);}
+  AD3 ARcpD3(AD3 x){return rcp(x);}
+  AD4 ARcpD4(AD4 x){return rcp(x);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AD1 ARsqD1(AD1 x){return rsqrt(x);}
+  AD2 ARsqD2(AD2 x){return rsqrt(x);}
+  AD3 ARsqD3(AD3 x){return rsqrt(x);}
+  AD4 ARsqD4(AD4 x){return rsqrt(x);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AD1 ASatD1(AD1 x){return saturate(x);}
+  AD2 ASatD2(AD2 x){return saturate(x);}
+  AD3 ASatD3(AD3 x){return saturate(x);}
+  AD4 ASatD4(AD4 x){return saturate(x);}
+ #endif
+//==============================================================================================================================
+//                                                         HLSL WAVE
+//==============================================================================================================================
+ #ifdef A_WAVE
+  // Where 'x' must be a compile time literal.
+  AF1 AWaveXorF1(AF1 v,AU1 x){return WaveReadLaneAt(v,WaveGetLaneIndex()^x);}
+  AF2 AWaveXorF2(AF2 v,AU1 x){return WaveReadLaneAt(v,WaveGetLaneIndex()^x);}
+  AF3 AWaveXorF3(AF3 v,AU1 x){return WaveReadLaneAt(v,WaveGetLaneIndex()^x);}
+  AF4 AWaveXorF4(AF4 v,AU1 x){return WaveReadLaneAt(v,WaveGetLaneIndex()^x);}
+  AU1 AWaveXorU1(AU1 v,AU1 x){return WaveReadLaneAt(v,WaveGetLaneIndex()^x);}
+  AU2 AWaveXorU1(AU2 v,AU1 x){return WaveReadLaneAt(v,WaveGetLaneIndex()^x);}
+  AU3 AWaveXorU1(AU3 v,AU1 x){return WaveReadLaneAt(v,WaveGetLaneIndex()^x);}
+  AU4 AWaveXorU1(AU4 v,AU1 x){return WaveReadLaneAt(v,WaveGetLaneIndex()^x);}
+//------------------------------------------------------------------------------------------------------------------------------
+  #ifdef A_HALF
+   AH2 AWaveXorH2(AH2 v,AU1 x){return AH2_AU1(WaveReadLaneAt(AU1_AH2(v),WaveGetLaneIndex()^x));}
+   AH4 AWaveXorH4(AH4 v,AU1 x){return AH4_AU2(WaveReadLaneAt(AU2_AH4(v),WaveGetLaneIndex()^x));}
+   AW2 AWaveXorW2(AW2 v,AU1 x){return AW2_AU1(WaveReadLaneAt(AU1_AW2(v),WaveGetLaneIndex()^x));}
+   AW4 AWaveXorW4(AW4 v,AU1 x){return AW4_AU1(WaveReadLaneAt(AU1_AW4(v),WaveGetLaneIndex()^x));}
+  #endif
+ #endif
+//==============================================================================================================================
+#endif
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//
+//
+//                                                          GPU COMMON
+//
+//
+//==============================================================================================================================
+#ifdef A_GPU
+ // Negative and positive infinity.
+ #define A_INFP_F AF1_AU1(0x7f800000u)
+ #define A_INFN_F AF1_AU1(0xff800000u)
+//------------------------------------------------------------------------------------------------------------------------------
+ // Copy sign from 's' to positive 'd'.
+ AF1 ACpySgnF1(AF1 d,AF1 s){return AF1_AU1(AU1_AF1(d)|(AU1_AF1(s)&AU1_(0x80000000u)));}
+ AF2 ACpySgnF2(AF2 d,AF2 s){return AF2_AU2(AU2_AF2(d)|(AU2_AF2(s)&AU2_(0x80000000u)));}
+ AF3 ACpySgnF3(AF3 d,AF3 s){return AF3_AU3(AU3_AF3(d)|(AU3_AF3(s)&AU3_(0x80000000u)));}
+ AF4 ACpySgnF4(AF4 d,AF4 s){return AF4_AU4(AU4_AF4(d)|(AU4_AF4(s)&AU4_(0x80000000u)));}
+//------------------------------------------------------------------------------------------------------------------------------
+ // Single operation to return (useful to create a mask to use in lerp for branch free logic),
+ //  m=NaN := 0
+ //  m>=0  := 0
+ //  m<0   := 1
+ // Uses the following useful floating point logic,
+ //  saturate(+a*(-INF)==-INF) := 0
+ //  saturate( 0*(-INF)== NaN) := 0
+ //  saturate(-a*(-INF)==+INF) := 1
+ AF1 ASignedF1(AF1 m){return ASatF1(m*AF1_(A_INFN_F));}
+ AF2 ASignedF2(AF2 m){return ASatF2(m*AF2_(A_INFN_F));}
+ AF3 ASignedF3(AF3 m){return ASatF3(m*AF3_(A_INFN_F));}
+ AF4 ASignedF4(AF4 m){return ASatF4(m*AF4_(A_INFN_F));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF1 AGtZeroF1(AF1 m){return ASatF1(m*AF1_(A_INFP_F));}
+ AF2 AGtZeroF2(AF2 m){return ASatF2(m*AF2_(A_INFP_F));}
+ AF3 AGtZeroF3(AF3 m){return ASatF3(m*AF3_(A_INFP_F));}
+ AF4 AGtZeroF4(AF4 m){return ASatF4(m*AF4_(A_INFP_F));}
+//==============================================================================================================================
+ #ifdef A_HALF
+  #ifdef A_HLSL_6_2
+   #define A_INFP_H AH1_AW1((uint16_t)0x7c00u)
+   #define A_INFN_H AH1_AW1((uint16_t)0xfc00u)
+  #else
+   #define A_INFP_H AH1_AW1(0x7c00u)
+   #define A_INFN_H AH1_AW1(0xfc00u)
+  #endif
+
+//------------------------------------------------------------------------------------------------------------------------------
+  AH1 ACpySgnH1(AH1 d,AH1 s){return AH1_AW1(AW1_AH1(d)|(AW1_AH1(s)&AW1_(0x8000u)));}
+  AH2 ACpySgnH2(AH2 d,AH2 s){return AH2_AW2(AW2_AH2(d)|(AW2_AH2(s)&AW2_(0x8000u)));}
+  AH3 ACpySgnH3(AH3 d,AH3 s){return AH3_AW3(AW3_AH3(d)|(AW3_AH3(s)&AW3_(0x8000u)));}
+  AH4 ACpySgnH4(AH4 d,AH4 s){return AH4_AW4(AW4_AH4(d)|(AW4_AH4(s)&AW4_(0x8000u)));}
+//------------------------------------------------------------------------------------------------------------------------------
+  AH1 ASignedH1(AH1 m){return ASatH1(m*AH1_(A_INFN_H));}
+  AH2 ASignedH2(AH2 m){return ASatH2(m*AH2_(A_INFN_H));}
+  AH3 ASignedH3(AH3 m){return ASatH3(m*AH3_(A_INFN_H));}
+  AH4 ASignedH4(AH4 m){return ASatH4(m*AH4_(A_INFN_H));}
+//------------------------------------------------------------------------------------------------------------------------------
+  AH1 AGtZeroH1(AH1 m){return ASatH1(m*AH1_(A_INFP_H));}
+  AH2 AGtZeroH2(AH2 m){return ASatH2(m*AH2_(A_INFP_H));}
+  AH3 AGtZeroH3(AH3 m){return ASatH3(m*AH3_(A_INFP_H));}
+  AH4 AGtZeroH4(AH4 m){return ASatH4(m*AH4_(A_INFP_H));}
+ #endif
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                [FIS] FLOAT INTEGER SORTABLE
+//------------------------------------------------------------------------------------------------------------------------------
+// Float to integer sortable.
+//  - If sign bit=0, flip the sign bit (positives).
+//  - If sign bit=1, flip all bits     (negatives).
+// Integer sortable to float.
+//  - If sign bit=1, flip the sign bit (positives).
+//  - If sign bit=0, flip all bits     (negatives).
+// Has nice side effects.
+//  - Larger integers are more positive values.
+//  - Float zero is mapped to center of integers (so clear to integer zero is a nice default for atomic max usage).
+// Burns 3 ops for conversion {shift,or,xor}.
+//==============================================================================================================================
+ AU1 AFisToU1(AU1 x){return x^(( AShrSU1(x,AU1_(31)))|AU1_(0x80000000));}
+ AU1 AFisFromU1(AU1 x){return x^((~AShrSU1(x,AU1_(31)))|AU1_(0x80000000));}
+//------------------------------------------------------------------------------------------------------------------------------
+ // Just adjust high 16-bit value (useful when upper part of 32-bit word is a 16-bit float value).
+ AU1 AFisToHiU1(AU1 x){return x^(( AShrSU1(x,AU1_(15)))|AU1_(0x80000000));}
+ AU1 AFisFromHiU1(AU1 x){return x^((~AShrSU1(x,AU1_(15)))|AU1_(0x80000000));}
+//------------------------------------------------------------------------------------------------------------------------------
+ #ifdef A_HALF
+  AW1 AFisToW1(AW1 x){return x^(( AShrSW1(x,AW1_(15)))|AW1_(0x8000));}
+  AW1 AFisFromW1(AW1 x){return x^((~AShrSW1(x,AW1_(15)))|AW1_(0x8000));}
+//------------------------------------------------------------------------------------------------------------------------------
+  AW2 AFisToW2(AW2 x){return x^(( AShrSW2(x,AW2_(15)))|AW2_(0x8000));}
+  AW2 AFisFromW2(AW2 x){return x^((~AShrSW2(x,AW2_(15)))|AW2_(0x8000));}
+ #endif
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                      [PERM] V_PERM_B32
+//------------------------------------------------------------------------------------------------------------------------------
+// Support for V_PERM_B32 started in the 3rd generation of GCN.
+//------------------------------------------------------------------------------------------------------------------------------
+// yyyyxxxx - The 'i' input.
+// 76543210
+// ========
+// HGFEDCBA - Naming on permutation.
+//------------------------------------------------------------------------------------------------------------------------------
+// TODO
+// ====
+//  - Make sure compiler optimizes this.
+//==============================================================================================================================
+ #ifdef A_HALF
+  AU1 APerm0E0A(AU2 i){return((i.x    )&0xffu)|((i.y<<16)&0xff0000u);}
+  AU1 APerm0F0B(AU2 i){return((i.x>> 8)&0xffu)|((i.y<< 8)&0xff0000u);}
+  AU1 APerm0G0C(AU2 i){return((i.x>>16)&0xffu)|((i.y    )&0xff0000u);}
+  AU1 APerm0H0D(AU2 i){return((i.x>>24)&0xffu)|((i.y>> 8)&0xff0000u);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AU1 APermHGFA(AU2 i){return((i.x    )&0x000000ffu)|(i.y&0xffffff00u);}
+  AU1 APermHGFC(AU2 i){return((i.x>>16)&0x000000ffu)|(i.y&0xffffff00u);}
+  AU1 APermHGAE(AU2 i){return((i.x<< 8)&0x0000ff00u)|(i.y&0xffff00ffu);}
+  AU1 APermHGCE(AU2 i){return((i.x>> 8)&0x0000ff00u)|(i.y&0xffff00ffu);}
+  AU1 APermHAFE(AU2 i){return((i.x<<16)&0x00ff0000u)|(i.y&0xff00ffffu);}
+  AU1 APermHCFE(AU2 i){return((i.x    )&0x00ff0000u)|(i.y&0xff00ffffu);}
+  AU1 APermAGFE(AU2 i){return((i.x<<24)&0xff000000u)|(i.y&0x00ffffffu);}
+  AU1 APermCGFE(AU2 i){return((i.x<< 8)&0xff000000u)|(i.y&0x00ffffffu);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AU1 APermGCEA(AU2 i){return((i.x)&0x00ff00ffu)|((i.y<<8)&0xff00ff00u);}
+  AU1 APermGECA(AU2 i){return(((i.x)&0xffu)|((i.x>>8)&0xff00u)|((i.y<<16)&0xff0000u)|((i.y<<8)&0xff000000u));}
+ #endif
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                               [BUC] BYTE UNSIGNED CONVERSION
+//------------------------------------------------------------------------------------------------------------------------------
+// Designed to use the optimal conversion, enables the scaling to possibly be factored into other computation.
+// Works on a range of {0 to A_BUC_<32,16>}, for <32-bit, and 16-bit> respectively.
+//------------------------------------------------------------------------------------------------------------------------------
+// OPCODE NOTES
+// ============
+// GCN does not do UNORM or SNORM for bytes in opcodes.
+//  - V_CVT_F32_UBYTE{0,1,2,3} - Unsigned byte to float.
+//  - V_CVT_PKACC_U8_F32 - Float to unsigned byte (does bit-field insert into 32-bit integer).
+// V_PERM_B32 does byte packing with ability to zero fill bytes as well.
+//  - Can pull out byte values from two sources, and zero fill upper 8-bits of packed hi and lo. 
+//------------------------------------------------------------------------------------------------------------------------------
+// BYTE : FLOAT - ABuc{0,1,2,3}{To,From}U1() - Designed for V_CVT_F32_UBYTE* and V_CVT_PKACCUM_U8_F32 ops.
+// ====   =====
+//    0 : 0
+//    1 : 1
+//     ...
+//  255 : 255
+//      : 256 (just outside the encoding range)
+//------------------------------------------------------------------------------------------------------------------------------
+// BYTE : FLOAT - ABuc{0,1,2,3}{To,From}U2() - Designed for 16-bit denormal tricks and V_PERM_B32.
+// ====   =====
+//    0 : 0
+//    1 : 1/512
+//    2 : 1/256
+//     ...
+//   64 : 1/8
+//  128 : 1/4
+//  255 : 255/512
+//      : 1/2 (just outside the encoding range)
+//------------------------------------------------------------------------------------------------------------------------------
+// OPTIMAL IMPLEMENTATIONS ON AMD ARCHITECTURES
+// ============================================
+// r=ABuc0FromU1(i)
+//   V_CVT_F32_UBYTE0 r,i
+// --------------------------------------------
+// r=ABuc0ToU1(d,i)
+//   V_CVT_PKACCUM_U8_F32 r,i,0,d
+// --------------------------------------------
+// d=ABuc0FromU2(i)
+//   Where 'k0' is an SGPR with 0x0E0A
+//   Where 'k1' is an SGPR with {32768.0} packed into the lower 16-bits
+//   V_PERM_B32 d,i.x,i.y,k0
+//   V_PK_FMA_F16 d,d,k1.x,0
+// --------------------------------------------
+// r=ABuc0ToU2(d,i)
+//   Where 'k0' is an SGPR with {1.0/32768.0} packed into the lower 16-bits
+//   Where 'k1' is an SGPR with 0x????
+//   Where 'k2' is an SGPR with 0x????
+//   V_PK_FMA_F16 i,i,k0.x,0
+//   V_PERM_B32 r.x,i,i,k1
+//   V_PERM_B32 r.y,i,i,k2
+//==============================================================================================================================
+ // Peak range for 32-bit and 16-bit operations.
+ #define A_BUC_32 (255.0)
+ #define A_BUC_16 (255.0/512.0)
+//==============================================================================================================================
+ #if 1
+  // Designed to be one V_CVT_PKACCUM_U8_F32.
+  // The extra min is required to pattern match to V_CVT_PKACCUM_U8_F32.
+  AU1 ABuc0ToU1(AU1 d,AF1 i){return (d&0xffffff00u)|((min(AU1(i),255u)    )&(0x000000ffu));}
+  AU1 ABuc1ToU1(AU1 d,AF1 i){return (d&0xffff00ffu)|((min(AU1(i),255u)<< 8)&(0x0000ff00u));}
+  AU1 ABuc2ToU1(AU1 d,AF1 i){return (d&0xff00ffffu)|((min(AU1(i),255u)<<16)&(0x00ff0000u));}
+  AU1 ABuc3ToU1(AU1 d,AF1 i){return (d&0x00ffffffu)|((min(AU1(i),255u)<<24)&(0xff000000u));}
+//------------------------------------------------------------------------------------------------------------------------------
+  // Designed to be one V_CVT_F32_UBYTE*.
+  AF1 ABuc0FromU1(AU1 i){return AF1((i    )&255u);}
+  AF1 ABuc1FromU1(AU1 i){return AF1((i>> 8)&255u);}
+  AF1 ABuc2FromU1(AU1 i){return AF1((i>>16)&255u);}
+  AF1 ABuc3FromU1(AU1 i){return AF1((i>>24)&255u);}
+ #endif
+//==============================================================================================================================
+ #ifdef A_HALF
+  // Takes {x0,x1} and {y0,y1} and builds {{x0,y0},{x1,y1}}.
+  AW2 ABuc01ToW2(AH2 x,AH2 y){x*=AH2_(1.0/32768.0);y*=AH2_(1.0/32768.0);
+   return AW2_AU1(APermGCEA(AU2(AU1_AW2(AW2_AH2(x)),AU1_AW2(AW2_AH2(y)))));}
+//------------------------------------------------------------------------------------------------------------------------------
+  // Designed for 3 ops to do SOA to AOS and conversion.
+  AU2 ABuc0ToU2(AU2 d,AH2 i){AU1 b=AU1_AW2(AW2_AH2(i*AH2_(1.0/32768.0)));
+   return AU2(APermHGFA(AU2(d.x,b)),APermHGFC(AU2(d.y,b)));}
+  AU2 ABuc1ToU2(AU2 d,AH2 i){AU1 b=AU1_AW2(AW2_AH2(i*AH2_(1.0/32768.0)));
+   return AU2(APermHGAE(AU2(d.x,b)),APermHGCE(AU2(d.y,b)));}
+  AU2 ABuc2ToU2(AU2 d,AH2 i){AU1 b=AU1_AW2(AW2_AH2(i*AH2_(1.0/32768.0)));
+   return AU2(APermHAFE(AU2(d.x,b)),APermHCFE(AU2(d.y,b)));}
+  AU2 ABuc3ToU2(AU2 d,AH2 i){AU1 b=AU1_AW2(AW2_AH2(i*AH2_(1.0/32768.0)));
+   return AU2(APermAGFE(AU2(d.x,b)),APermCGFE(AU2(d.y,b)));}
+//------------------------------------------------------------------------------------------------------------------------------
+  // Designed for 2 ops to do both AOS to SOA, and conversion.
+  AH2 ABuc0FromU2(AU2 i){return AH2_AW2(AW2_AU1(APerm0E0A(i)))*AH2_(32768.0);}
+  AH2 ABuc1FromU2(AU2 i){return AH2_AW2(AW2_AU1(APerm0F0B(i)))*AH2_(32768.0);}
+  AH2 ABuc2FromU2(AU2 i){return AH2_AW2(AW2_AU1(APerm0G0C(i)))*AH2_(32768.0);}
+  AH2 ABuc3FromU2(AU2 i){return AH2_AW2(AW2_AU1(APerm0H0D(i)))*AH2_(32768.0);}
+ #endif
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                 [BSC] BYTE SIGNED CONVERSION
+//------------------------------------------------------------------------------------------------------------------------------
+// Similar to [BUC].
+// Works on a range of {-/+ A_BSC_<32,16>}, for <32-bit, and 16-bit> respectively.
+//------------------------------------------------------------------------------------------------------------------------------
+// ENCODING (without zero-based encoding)
+// ========
+//   0 = unused (can be used to mean something else)
+//   1 = lowest value 
+// 128 = exact zero center (zero based encoding 
+// 255 = highest value
+//------------------------------------------------------------------------------------------------------------------------------
+// Zero-based [Zb] flips the MSB bit of the byte (making 128 "exact zero" actually zero).
+// This is useful if there is a desire for cleared values to decode as zero.
+//------------------------------------------------------------------------------------------------------------------------------
+// BYTE : FLOAT - ABsc{0,1,2,3}{To,From}U2() - Designed for 16-bit denormal tricks and V_PERM_B32.
+// ====   =====
+//    0 : -127/512 (unused)
+//    1 : -126/512
+//    2 : -125/512
+//     ...
+//  128 : 0 
+//     ... 
+//  255 : 127/512
+//      : 1/4 (just outside the encoding range)
+//==============================================================================================================================
+ // Peak range for 32-bit and 16-bit operations.
+ #define A_BSC_32 (127.0)
+ #define A_BSC_16 (127.0/512.0)
+//==============================================================================================================================
+ #if 1
+  AU1 ABsc0ToU1(AU1 d,AF1 i){return (d&0xffffff00u)|((min(AU1(i+128.0),255u)    )&(0x000000ffu));}
+  AU1 ABsc1ToU1(AU1 d,AF1 i){return (d&0xffff00ffu)|((min(AU1(i+128.0),255u)<< 8)&(0x0000ff00u));}
+  AU1 ABsc2ToU1(AU1 d,AF1 i){return (d&0xff00ffffu)|((min(AU1(i+128.0),255u)<<16)&(0x00ff0000u));}
+  AU1 ABsc3ToU1(AU1 d,AF1 i){return (d&0x00ffffffu)|((min(AU1(i+128.0),255u)<<24)&(0xff000000u));}
+//------------------------------------------------------------------------------------------------------------------------------
+  AU1 ABsc0ToZbU1(AU1 d,AF1 i){return ((d&0xffffff00u)|((min(AU1(trunc(i)+128.0),255u)    )&(0x000000ffu)))^0x00000080u;}
+  AU1 ABsc1ToZbU1(AU1 d,AF1 i){return ((d&0xffff00ffu)|((min(AU1(trunc(i)+128.0),255u)<< 8)&(0x0000ff00u)))^0x00008000u;}
+  AU1 ABsc2ToZbU1(AU1 d,AF1 i){return ((d&0xff00ffffu)|((min(AU1(trunc(i)+128.0),255u)<<16)&(0x00ff0000u)))^0x00800000u;}
+  AU1 ABsc3ToZbU1(AU1 d,AF1 i){return ((d&0x00ffffffu)|((min(AU1(trunc(i)+128.0),255u)<<24)&(0xff000000u)))^0x80000000u;}
+//------------------------------------------------------------------------------------------------------------------------------
+  AF1 ABsc0FromU1(AU1 i){return AF1((i    )&255u)-128.0;}
+  AF1 ABsc1FromU1(AU1 i){return AF1((i>> 8)&255u)-128.0;}
+  AF1 ABsc2FromU1(AU1 i){return AF1((i>>16)&255u)-128.0;}
+  AF1 ABsc3FromU1(AU1 i){return AF1((i>>24)&255u)-128.0;}
+//------------------------------------------------------------------------------------------------------------------------------
+  AF1 ABsc0FromZbU1(AU1 i){return AF1(((i    )&255u)^0x80u)-128.0;}
+  AF1 ABsc1FromZbU1(AU1 i){return AF1(((i>> 8)&255u)^0x80u)-128.0;}
+  AF1 ABsc2FromZbU1(AU1 i){return AF1(((i>>16)&255u)^0x80u)-128.0;}
+  AF1 ABsc3FromZbU1(AU1 i){return AF1(((i>>24)&255u)^0x80u)-128.0;}
+ #endif
+//==============================================================================================================================
+ #ifdef A_HALF
+  // Takes {x0,x1} and {y0,y1} and builds {{x0,y0},{x1,y1}}.
+  AW2 ABsc01ToW2(AH2 x,AH2 y){x=x*AH2_(1.0/32768.0)+AH2_(0.25/32768.0);y=y*AH2_(1.0/32768.0)+AH2_(0.25/32768.0);
+   return AW2_AU1(APermGCEA(AU2(AU1_AW2(AW2_AH2(x)),AU1_AW2(AW2_AH2(y)))));}
+//------------------------------------------------------------------------------------------------------------------------------
+  AU2 ABsc0ToU2(AU2 d,AH2 i){AU1 b=AU1_AW2(AW2_AH2(i*AH2_(1.0/32768.0)+AH2_(0.25/32768.0)));
+   return AU2(APermHGFA(AU2(d.x,b)),APermHGFC(AU2(d.y,b)));}
+  AU2 ABsc1ToU2(AU2 d,AH2 i){AU1 b=AU1_AW2(AW2_AH2(i*AH2_(1.0/32768.0)+AH2_(0.25/32768.0)));
+   return AU2(APermHGAE(AU2(d.x,b)),APermHGCE(AU2(d.y,b)));}
+  AU2 ABsc2ToU2(AU2 d,AH2 i){AU1 b=AU1_AW2(AW2_AH2(i*AH2_(1.0/32768.0)+AH2_(0.25/32768.0)));
+   return AU2(APermHAFE(AU2(d.x,b)),APermHCFE(AU2(d.y,b)));}
+  AU2 ABsc3ToU2(AU2 d,AH2 i){AU1 b=AU1_AW2(AW2_AH2(i*AH2_(1.0/32768.0)+AH2_(0.25/32768.0)));
+   return AU2(APermAGFE(AU2(d.x,b)),APermCGFE(AU2(d.y,b)));}
+//------------------------------------------------------------------------------------------------------------------------------
+  AU2 ABsc0ToZbU2(AU2 d,AH2 i){AU1 b=AU1_AW2(AW2_AH2(i*AH2_(1.0/32768.0)+AH2_(0.25/32768.0)))^0x00800080u;
+   return AU2(APermHGFA(AU2(d.x,b)),APermHGFC(AU2(d.y,b)));}
+  AU2 ABsc1ToZbU2(AU2 d,AH2 i){AU1 b=AU1_AW2(AW2_AH2(i*AH2_(1.0/32768.0)+AH2_(0.25/32768.0)))^0x00800080u;
+   return AU2(APermHGAE(AU2(d.x,b)),APermHGCE(AU2(d.y,b)));}
+  AU2 ABsc2ToZbU2(AU2 d,AH2 i){AU1 b=AU1_AW2(AW2_AH2(i*AH2_(1.0/32768.0)+AH2_(0.25/32768.0)))^0x00800080u;
+   return AU2(APermHAFE(AU2(d.x,b)),APermHCFE(AU2(d.y,b)));}
+  AU2 ABsc3ToZbU2(AU2 d,AH2 i){AU1 b=AU1_AW2(AW2_AH2(i*AH2_(1.0/32768.0)+AH2_(0.25/32768.0)))^0x00800080u;
+   return AU2(APermAGFE(AU2(d.x,b)),APermCGFE(AU2(d.y,b)));}
+//------------------------------------------------------------------------------------------------------------------------------
+  AH2 ABsc0FromU2(AU2 i){return AH2_AW2(AW2_AU1(APerm0E0A(i)))*AH2_(32768.0)-AH2_(0.25);}
+  AH2 ABsc1FromU2(AU2 i){return AH2_AW2(AW2_AU1(APerm0F0B(i)))*AH2_(32768.0)-AH2_(0.25);}
+  AH2 ABsc2FromU2(AU2 i){return AH2_AW2(AW2_AU1(APerm0G0C(i)))*AH2_(32768.0)-AH2_(0.25);}
+  AH2 ABsc3FromU2(AU2 i){return AH2_AW2(AW2_AU1(APerm0H0D(i)))*AH2_(32768.0)-AH2_(0.25);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AH2 ABsc0FromZbU2(AU2 i){return AH2_AW2(AW2_AU1(APerm0E0A(i)^0x00800080u))*AH2_(32768.0)-AH2_(0.25);}
+  AH2 ABsc1FromZbU2(AU2 i){return AH2_AW2(AW2_AU1(APerm0F0B(i)^0x00800080u))*AH2_(32768.0)-AH2_(0.25);}
+  AH2 ABsc2FromZbU2(AU2 i){return AH2_AW2(AW2_AU1(APerm0G0C(i)^0x00800080u))*AH2_(32768.0)-AH2_(0.25);}
+  AH2 ABsc3FromZbU2(AU2 i){return AH2_AW2(AW2_AU1(APerm0H0D(i)^0x00800080u))*AH2_(32768.0)-AH2_(0.25);}
+ #endif
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                     HALF APPROXIMATIONS
+//------------------------------------------------------------------------------------------------------------------------------
+// These support only positive inputs.
+// Did not see value yet in specialization for range.
+// Using quick testing, ended up mostly getting the same "best" approximation for various ranges.
+// With hardware that can co-execute transcendentals, the value in approximations could be less than expected.
+// However from a latency perspective, if execution of a transcendental is 4 clk, with no packed support, -> 8 clk total.
+// And co-execution would require a compiler interleaving a lot of independent work for packed usage.
+//------------------------------------------------------------------------------------------------------------------------------
+// The one Newton Raphson iteration form of rsq() was skipped (requires 6 ops total).
+// Same with sqrt(), as this could be x*rsq() (7 ops).
+//==============================================================================================================================
+ #ifdef A_HALF
+  // Minimize squared error across full positive range, 2 ops.
+  // The 0x1de2 based approximation maps {0 to 1} input maps to < 1 output.
+  AH1 APrxLoSqrtH1(AH1 a){return AH1_AW1((AW1_AH1(a)>>AW1_(1))+AW1_(0x1de2));}
+  AH2 APrxLoSqrtH2(AH2 a){return AH2_AW2((AW2_AH2(a)>>AW2_(1))+AW2_(0x1de2));}
+  AH3 APrxLoSqrtH3(AH3 a){return AH3_AW3((AW3_AH3(a)>>AW3_(1))+AW3_(0x1de2));}
+  AH4 APrxLoSqrtH4(AH4 a){return AH4_AW4((AW4_AH4(a)>>AW4_(1))+AW4_(0x1de2));}
+//------------------------------------------------------------------------------------------------------------------------------
+  // Lower precision estimation, 1 op.
+  // Minimize squared error across {smallest normal to 16384.0}.
+  AH1 APrxLoRcpH1(AH1 a){return AH1_AW1(AW1_(0x7784)-AW1_AH1(a));}
+  AH2 APrxLoRcpH2(AH2 a){return AH2_AW2(AW2_(0x7784)-AW2_AH2(a));}
+  AH3 APrxLoRcpH3(AH3 a){return AH3_AW3(AW3_(0x7784)-AW3_AH3(a));}
+  AH4 APrxLoRcpH4(AH4 a){return AH4_AW4(AW4_(0x7784)-AW4_AH4(a));}
+//------------------------------------------------------------------------------------------------------------------------------
+  // Medium precision estimation, one Newton Raphson iteration, 3 ops.
+  AH1 APrxMedRcpH1(AH1 a){AH1 b=AH1_AW1(AW1_(0x778d)-AW1_AH1(a));return b*(-b*a+AH1_(2.0));}
+  AH2 APrxMedRcpH2(AH2 a){AH2 b=AH2_AW2(AW2_(0x778d)-AW2_AH2(a));return b*(-b*a+AH2_(2.0));}
+  AH3 APrxMedRcpH3(AH3 a){AH3 b=AH3_AW3(AW3_(0x778d)-AW3_AH3(a));return b*(-b*a+AH3_(2.0));}
+  AH4 APrxMedRcpH4(AH4 a){AH4 b=AH4_AW4(AW4_(0x778d)-AW4_AH4(a));return b*(-b*a+AH4_(2.0));}
+//------------------------------------------------------------------------------------------------------------------------------
+  // Minimize squared error across {smallest normal to 16384.0}, 2 ops.
+  AH1 APrxLoRsqH1(AH1 a){return AH1_AW1(AW1_(0x59a3)-(AW1_AH1(a)>>AW1_(1)));}
+  AH2 APrxLoRsqH2(AH2 a){return AH2_AW2(AW2_(0x59a3)-(AW2_AH2(a)>>AW2_(1)));}
+  AH3 APrxLoRsqH3(AH3 a){return AH3_AW3(AW3_(0x59a3)-(AW3_AH3(a)>>AW3_(1)));}
+  AH4 APrxLoRsqH4(AH4 a){return AH4_AW4(AW4_(0x59a3)-(AW4_AH4(a)>>AW4_(1)));}
+ #endif
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                    FLOAT APPROXIMATIONS
+//------------------------------------------------------------------------------------------------------------------------------
+// Michal Drobot has an excellent presentation on these: "Low Level Optimizations For GCN",
+//  - Idea dates back to SGI, then to Quake 3, etc.
+//  - https://michaldrobot.files.wordpress.com/2014/05/gcn_alu_opt_digitaldragons2014.pdf
+//     - sqrt(x)=rsqrt(x)*x
+//     - rcp(x)=rsqrt(x)*rsqrt(x) for positive x
+//  - https://github.com/michaldrobot/ShaderFastLibs/blob/master/ShaderFastMathLib.h
+//------------------------------------------------------------------------------------------------------------------------------
+// These below are from perhaps less complete searching for optimal.
+// Used FP16 normal range for testing with +4096 32-bit step size for sampling error.
+// So these match up well with the half approximations.
+//==============================================================================================================================
+ AF1 APrxLoSqrtF1(AF1 a){return AF1_AU1((AU1_AF1(a)>>AU1_(1))+AU1_(0x1fbc4639));}
+ AF1 APrxLoRcpF1(AF1 a){return AF1_AU1(AU1_(0x7ef07ebb)-AU1_AF1(a));}
+ AF1 APrxMedRcpF1(AF1 a){AF1 b=AF1_AU1(AU1_(0x7ef19fff)-AU1_AF1(a));return b*(-b*a+AF1_(2.0));}
+ AF1 APrxLoRsqF1(AF1 a){return AF1_AU1(AU1_(0x5f347d74)-(AU1_AF1(a)>>AU1_(1)));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF2 APrxLoSqrtF2(AF2 a){return AF2_AU2((AU2_AF2(a)>>AU2_(1))+AU2_(0x1fbc4639));}
+ AF2 APrxLoRcpF2(AF2 a){return AF2_AU2(AU2_(0x7ef07ebb)-AU2_AF2(a));}
+ AF2 APrxMedRcpF2(AF2 a){AF2 b=AF2_AU2(AU2_(0x7ef19fff)-AU2_AF2(a));return b*(-b*a+AF2_(2.0));}
+ AF2 APrxLoRsqF2(AF2 a){return AF2_AU2(AU2_(0x5f347d74)-(AU2_AF2(a)>>AU2_(1)));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF3 APrxLoSqrtF3(AF3 a){return AF3_AU3((AU3_AF3(a)>>AU3_(1))+AU3_(0x1fbc4639));}
+ AF3 APrxLoRcpF3(AF3 a){return AF3_AU3(AU3_(0x7ef07ebb)-AU3_AF3(a));}
+ AF3 APrxMedRcpF3(AF3 a){AF3 b=AF3_AU3(AU3_(0x7ef19fff)-AU3_AF3(a));return b*(-b*a+AF3_(2.0));}
+ AF3 APrxLoRsqF3(AF3 a){return AF3_AU3(AU3_(0x5f347d74)-(AU3_AF3(a)>>AU3_(1)));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF4 APrxLoSqrtF4(AF4 a){return AF4_AU4((AU4_AF4(a)>>AU4_(1))+AU4_(0x1fbc4639));}
+ AF4 APrxLoRcpF4(AF4 a){return AF4_AU4(AU4_(0x7ef07ebb)-AU4_AF4(a));}
+ AF4 APrxMedRcpF4(AF4 a){AF4 b=AF4_AU4(AU4_(0x7ef19fff)-AU4_AF4(a));return b*(-b*a+AF4_(2.0));}
+ AF4 APrxLoRsqF4(AF4 a){return AF4_AU4(AU4_(0x5f347d74)-(AU4_AF4(a)>>AU4_(1)));}
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                    PQ APPROXIMATIONS
+//------------------------------------------------------------------------------------------------------------------------------
+// PQ is very close to x^(1/8). The functions below Use the fast float approximation method to do
+// PQ<~>Gamma2 (4th power and fast 4th root) and PQ<~>Linear (8th power and fast 8th root). Maximum error is ~0.2%.
+//==============================================================================================================================
+// Helpers
+ AF1 Quart(AF1 a) { a = a * a; return a * a;}
+ AF1 Oct(AF1 a) { a = a * a; a = a * a; return a * a; }
+ AF2 Quart(AF2 a) { a = a * a; return a * a; }
+ AF2 Oct(AF2 a) { a = a * a; a = a * a; return a * a; }
+ AF3 Quart(AF3 a) { a = a * a; return a * a; }
+ AF3 Oct(AF3 a) { a = a * a; a = a * a; return a * a; }
+ AF4 Quart(AF4 a) { a = a * a; return a * a; }
+ AF4 Oct(AF4 a) { a = a * a; a = a * a; return a * a; }
+ //------------------------------------------------------------------------------------------------------------------------------
+ AF1 APrxPQToGamma2(AF1 a) { return Quart(a); }
+ AF1 APrxPQToLinear(AF1 a) { return Oct(a); }
+ AF1 APrxLoGamma2ToPQ(AF1 a) { return AF1_AU1((AU1_AF1(a) >> AU1_(2)) + AU1_(0x2F9A4E46)); }
+ AF1 APrxMedGamma2ToPQ(AF1 a) { AF1 b = AF1_AU1((AU1_AF1(a) >> AU1_(2)) + AU1_(0x2F9A4E46)); AF1 b4 = Quart(b); return b - b * (b4 - a) / (AF1_(4.0) * b4); }
+ AF1 APrxHighGamma2ToPQ(AF1 a) { return sqrt(sqrt(a)); }
+ AF1 APrxLoLinearToPQ(AF1 a) { return AF1_AU1((AU1_AF1(a) >> AU1_(3)) + AU1_(0x378D8723)); }
+ AF1 APrxMedLinearToPQ(AF1 a) { AF1 b = AF1_AU1((AU1_AF1(a) >> AU1_(3)) + AU1_(0x378D8723)); AF1 b8 = Oct(b); return b - b * (b8 - a) / (AF1_(8.0) * b8); }
+ AF1 APrxHighLinearToPQ(AF1 a) { return sqrt(sqrt(sqrt(a))); }
+ //------------------------------------------------------------------------------------------------------------------------------
+ AF2 APrxPQToGamma2(AF2 a) { return Quart(a); }
+ AF2 APrxPQToLinear(AF2 a) { return Oct(a); }
+ AF2 APrxLoGamma2ToPQ(AF2 a) { return AF2_AU2((AU2_AF2(a) >> AU2_(2)) + AU2_(0x2F9A4E46)); }
+ AF2 APrxMedGamma2ToPQ(AF2 a) { AF2 b = AF2_AU2((AU2_AF2(a) >> AU2_(2)) + AU2_(0x2F9A4E46)); AF2 b4 = Quart(b); return b - b * (b4 - a) / (AF1_(4.0) * b4); }
+ AF2 APrxHighGamma2ToPQ(AF2 a) { return sqrt(sqrt(a)); }
+ AF2 APrxLoLinearToPQ(AF2 a) { return AF2_AU2((AU2_AF2(a) >> AU2_(3)) + AU2_(0x378D8723)); }
+ AF2 APrxMedLinearToPQ(AF2 a) { AF2 b = AF2_AU2((AU2_AF2(a) >> AU2_(3)) + AU2_(0x378D8723)); AF2 b8 = Oct(b); return b - b * (b8 - a) / (AF1_(8.0) * b8); }
+ AF2 APrxHighLinearToPQ(AF2 a) { return sqrt(sqrt(sqrt(a))); }
+ //------------------------------------------------------------------------------------------------------------------------------
+ AF3 APrxPQToGamma2(AF3 a) { return Quart(a); }
+ AF3 APrxPQToLinear(AF3 a) { return Oct(a); }
+ AF3 APrxLoGamma2ToPQ(AF3 a) { return AF3_AU3((AU3_AF3(a) >> AU3_(2)) + AU3_(0x2F9A4E46)); }
+ AF3 APrxMedGamma2ToPQ(AF3 a) { AF3 b = AF3_AU3((AU3_AF3(a) >> AU3_(2)) + AU3_(0x2F9A4E46)); AF3 b4 = Quart(b); return b - b * (b4 - a) / (AF1_(4.0) * b4); }
+ AF3 APrxHighGamma2ToPQ(AF3 a) { return sqrt(sqrt(a)); }
+ AF3 APrxLoLinearToPQ(AF3 a) { return AF3_AU3((AU3_AF3(a) >> AU3_(3)) + AU3_(0x378D8723)); }
+ AF3 APrxMedLinearToPQ(AF3 a) { AF3 b = AF3_AU3((AU3_AF3(a) >> AU3_(3)) + AU3_(0x378D8723)); AF3 b8 = Oct(b); return b - b * (b8 - a) / (AF1_(8.0) * b8); }
+ AF3 APrxHighLinearToPQ(AF3 a) { return sqrt(sqrt(sqrt(a))); }
+ //------------------------------------------------------------------------------------------------------------------------------
+ AF4 APrxPQToGamma2(AF4 a) { return Quart(a); }
+ AF4 APrxPQToLinear(AF4 a) { return Oct(a); }
+ AF4 APrxLoGamma2ToPQ(AF4 a) { return AF4_AU4((AU4_AF4(a) >> AU4_(2)) + AU4_(0x2F9A4E46)); }
+ AF4 APrxMedGamma2ToPQ(AF4 a) { AF4 b = AF4_AU4((AU4_AF4(a) >> AU4_(2)) + AU4_(0x2F9A4E46)); AF4 b4 = Quart(b); return b - b * (b4 - a) / (AF1_(4.0) * b4); }
+ AF4 APrxHighGamma2ToPQ(AF4 a) { return sqrt(sqrt(a)); }
+ AF4 APrxLoLinearToPQ(AF4 a) { return AF4_AU4((AU4_AF4(a) >> AU4_(3)) + AU4_(0x378D8723)); }
+ AF4 APrxMedLinearToPQ(AF4 a) { AF4 b = AF4_AU4((AU4_AF4(a) >> AU4_(3)) + AU4_(0x378D8723)); AF4 b8 = Oct(b); return b - b * (b8 - a) / (AF1_(8.0) * b8); }
+ AF4 APrxHighLinearToPQ(AF4 a) { return sqrt(sqrt(sqrt(a))); }
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                    PARABOLIC SIN & COS
+//------------------------------------------------------------------------------------------------------------------------------
+// Approximate answers to transcendental questions.
+//------------------------------------------------------------------------------------------------------------------------------
+//==============================================================================================================================
+ #if 1
+  // Valid input range is {-1 to 1} representing {0 to 2 pi}.
+  // Output range is {-1/4 to 1/4} representing {-1 to 1}.
+  AF1 APSinF1(AF1 x){return x*abs(x)-x;} // MAD.
+  AF2 APSinF2(AF2 x){return x*abs(x)-x;}
+  AF1 APCosF1(AF1 x){x=AFractF1(x*AF1_(0.5)+AF1_(0.75));x=x*AF1_(2.0)-AF1_(1.0);return APSinF1(x);} // 3x MAD, FRACT
+  AF2 APCosF2(AF2 x){x=AFractF2(x*AF2_(0.5)+AF2_(0.75));x=x*AF2_(2.0)-AF2_(1.0);return APSinF2(x);}
+  AF2 APSinCosF1(AF1 x){AF1 y=AFractF1(x*AF1_(0.5)+AF1_(0.75));y=y*AF1_(2.0)-AF1_(1.0);return APSinF2(AF2(x,y));}
+ #endif
+//------------------------------------------------------------------------------------------------------------------------------
+ #ifdef A_HALF
+  // For a packed {sin,cos} pair,
+  //  - Native takes 16 clocks and 4 issue slots (no packed transcendentals).
+  //  - Parabolic takes 8 clocks and 8 issue slots (only fract is non-packed).
+  AH1 APSinH1(AH1 x){return x*abs(x)-x;}
+  AH2 APSinH2(AH2 x){return x*abs(x)-x;} // AND,FMA
+  AH1 APCosH1(AH1 x){x=AFractH1(x*AH1_(0.5)+AH1_(0.75));x=x*AH1_(2.0)-AH1_(1.0);return APSinH1(x);} 
+  AH2 APCosH2(AH2 x){x=AFractH2(x*AH2_(0.5)+AH2_(0.75));x=x*AH2_(2.0)-AH2_(1.0);return APSinH2(x);} // 3x FMA, 2xFRACT, AND
+  AH2 APSinCosH1(AH1 x){AH1 y=AFractH1(x*AH1_(0.5)+AH1_(0.75));y=y*AH1_(2.0)-AH1_(1.0);return APSinH2(AH2(x,y));}
+ #endif
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                     [ZOL] ZERO ONE LOGIC
+//------------------------------------------------------------------------------------------------------------------------------
+// Conditional free logic designed for easy 16-bit packing, and backwards porting to 32-bit.
+//------------------------------------------------------------------------------------------------------------------------------
+// 0 := false
+// 1 := true
+//------------------------------------------------------------------------------------------------------------------------------
+// AndNot(x,y)   -> !(x&y) .... One op.
+// AndOr(x,y,z)  -> (x&y)|z ... One op.
+// GtZero(x)     -> x>0.0 ..... One op.
+// Sel(x,y,z)    -> x?y:z ..... Two ops, has no precision loss.
+// Signed(x)     -> x<0.0 ..... One op.
+// ZeroPass(x,y) -> x?0:y ..... Two ops, 'y' is a pass through safe for aliasing as integer.
+//------------------------------------------------------------------------------------------------------------------------------
+// OPTIMIZATION NOTES
+// ==================
+// - On Vega to use 2 constants in a packed op, pass in as one AW2 or one AH2 'k.xy' and use as 'k.xx' and 'k.yy'.
+//   For example 'a.xy*k.xx+k.yy'.
+//==============================================================================================================================
+ #if 1
+  AU1 AZolAndU1(AU1 x,AU1 y){return min(x,y);}
+  AU2 AZolAndU2(AU2 x,AU2 y){return min(x,y);}
+  AU3 AZolAndU3(AU3 x,AU3 y){return min(x,y);}
+  AU4 AZolAndU4(AU4 x,AU4 y){return min(x,y);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AU1 AZolNotU1(AU1 x){return x^AU1_(1);}
+  AU2 AZolNotU2(AU2 x){return x^AU2_(1);}
+  AU3 AZolNotU3(AU3 x){return x^AU3_(1);}
+  AU4 AZolNotU4(AU4 x){return x^AU4_(1);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AU1 AZolOrU1(AU1 x,AU1 y){return max(x,y);}
+  AU2 AZolOrU2(AU2 x,AU2 y){return max(x,y);}
+  AU3 AZolOrU3(AU3 x,AU3 y){return max(x,y);}
+  AU4 AZolOrU4(AU4 x,AU4 y){return max(x,y);}
+//==============================================================================================================================
+  AU1 AZolF1ToU1(AF1 x){return AU1(x);}
+  AU2 AZolF2ToU2(AF2 x){return AU2(x);}
+  AU3 AZolF3ToU3(AF3 x){return AU3(x);}
+  AU4 AZolF4ToU4(AF4 x){return AU4(x);}
+//------------------------------------------------------------------------------------------------------------------------------
+  // 2 ops, denormals don't work in 32-bit on PC (and if they are enabled, OMOD is disabled).
+  AU1 AZolNotF1ToU1(AF1 x){return AU1(AF1_(1.0)-x);}
+  AU2 AZolNotF2ToU2(AF2 x){return AU2(AF2_(1.0)-x);}
+  AU3 AZolNotF3ToU3(AF3 x){return AU3(AF3_(1.0)-x);}
+  AU4 AZolNotF4ToU4(AF4 x){return AU4(AF4_(1.0)-x);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AF1 AZolU1ToF1(AU1 x){return AF1(x);}
+  AF2 AZolU2ToF2(AU2 x){return AF2(x);}
+  AF3 AZolU3ToF3(AU3 x){return AF3(x);}
+  AF4 AZolU4ToF4(AU4 x){return AF4(x);}
+//==============================================================================================================================
+  AF1 AZolAndF1(AF1 x,AF1 y){return min(x,y);}
+  AF2 AZolAndF2(AF2 x,AF2 y){return min(x,y);}
+  AF3 AZolAndF3(AF3 x,AF3 y){return min(x,y);}
+  AF4 AZolAndF4(AF4 x,AF4 y){return min(x,y);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AF1 ASolAndNotF1(AF1 x,AF1 y){return (-x)*y+AF1_(1.0);}
+  AF2 ASolAndNotF2(AF2 x,AF2 y){return (-x)*y+AF2_(1.0);}
+  AF3 ASolAndNotF3(AF3 x,AF3 y){return (-x)*y+AF3_(1.0);}
+  AF4 ASolAndNotF4(AF4 x,AF4 y){return (-x)*y+AF4_(1.0);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AF1 AZolAndOrF1(AF1 x,AF1 y,AF1 z){return ASatF1(x*y+z);}
+  AF2 AZolAndOrF2(AF2 x,AF2 y,AF2 z){return ASatF2(x*y+z);}
+  AF3 AZolAndOrF3(AF3 x,AF3 y,AF3 z){return ASatF3(x*y+z);}
+  AF4 AZolAndOrF4(AF4 x,AF4 y,AF4 z){return ASatF4(x*y+z);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AF1 AZolGtZeroF1(AF1 x){return ASatF1(x*AF1_(A_INFP_F));}
+  AF2 AZolGtZeroF2(AF2 x){return ASatF2(x*AF2_(A_INFP_F));}
+  AF3 AZolGtZeroF3(AF3 x){return ASatF3(x*AF3_(A_INFP_F));}
+  AF4 AZolGtZeroF4(AF4 x){return ASatF4(x*AF4_(A_INFP_F));}
+//------------------------------------------------------------------------------------------------------------------------------
+  AF1 AZolNotF1(AF1 x){return AF1_(1.0)-x;}
+  AF2 AZolNotF2(AF2 x){return AF2_(1.0)-x;}
+  AF3 AZolNotF3(AF3 x){return AF3_(1.0)-x;}
+  AF4 AZolNotF4(AF4 x){return AF4_(1.0)-x;}
+//------------------------------------------------------------------------------------------------------------------------------
+  AF1 AZolOrF1(AF1 x,AF1 y){return max(x,y);}
+  AF2 AZolOrF2(AF2 x,AF2 y){return max(x,y);}
+  AF3 AZolOrF3(AF3 x,AF3 y){return max(x,y);}
+  AF4 AZolOrF4(AF4 x,AF4 y){return max(x,y);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AF1 AZolSelF1(AF1 x,AF1 y,AF1 z){AF1 r=(-x)*z+z;return x*y+r;}
+  AF2 AZolSelF2(AF2 x,AF2 y,AF2 z){AF2 r=(-x)*z+z;return x*y+r;}
+  AF3 AZolSelF3(AF3 x,AF3 y,AF3 z){AF3 r=(-x)*z+z;return x*y+r;}
+  AF4 AZolSelF4(AF4 x,AF4 y,AF4 z){AF4 r=(-x)*z+z;return x*y+r;}
+//------------------------------------------------------------------------------------------------------------------------------
+  AF1 AZolSignedF1(AF1 x){return ASatF1(x*AF1_(A_INFN_F));}
+  AF2 AZolSignedF2(AF2 x){return ASatF2(x*AF2_(A_INFN_F));}
+  AF3 AZolSignedF3(AF3 x){return ASatF3(x*AF3_(A_INFN_F));}
+  AF4 AZolSignedF4(AF4 x){return ASatF4(x*AF4_(A_INFN_F));}
+//------------------------------------------------------------------------------------------------------------------------------
+  AF1 AZolZeroPassF1(AF1 x,AF1 y){return AF1_AU1((AU1_AF1(x)!=AU1_(0))?AU1_(0):AU1_AF1(y));}
+  AF2 AZolZeroPassF2(AF2 x,AF2 y){return AF2_AU2((AU2_AF2(x)!=AU2_(0))?AU2_(0):AU2_AF2(y));}
+  AF3 AZolZeroPassF3(AF3 x,AF3 y){return AF3_AU3((AU3_AF3(x)!=AU3_(0))?AU3_(0):AU3_AF3(y));}
+  AF4 AZolZeroPassF4(AF4 x,AF4 y){return AF4_AU4((AU4_AF4(x)!=AU4_(0))?AU4_(0):AU4_AF4(y));}
+ #endif
+//==============================================================================================================================
+ #ifdef A_HALF
+  AW1 AZolAndW1(AW1 x,AW1 y){return min(x,y);}
+  AW2 AZolAndW2(AW2 x,AW2 y){return min(x,y);}
+  AW3 AZolAndW3(AW3 x,AW3 y){return min(x,y);}
+  AW4 AZolAndW4(AW4 x,AW4 y){return min(x,y);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AW1 AZolNotW1(AW1 x){return x^AW1_(1);}
+  AW2 AZolNotW2(AW2 x){return x^AW2_(1);}
+  AW3 AZolNotW3(AW3 x){return x^AW3_(1);}
+  AW4 AZolNotW4(AW4 x){return x^AW4_(1);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AW1 AZolOrW1(AW1 x,AW1 y){return max(x,y);}
+  AW2 AZolOrW2(AW2 x,AW2 y){return max(x,y);}
+  AW3 AZolOrW3(AW3 x,AW3 y){return max(x,y);}
+  AW4 AZolOrW4(AW4 x,AW4 y){return max(x,y);}
+//==============================================================================================================================
+  // Uses denormal trick.
+  AW1 AZolH1ToW1(AH1 x){return AW1_AH1(x*AH1_AW1(AW1_(1)));}
+  AW2 AZolH2ToW2(AH2 x){return AW2_AH2(x*AH2_AW2(AW2_(1)));}
+  AW3 AZolH3ToW3(AH3 x){return AW3_AH3(x*AH3_AW3(AW3_(1)));}
+  AW4 AZolH4ToW4(AH4 x){return AW4_AH4(x*AH4_AW4(AW4_(1)));}
+//------------------------------------------------------------------------------------------------------------------------------
+  // AMD arch lacks a packed conversion opcode.
+  AH1 AZolW1ToH1(AW1 x){return AH1_AW1(x*AW1_AH1(AH1_(1.0)));}
+  AH2 AZolW2ToH2(AW2 x){return AH2_AW2(x*AW2_AH2(AH2_(1.0)));}
+  AH3 AZolW1ToH3(AW3 x){return AH3_AW3(x*AW3_AH3(AH3_(1.0)));}
+  AH4 AZolW2ToH4(AW4 x){return AH4_AW4(x*AW4_AH4(AH4_(1.0)));}
+//==============================================================================================================================
+  AH1 AZolAndH1(AH1 x,AH1 y){return min(x,y);}
+  AH2 AZolAndH2(AH2 x,AH2 y){return min(x,y);}
+  AH3 AZolAndH3(AH3 x,AH3 y){return min(x,y);}
+  AH4 AZolAndH4(AH4 x,AH4 y){return min(x,y);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AH1 ASolAndNotH1(AH1 x,AH1 y){return (-x)*y+AH1_(1.0);}
+  AH2 ASolAndNotH2(AH2 x,AH2 y){return (-x)*y+AH2_(1.0);}
+  AH3 ASolAndNotH3(AH3 x,AH3 y){return (-x)*y+AH3_(1.0);}
+  AH4 ASolAndNotH4(AH4 x,AH4 y){return (-x)*y+AH4_(1.0);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AH1 AZolAndOrH1(AH1 x,AH1 y,AH1 z){return ASatH1(x*y+z);}
+  AH2 AZolAndOrH2(AH2 x,AH2 y,AH2 z){return ASatH2(x*y+z);}
+  AH3 AZolAndOrH3(AH3 x,AH3 y,AH3 z){return ASatH3(x*y+z);}
+  AH4 AZolAndOrH4(AH4 x,AH4 y,AH4 z){return ASatH4(x*y+z);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AH1 AZolGtZeroH1(AH1 x){return ASatH1(x*AH1_(A_INFP_H));}
+  AH2 AZolGtZeroH2(AH2 x){return ASatH2(x*AH2_(A_INFP_H));}
+  AH3 AZolGtZeroH3(AH3 x){return ASatH3(x*AH3_(A_INFP_H));}
+  AH4 AZolGtZeroH4(AH4 x){return ASatH4(x*AH4_(A_INFP_H));}
+//------------------------------------------------------------------------------------------------------------------------------
+  AH1 AZolNotH1(AH1 x){return AH1_(1.0)-x;}
+  AH2 AZolNotH2(AH2 x){return AH2_(1.0)-x;}
+  AH3 AZolNotH3(AH3 x){return AH3_(1.0)-x;}
+  AH4 AZolNotH4(AH4 x){return AH4_(1.0)-x;}
+//------------------------------------------------------------------------------------------------------------------------------
+  AH1 AZolOrH1(AH1 x,AH1 y){return max(x,y);}
+  AH2 AZolOrH2(AH2 x,AH2 y){return max(x,y);}
+  AH3 AZolOrH3(AH3 x,AH3 y){return max(x,y);}
+  AH4 AZolOrH4(AH4 x,AH4 y){return max(x,y);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AH1 AZolSelH1(AH1 x,AH1 y,AH1 z){AH1 r=(-x)*z+z;return x*y+r;}
+  AH2 AZolSelH2(AH2 x,AH2 y,AH2 z){AH2 r=(-x)*z+z;return x*y+r;}
+  AH3 AZolSelH3(AH3 x,AH3 y,AH3 z){AH3 r=(-x)*z+z;return x*y+r;}
+  AH4 AZolSelH4(AH4 x,AH4 y,AH4 z){AH4 r=(-x)*z+z;return x*y+r;}
+//------------------------------------------------------------------------------------------------------------------------------
+  AH1 AZolSignedH1(AH1 x){return ASatH1(x*AH1_(A_INFN_H));}
+  AH2 AZolSignedH2(AH2 x){return ASatH2(x*AH2_(A_INFN_H));}
+  AH3 AZolSignedH3(AH3 x){return ASatH3(x*AH3_(A_INFN_H));}
+  AH4 AZolSignedH4(AH4 x){return ASatH4(x*AH4_(A_INFN_H));}
+ #endif
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                      COLOR CONVERSIONS
+//------------------------------------------------------------------------------------------------------------------------------
+// These are all linear to/from some other space (where 'linear' has been shortened out of the function name).
+// So 'ToGamma' is 'LinearToGamma', and 'FromGamma' is 'LinearFromGamma'.
+// These are branch free implementations.
+// The AToSrgbF1() function is useful for stores for compute shaders for GPUs without hardware linear->sRGB store conversion.
+//------------------------------------------------------------------------------------------------------------------------------
+// TRANSFER FUNCTIONS
+// ==================
+// 709 ..... Rec709 used for some HDTVs
+// Gamma ... Typically 2.2 for some PC displays, or 2.4-2.5 for CRTs, or 2.2 FreeSync2 native
+// Pq ...... PQ native for HDR10
+// Srgb .... The sRGB output, typical of PC displays, useful for 10-bit output, or storing to 8-bit UNORM without SRGB type
+// Two ..... Gamma 2.0, fastest conversion (useful for intermediate pass approximations)
+// Three ... Gamma 3.0, less fast, but good for HDR.
+//------------------------------------------------------------------------------------------------------------------------------
+// KEEPING TO SPEC
+// ===============
+// Both Rec.709 and sRGB have a linear segment which as spec'ed would intersect the curved segment 2 times.
+//  (a.) For 8-bit sRGB, steps {0 to 10.3} are in the linear region (4% of the encoding range).
+//  (b.) For 8-bit  709, steps {0 to 20.7} are in the linear region (8% of the encoding range).
+// Also there is a slight step in the transition regions.
+// Precision of the coefficients in the spec being the likely cause.
+// Main usage case of the sRGB code is to do the linear->sRGB converstion in a compute shader before store.
+// This is to work around lack of hardware (typically only ROP does the conversion for free).
+// To "correct" the linear segment, would be to introduce error, because hardware decode of sRGB->linear is fixed (and free).
+// So this header keeps with the spec.
+// For linear->sRGB transforms, the linear segment in some respects reduces error, because rounding in that region is linear.
+// Rounding in the curved region in hardware (and fast software code) introduces error due to rounding in non-linear.
+//------------------------------------------------------------------------------------------------------------------------------
+// FOR PQ
+// ======
+// Both input and output is {0.0-1.0}, and where output 1.0 represents 10000.0 cd/m^2.
+// All constants are only specified to FP32 precision.
+// External PQ source reference,
+//  - https://github.com/ampas/aces-dev/blob/master/transforms/ctl/utilities/ACESlib.Utilities_Color.a1.0.1.ctl
+//------------------------------------------------------------------------------------------------------------------------------
+// PACKED VERSIONS
+// ===============
+// These are the A*H2() functions.
+// There is no PQ functions as FP16 seemed to not have enough precision for the conversion.
+// The remaining functions are "good enough" for 8-bit, and maybe 10-bit if not concerned about a few 1-bit errors.
+// Precision is lowest in the 709 conversion, higher in sRGB, higher still in Two and Gamma (when using 2.2 at least).
+//------------------------------------------------------------------------------------------------------------------------------
+// NOTES
+// =====
+// Could be faster for PQ conversions to be in ALU or a texture lookup depending on usage case.
+//==============================================================================================================================
+ #if 1
+  AF1 ATo709F1(AF1 c){AF3 j=AF3(0.018*4.5,4.5,0.45);AF2 k=AF2(1.099,-0.099);
+   return clamp(j.x  ,c*j.y  ,pow(c,j.z  )*k.x  +k.y  );}
+  AF2 ATo709F2(AF2 c){AF3 j=AF3(0.018*4.5,4.5,0.45);AF2 k=AF2(1.099,-0.099);
+   return clamp(j.xx ,c*j.yy ,pow(c,j.zz )*k.xx +k.yy );}
+  AF3 ATo709F3(AF3 c){AF3 j=AF3(0.018*4.5,4.5,0.45);AF2 k=AF2(1.099,-0.099);
+   return clamp(j.xxx,c*j.yyy,pow(c,j.zzz)*k.xxx+k.yyy);}
+//------------------------------------------------------------------------------------------------------------------------------
+  // Note 'rcpX' is '1/x', where the 'x' is what would be used in AFromGamma().
+  AF1 AToGammaF1(AF1 c,AF1 rcpX){return pow(c,AF1_(rcpX));} 
+  AF2 AToGammaF2(AF2 c,AF1 rcpX){return pow(c,AF2_(rcpX));} 
+  AF3 AToGammaF3(AF3 c,AF1 rcpX){return pow(c,AF3_(rcpX));} 
+//------------------------------------------------------------------------------------------------------------------------------
+  AF1 AToPqF1(AF1 x){AF1 p=pow(x,AF1_(0.159302));
+   return pow((AF1_(0.835938)+AF1_(18.8516)*p)/(AF1_(1.0)+AF1_(18.6875)*p),AF1_(78.8438));}
+  AF2 AToPqF1(AF2 x){AF2 p=pow(x,AF2_(0.159302));
+   return pow((AF2_(0.835938)+AF2_(18.8516)*p)/(AF2_(1.0)+AF2_(18.6875)*p),AF2_(78.8438));}
+  AF3 AToPqF1(AF3 x){AF3 p=pow(x,AF3_(0.159302));
+   return pow((AF3_(0.835938)+AF3_(18.8516)*p)/(AF3_(1.0)+AF3_(18.6875)*p),AF3_(78.8438));}
+//------------------------------------------------------------------------------------------------------------------------------
+  AF1 AToSrgbF1(AF1 c){AF3 j=AF3(0.0031308*12.92,12.92,1.0/2.4);AF2 k=AF2(1.055,-0.055);
+   return clamp(j.x  ,c*j.y  ,pow(c,j.z  )*k.x  +k.y  );}
+  AF2 AToSrgbF2(AF2 c){AF3 j=AF3(0.0031308*12.92,12.92,1.0/2.4);AF2 k=AF2(1.055,-0.055);
+   return clamp(j.xx ,c*j.yy ,pow(c,j.zz )*k.xx +k.yy );}
+  AF3 AToSrgbF3(AF3 c){AF3 j=AF3(0.0031308*12.92,12.92,1.0/2.4);AF2 k=AF2(1.055,-0.055);
+   return clamp(j.xxx,c*j.yyy,pow(c,j.zzz)*k.xxx+k.yyy);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AF1 AToTwoF1(AF1 c){return sqrt(c);}
+  AF2 AToTwoF2(AF2 c){return sqrt(c);}
+  AF3 AToTwoF3(AF3 c){return sqrt(c);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AF1 AToThreeF1(AF1 c){return pow(c,AF1_(1.0/3.0));}
+  AF2 AToThreeF2(AF2 c){return pow(c,AF2_(1.0/3.0));}
+  AF3 AToThreeF3(AF3 c){return pow(c,AF3_(1.0/3.0));}
+ #endif
+//==============================================================================================================================
+ #if 1
+  // Unfortunately median won't work here.
+  AF1 AFrom709F1(AF1 c){AF3 j=AF3(0.081/4.5,1.0/4.5,1.0/0.45);AF2 k=AF2(1.0/1.099,0.099/1.099);
+   return AZolSelF1(AZolSignedF1(c-j.x  ),c*j.y  ,pow(c*k.x  +k.y  ,j.z  ));}
+  AF2 AFrom709F2(AF2 c){AF3 j=AF3(0.081/4.5,1.0/4.5,1.0/0.45);AF2 k=AF2(1.0/1.099,0.099/1.099);
+   return AZolSelF2(AZolSignedF2(c-j.xx ),c*j.yy ,pow(c*k.xx +k.yy ,j.zz ));}
+  AF3 AFrom709F3(AF3 c){AF3 j=AF3(0.081/4.5,1.0/4.5,1.0/0.45);AF2 k=AF2(1.0/1.099,0.099/1.099);
+   return AZolSelF3(AZolSignedF3(c-j.xxx),c*j.yyy,pow(c*k.xxx+k.yyy,j.zzz));}
+//------------------------------------------------------------------------------------------------------------------------------
+  AF1 AFromGammaF1(AF1 c,AF1 x){return pow(c,AF1_(x));} 
+  AF2 AFromGammaF2(AF2 c,AF1 x){return pow(c,AF2_(x));} 
+  AF3 AFromGammaF3(AF3 c,AF1 x){return pow(c,AF3_(x));} 
+//------------------------------------------------------------------------------------------------------------------------------
+  AF1 AFromPqF1(AF1 x){AF1 p=pow(x,AF1_(0.0126833));
+   return pow(ASatF1(p-AF1_(0.835938))/(AF1_(18.8516)-AF1_(18.6875)*p),AF1_(6.27739));}
+  AF2 AFromPqF1(AF2 x){AF2 p=pow(x,AF2_(0.0126833));
+   return pow(ASatF2(p-AF2_(0.835938))/(AF2_(18.8516)-AF2_(18.6875)*p),AF2_(6.27739));}
+  AF3 AFromPqF1(AF3 x){AF3 p=pow(x,AF3_(0.0126833));
+   return pow(ASatF3(p-AF3_(0.835938))/(AF3_(18.8516)-AF3_(18.6875)*p),AF3_(6.27739));}
+//------------------------------------------------------------------------------------------------------------------------------
+  // Unfortunately median won't work here.
+  AF1 AFromSrgbF1(AF1 c){AF3 j=AF3(0.04045/12.92,1.0/12.92,2.4);AF2 k=AF2(1.0/1.055,0.055/1.055);
+   return AZolSelF1(AZolSignedF1(c-j.x  ),c*j.y  ,pow(c*k.x  +k.y  ,j.z  ));}
+  AF2 AFromSrgbF2(AF2 c){AF3 j=AF3(0.04045/12.92,1.0/12.92,2.4);AF2 k=AF2(1.0/1.055,0.055/1.055);
+   return AZolSelF2(AZolSignedF2(c-j.xx ),c*j.yy ,pow(c*k.xx +k.yy ,j.zz ));}
+  AF3 AFromSrgbF3(AF3 c){AF3 j=AF3(0.04045/12.92,1.0/12.92,2.4);AF2 k=AF2(1.0/1.055,0.055/1.055);
+   return AZolSelF3(AZolSignedF3(c-j.xxx),c*j.yyy,pow(c*k.xxx+k.yyy,j.zzz));}
+//------------------------------------------------------------------------------------------------------------------------------
+  AF1 AFromTwoF1(AF1 c){return c*c;}
+  AF2 AFromTwoF2(AF2 c){return c*c;}
+  AF3 AFromTwoF3(AF3 c){return c*c;}
+//------------------------------------------------------------------------------------------------------------------------------
+  AF1 AFromThreeF1(AF1 c){return c*c*c;}
+  AF2 AFromThreeF2(AF2 c){return c*c*c;}
+  AF3 AFromThreeF3(AF3 c){return c*c*c;}
+ #endif
+//==============================================================================================================================
+ #ifdef A_HALF
+  AH1 ATo709H1(AH1 c){AH3 j=AH3(0.018*4.5,4.5,0.45);AH2 k=AH2(1.099,-0.099);
+   return clamp(j.x  ,c*j.y  ,pow(c,j.z  )*k.x  +k.y  );}
+  AH2 ATo709H2(AH2 c){AH3 j=AH3(0.018*4.5,4.5,0.45);AH2 k=AH2(1.099,-0.099);
+   return clamp(j.xx ,c*j.yy ,pow(c,j.zz )*k.xx +k.yy );}
+  AH3 ATo709H3(AH3 c){AH3 j=AH3(0.018*4.5,4.5,0.45);AH2 k=AH2(1.099,-0.099);
+   return clamp(j.xxx,c*j.yyy,pow(c,j.zzz)*k.xxx+k.yyy);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AH1 AToGammaH1(AH1 c,AH1 rcpX){return pow(c,AH1_(rcpX));}
+  AH2 AToGammaH2(AH2 c,AH1 rcpX){return pow(c,AH2_(rcpX));}
+  AH3 AToGammaH3(AH3 c,AH1 rcpX){return pow(c,AH3_(rcpX));}
+//------------------------------------------------------------------------------------------------------------------------------
+  AH1 AToSrgbH1(AH1 c){AH3 j=AH3(0.0031308*12.92,12.92,1.0/2.4);AH2 k=AH2(1.055,-0.055);
+   return clamp(j.x  ,c*j.y  ,pow(c,j.z  )*k.x  +k.y  );}
+  AH2 AToSrgbH2(AH2 c){AH3 j=AH3(0.0031308*12.92,12.92,1.0/2.4);AH2 k=AH2(1.055,-0.055);
+   return clamp(j.xx ,c*j.yy ,pow(c,j.zz )*k.xx +k.yy );}
+  AH3 AToSrgbH3(AH3 c){AH3 j=AH3(0.0031308*12.92,12.92,1.0/2.4);AH2 k=AH2(1.055,-0.055);
+   return clamp(j.xxx,c*j.yyy,pow(c,j.zzz)*k.xxx+k.yyy);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AH1 AToTwoH1(AH1 c){return sqrt(c);}
+  AH2 AToTwoH2(AH2 c){return sqrt(c);}
+  AH3 AToTwoH3(AH3 c){return sqrt(c);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AH1 AToThreeF1(AH1 c){return pow(c,AH1_(1.0/3.0));}
+  AH2 AToThreeF2(AH2 c){return pow(c,AH2_(1.0/3.0));}
+  AH3 AToThreeF3(AH3 c){return pow(c,AH3_(1.0/3.0));}
+ #endif
+//==============================================================================================================================
+ #ifdef A_HALF
+  AH1 AFrom709H1(AH1 c){AH3 j=AH3(0.081/4.5,1.0/4.5,1.0/0.45);AH2 k=AH2(1.0/1.099,0.099/1.099);
+   return AZolSelH1(AZolSignedH1(c-j.x  ),c*j.y  ,pow(c*k.x  +k.y  ,j.z  ));}
+  AH2 AFrom709H2(AH2 c){AH3 j=AH3(0.081/4.5,1.0/4.5,1.0/0.45);AH2 k=AH2(1.0/1.099,0.099/1.099);
+   return AZolSelH2(AZolSignedH2(c-j.xx ),c*j.yy ,pow(c*k.xx +k.yy ,j.zz ));}
+  AH3 AFrom709H3(AH3 c){AH3 j=AH3(0.081/4.5,1.0/4.5,1.0/0.45);AH2 k=AH2(1.0/1.099,0.099/1.099);
+   return AZolSelH3(AZolSignedH3(c-j.xxx),c*j.yyy,pow(c*k.xxx+k.yyy,j.zzz));}
+//------------------------------------------------------------------------------------------------------------------------------
+  AH1 AFromGammaH1(AH1 c,AH1 x){return pow(c,AH1_(x));}
+  AH2 AFromGammaH2(AH2 c,AH1 x){return pow(c,AH2_(x));}
+  AH3 AFromGammaH3(AH3 c,AH1 x){return pow(c,AH3_(x));}
+//------------------------------------------------------------------------------------------------------------------------------
+  AH1 AHromSrgbF1(AH1 c){AH3 j=AH3(0.04045/12.92,1.0/12.92,2.4);AH2 k=AH2(1.0/1.055,0.055/1.055);
+   return AZolSelH1(AZolSignedH1(c-j.x  ),c*j.y  ,pow(c*k.x  +k.y  ,j.z  ));}
+  AH2 AHromSrgbF2(AH2 c){AH3 j=AH3(0.04045/12.92,1.0/12.92,2.4);AH2 k=AH2(1.0/1.055,0.055/1.055);
+   return AZolSelH2(AZolSignedH2(c-j.xx ),c*j.yy ,pow(c*k.xx +k.yy ,j.zz ));}
+  AH3 AHromSrgbF3(AH3 c){AH3 j=AH3(0.04045/12.92,1.0/12.92,2.4);AH2 k=AH2(1.0/1.055,0.055/1.055);
+   return AZolSelH3(AZolSignedH3(c-j.xxx),c*j.yyy,pow(c*k.xxx+k.yyy,j.zzz));}
+//------------------------------------------------------------------------------------------------------------------------------
+  AH1 AFromTwoH1(AH1 c){return c*c;}
+  AH2 AFromTwoH2(AH2 c){return c*c;}
+  AH3 AFromTwoH3(AH3 c){return c*c;}
+//------------------------------------------------------------------------------------------------------------------------------
+  AH1 AFromThreeH1(AH1 c){return c*c*c;}
+  AH2 AFromThreeH2(AH2 c){return c*c*c;}
+  AH3 AFromThreeH3(AH3 c){return c*c*c;}
+ #endif
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                          CS REMAP
+//==============================================================================================================================
+ // Simple remap 64x1 to 8x8 with rotated 2x2 pixel quads in quad linear.
+ //  543210
+ //  ======
+ //  ..xxx.
+ //  yy...y
+ AU2 ARmp8x8(AU1 a){return AU2(ABfe(a,1u,3u),ABfiM(ABfe(a,3u,3u),a,1u));}
+//==============================================================================================================================
+ // More complex remap 64x1 to 8x8 which is necessary for 2D wave reductions.
+ //  543210
+ //  ======
+ //  .xx..x
+ //  y..yy.
+ // Details,
+ //  LANE TO 8x8 MAPPING
+ //  ===================
+ //  00 01 08 09 10 11 18 19 
+ //  02 03 0a 0b 12 13 1a 1b
+ //  04 05 0c 0d 14 15 1c 1d
+ //  06 07 0e 0f 16 17 1e 1f 
+ //  20 21 28 29 30 31 38 39 
+ //  22 23 2a 2b 32 33 3a 3b
+ //  24 25 2c 2d 34 35 3c 3d
+ //  26 27 2e 2f 36 37 3e 3f 
+ AU2 ARmpRed8x8(AU1 a){return AU2(ABfiM(ABfe(a,2u,3u),a,1u),ABfiM(ABfe(a,3u,3u),ABfe(a,1u,2u),2u));}
+//==============================================================================================================================
+ #ifdef A_HALF
+  AW2 ARmp8x8H(AU1 a){return AW2(ABfe(a,1u,3u),ABfiM(ABfe(a,3u,3u),a,1u));}
+  AW2 ARmpRed8x8H(AU1 a){return AW2(ABfiM(ABfe(a,2u,3u),a,1u),ABfiM(ABfe(a,3u,3u),ABfe(a,1u,2u),2u));}
+ #endif
+#endif
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//
+//                                                          REFERENCE
+//
+//------------------------------------------------------------------------------------------------------------------------------
+// IEEE FLOAT RULES
+// ================
+//  - saturate(NaN)=0, saturate(-INF)=0, saturate(+INF)=1
+//  - {+/-}0 * {+/-}INF = NaN
+//  - -INF + (+INF) = NaN
+//  - {+/-}0 / {+/-}0 = NaN
+//  - {+/-}INF / {+/-}INF = NaN
+//  - a<(-0) := sqrt(a) = NaN (a=-0.0 won't NaN)
+//  - 0 == -0
+//  - 4/0 = +INF
+//  - 4/-0 = -INF
+//  - 4+INF = +INF
+//  - 4-INF = -INF
+//  - 4*(+INF) = +INF
+//  - 4*(-INF) = -INF
+//  - -4*(+INF) = -INF
+//  - sqrt(+INF) = +INF
+//------------------------------------------------------------------------------------------------------------------------------
+// FP16 ENCODING
+// =============
+// fedcba9876543210
+// ----------------
+// ......mmmmmmmmmm  10-bit mantissa (encodes 11-bit 0.5 to 1.0 except for denormals)
+// .eeeee..........  5-bit exponent
+// .00000..........  denormals
+// .00001..........  -14 exponent
+// .11110..........   15 exponent
+// .111110000000000  infinity
+// .11111nnnnnnnnnn  NaN with n!=0
+// s...............  sign
+//------------------------------------------------------------------------------------------------------------------------------
+// FP16/INT16 ALIASING DENORMAL
+// ============================
+// 11-bit unsigned integers alias with half float denormal/normal values,
+//     1 = 2^(-24) = 1/16777216 ....................... first denormal value
+//     2 = 2^(-23)
+//   ...
+//  1023 = 2^(-14)*(1-2^(-10)) = 2^(-14)*(1-1/1024) ... last denormal value
+//  1024 = 2^(-14) = 1/16384 .......................... first normal value that still maps to integers
+//  2047 .............................................. last normal value that still maps to integers 
+// Scaling limits,
+//  2^15 = 32768 ...................................... largest power of 2 scaling
+// Largest pow2 conversion mapping is at *32768,
+//     1 : 2^(-9) = 1/512
+//     2 : 1/256
+//     4 : 1/128
+//     8 : 1/64
+//    16 : 1/32
+//    32 : 1/16
+//    64 : 1/8
+//   128 : 1/4
+//   256 : 1/2
+//   512 : 1
+//  1024 : 2
+//  2047 : a little less than 4
+//==============================================================================================================================
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//
+//
+//                                                     GPU/CPU PORTABILITY
+//
+//
+//------------------------------------------------------------------------------------------------------------------------------
+// This is the GPU implementation.
+// See the CPU implementation for docs.
+//==============================================================================================================================
+#ifdef A_GPU
+ #define A_TRUE true
+ #define A_FALSE false
+ #define A_STATIC
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                     VECTOR ARGUMENT/RETURN/INITIALIZATION PORTABILITY
+//==============================================================================================================================
+ #define retAD2 AD2
+ #define retAD3 AD3
+ #define retAD4 AD4
+ #define retAF2 AF2
+ #define retAF3 AF3
+ #define retAF4 AF4
+ #define retAL2 AL2
+ #define retAL3 AL3
+ #define retAL4 AL4
+ #define retAU2 AU2
+ #define retAU3 AU3
+ #define retAU4 AU4
+//------------------------------------------------------------------------------------------------------------------------------
+ #define inAD2 in AD2
+ #define inAD3 in AD3
+ #define inAD4 in AD4
+ #define inAF2 in AF2
+ #define inAF3 in AF3
+ #define inAF4 in AF4
+ #define inAL2 in AL2
+ #define inAL3 in AL3
+ #define inAL4 in AL4
+ #define inAU2 in AU2
+ #define inAU3 in AU3
+ #define inAU4 in AU4
+//------------------------------------------------------------------------------------------------------------------------------
+ #define inoutAD2 inout AD2
+ #define inoutAD3 inout AD3
+ #define inoutAD4 inout AD4
+ #define inoutAF2 inout AF2
+ #define inoutAF3 inout AF3
+ #define inoutAF4 inout AF4
+ #define inoutAL2 inout AL2
+ #define inoutAL3 inout AL3
+ #define inoutAL4 inout AL4
+ #define inoutAU2 inout AU2
+ #define inoutAU3 inout AU3
+ #define inoutAU4 inout AU4
+//------------------------------------------------------------------------------------------------------------------------------
+ #define outAD2 out AD2
+ #define outAD3 out AD3
+ #define outAD4 out AD4
+ #define outAF2 out AF2
+ #define outAF3 out AF3
+ #define outAF4 out AF4
+ #define outAL2 out AL2
+ #define outAL3 out AL3
+ #define outAL4 out AL4
+ #define outAU2 out AU2
+ #define outAU3 out AU3
+ #define outAU4 out AU4
+//------------------------------------------------------------------------------------------------------------------------------
+ #define varAD2(x) AD2 x
+ #define varAD3(x) AD3 x
+ #define varAD4(x) AD4 x
+ #define varAF2(x) AF2 x
+ #define varAF3(x) AF3 x
+ #define varAF4(x) AF4 x
+ #define varAL2(x) AL2 x
+ #define varAL3(x) AL3 x
+ #define varAL4(x) AL4 x
+ #define varAU2(x) AU2 x
+ #define varAU3(x) AU3 x
+ #define varAU4(x) AU4 x
+//------------------------------------------------------------------------------------------------------------------------------
+ #define initAD2(x,y) AD2(x,y)
+ #define initAD3(x,y,z) AD3(x,y,z)
+ #define initAD4(x,y,z,w) AD4(x,y,z,w)
+ #define initAF2(x,y) AF2(x,y)
+ #define initAF3(x,y,z) AF3(x,y,z)
+ #define initAF4(x,y,z,w) AF4(x,y,z,w)
+ #define initAL2(x,y) AL2(x,y)
+ #define initAL3(x,y,z) AL3(x,y,z)
+ #define initAL4(x,y,z,w) AL4(x,y,z,w)
+ #define initAU2(x,y) AU2(x,y)
+ #define initAU3(x,y,z) AU3(x,y,z)
+ #define initAU4(x,y,z,w) AU4(x,y,z,w)
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                     SCALAR RETURN OPS
+//==============================================================================================================================
+ #define AAbsD1(a) abs(AD1(a))
+ #define AAbsF1(a) abs(AF1(a))
+//------------------------------------------------------------------------------------------------------------------------------
+ #define ACosD1(a) cos(AD1(a))
+ #define ACosF1(a) cos(AF1(a))
+//------------------------------------------------------------------------------------------------------------------------------
+ #define ADotD2(a,b) dot(AD2(a),AD2(b))
+ #define ADotD3(a,b) dot(AD3(a),AD3(b))
+ #define ADotD4(a,b) dot(AD4(a),AD4(b))
+ #define ADotF2(a,b) dot(AF2(a),AF2(b))
+ #define ADotF3(a,b) dot(AF3(a),AF3(b))
+ #define ADotF4(a,b) dot(AF4(a),AF4(b))
+//------------------------------------------------------------------------------------------------------------------------------
+ #define AExp2D1(a) exp2(AD1(a))
+ #define AExp2F1(a) exp2(AF1(a))
+//------------------------------------------------------------------------------------------------------------------------------
+ #define AFloorD1(a) floor(AD1(a))
+ #define AFloorF1(a) floor(AF1(a))
+//------------------------------------------------------------------------------------------------------------------------------
+ #define ALog2D1(a) log2(AD1(a))
+ #define ALog2F1(a) log2(AF1(a))
+//------------------------------------------------------------------------------------------------------------------------------
+ #define AMaxD1(a,b) max(a,b)
+ #define AMaxF1(a,b) max(a,b)
+ #define AMaxL1(a,b) max(a,b)
+ #define AMaxU1(a,b) max(a,b)
+//------------------------------------------------------------------------------------------------------------------------------
+ #define AMinD1(a,b) min(a,b)
+ #define AMinF1(a,b) min(a,b)
+ #define AMinL1(a,b) min(a,b)
+ #define AMinU1(a,b) min(a,b)
+//------------------------------------------------------------------------------------------------------------------------------
+ #define ASinD1(a) sin(AD1(a))
+ #define ASinF1(a) sin(AF1(a))
+//------------------------------------------------------------------------------------------------------------------------------
+ #define ASqrtD1(a) sqrt(AD1(a))
+ #define ASqrtF1(a) sqrt(AF1(a))
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                               SCALAR RETURN OPS - DEPENDENT
+//==============================================================================================================================
+ #define APowD1(a,b) pow(AD1(a),AF1(b))
+ #define APowF1(a,b) pow(AF1(a),AF1(b))
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                         VECTOR OPS
+//------------------------------------------------------------------------------------------------------------------------------
+// These are added as needed for production or prototyping, so not necessarily a complete set.
+// They follow a convention of taking in a destination and also returning the destination value to increase utility.
+//==============================================================================================================================
+ #ifdef A_DUBL
+  AD2 opAAbsD2(outAD2 d,inAD2 a){d=abs(a);return d;}
+  AD3 opAAbsD3(outAD3 d,inAD3 a){d=abs(a);return d;}
+  AD4 opAAbsD4(outAD4 d,inAD4 a){d=abs(a);return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+  AD2 opAAddD2(outAD2 d,inAD2 a,inAD2 b){d=a+b;return d;}
+  AD3 opAAddD3(outAD3 d,inAD3 a,inAD3 b){d=a+b;return d;}
+  AD4 opAAddD4(outAD4 d,inAD4 a,inAD4 b){d=a+b;return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+  AD2 opAAddOneD2(outAD2 d,inAD2 a,AD1 b){d=a+AD2_(b);return d;}
+  AD3 opAAddOneD3(outAD3 d,inAD3 a,AD1 b){d=a+AD3_(b);return d;}
+  AD4 opAAddOneD4(outAD4 d,inAD4 a,AD1 b){d=a+AD4_(b);return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+  AD2 opACpyD2(outAD2 d,inAD2 a){d=a;return d;}
+  AD3 opACpyD3(outAD3 d,inAD3 a){d=a;return d;}
+  AD4 opACpyD4(outAD4 d,inAD4 a){d=a;return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+  AD2 opALerpD2(outAD2 d,inAD2 a,inAD2 b,inAD2 c){d=ALerpD2(a,b,c);return d;}
+  AD3 opALerpD3(outAD3 d,inAD3 a,inAD3 b,inAD3 c){d=ALerpD3(a,b,c);return d;}
+  AD4 opALerpD4(outAD4 d,inAD4 a,inAD4 b,inAD4 c){d=ALerpD4(a,b,c);return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+  AD2 opALerpOneD2(outAD2 d,inAD2 a,inAD2 b,AD1 c){d=ALerpD2(a,b,AD2_(c));return d;}
+  AD3 opALerpOneD3(outAD3 d,inAD3 a,inAD3 b,AD1 c){d=ALerpD3(a,b,AD3_(c));return d;}
+  AD4 opALerpOneD4(outAD4 d,inAD4 a,inAD4 b,AD1 c){d=ALerpD4(a,b,AD4_(c));return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+  AD2 opAMaxD2(outAD2 d,inAD2 a,inAD2 b){d=max(a,b);return d;}
+  AD3 opAMaxD3(outAD3 d,inAD3 a,inAD3 b){d=max(a,b);return d;}
+  AD4 opAMaxD4(outAD4 d,inAD4 a,inAD4 b){d=max(a,b);return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+  AD2 opAMinD2(outAD2 d,inAD2 a,inAD2 b){d=min(a,b);return d;}
+  AD3 opAMinD3(outAD3 d,inAD3 a,inAD3 b){d=min(a,b);return d;}
+  AD4 opAMinD4(outAD4 d,inAD4 a,inAD4 b){d=min(a,b);return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+  AD2 opAMulD2(outAD2 d,inAD2 a,inAD2 b){d=a*b;return d;}
+  AD3 opAMulD3(outAD3 d,inAD3 a,inAD3 b){d=a*b;return d;}
+  AD4 opAMulD4(outAD4 d,inAD4 a,inAD4 b){d=a*b;return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+  AD2 opAMulOneD2(outAD2 d,inAD2 a,AD1 b){d=a*AD2_(b);return d;}
+  AD3 opAMulOneD3(outAD3 d,inAD3 a,AD1 b){d=a*AD3_(b);return d;}
+  AD4 opAMulOneD4(outAD4 d,inAD4 a,AD1 b){d=a*AD4_(b);return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+  AD2 opANegD2(outAD2 d,inAD2 a){d=-a;return d;}
+  AD3 opANegD3(outAD3 d,inAD3 a){d=-a;return d;}
+  AD4 opANegD4(outAD4 d,inAD4 a){d=-a;return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+  AD2 opARcpD2(outAD2 d,inAD2 a){d=ARcpD2(a);return d;}
+  AD3 opARcpD3(outAD3 d,inAD3 a){d=ARcpD3(a);return d;}
+  AD4 opARcpD4(outAD4 d,inAD4 a){d=ARcpD4(a);return d;}
+ #endif
+//==============================================================================================================================
+ AF2 opAAbsF2(outAF2 d,inAF2 a){d=abs(a);return d;}
+ AF3 opAAbsF3(outAF3 d,inAF3 a){d=abs(a);return d;}
+ AF4 opAAbsF4(outAF4 d,inAF4 a){d=abs(a);return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF2 opAAddF2(outAF2 d,inAF2 a,inAF2 b){d=a+b;return d;}
+ AF3 opAAddF3(outAF3 d,inAF3 a,inAF3 b){d=a+b;return d;}
+ AF4 opAAddF4(outAF4 d,inAF4 a,inAF4 b){d=a+b;return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF2 opAAddOneF2(outAF2 d,inAF2 a,AF1 b){d=a+AF2_(b);return d;}
+ AF3 opAAddOneF3(outAF3 d,inAF3 a,AF1 b){d=a+AF3_(b);return d;}
+ AF4 opAAddOneF4(outAF4 d,inAF4 a,AF1 b){d=a+AF4_(b);return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF2 opACpyF2(outAF2 d,inAF2 a){d=a;return d;}
+ AF3 opACpyF3(outAF3 d,inAF3 a){d=a;return d;}
+ AF4 opACpyF4(outAF4 d,inAF4 a){d=a;return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF2 opALerpF2(outAF2 d,inAF2 a,inAF2 b,inAF2 c){d=ALerpF2(a,b,c);return d;}
+ AF3 opALerpF3(outAF3 d,inAF3 a,inAF3 b,inAF3 c){d=ALerpF3(a,b,c);return d;}
+ AF4 opALerpF4(outAF4 d,inAF4 a,inAF4 b,inAF4 c){d=ALerpF4(a,b,c);return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF2 opALerpOneF2(outAF2 d,inAF2 a,inAF2 b,AF1 c){d=ALerpF2(a,b,AF2_(c));return d;}
+ AF3 opALerpOneF3(outAF3 d,inAF3 a,inAF3 b,AF1 c){d=ALerpF3(a,b,AF3_(c));return d;}
+ AF4 opALerpOneF4(outAF4 d,inAF4 a,inAF4 b,AF1 c){d=ALerpF4(a,b,AF4_(c));return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF2 opAMaxF2(outAF2 d,inAF2 a,inAF2 b){d=max(a,b);return d;}
+ AF3 opAMaxF3(outAF3 d,inAF3 a,inAF3 b){d=max(a,b);return d;}
+ AF4 opAMaxF4(outAF4 d,inAF4 a,inAF4 b){d=max(a,b);return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF2 opAMinF2(outAF2 d,inAF2 a,inAF2 b){d=min(a,b);return d;}
+ AF3 opAMinF3(outAF3 d,inAF3 a,inAF3 b){d=min(a,b);return d;}
+ AF4 opAMinF4(outAF4 d,inAF4 a,inAF4 b){d=min(a,b);return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF2 opAMulF2(outAF2 d,inAF2 a,inAF2 b){d=a*b;return d;}
+ AF3 opAMulF3(outAF3 d,inAF3 a,inAF3 b){d=a*b;return d;}
+ AF4 opAMulF4(outAF4 d,inAF4 a,inAF4 b){d=a*b;return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF2 opAMulOneF2(outAF2 d,inAF2 a,AF1 b){d=a*AF2_(b);return d;}
+ AF3 opAMulOneF3(outAF3 d,inAF3 a,AF1 b){d=a*AF3_(b);return d;}
+ AF4 opAMulOneF4(outAF4 d,inAF4 a,AF1 b){d=a*AF4_(b);return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF2 opANegF2(outAF2 d,inAF2 a){d=-a;return d;}
+ AF3 opANegF3(outAF3 d,inAF3 a){d=-a;return d;}
+ AF4 opANegF4(outAF4 d,inAF4 a){d=-a;return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF2 opARcpF2(outAF2 d,inAF2 a){d=ARcpF2(a);return d;}
+ AF3 opARcpF3(outAF3 d,inAF3 a){d=ARcpF3(a);return d;}
+ AF4 opARcpF4(outAF4 d,inAF4 a){d=ARcpF4(a);return d;}
+#endif
diff --git a/src/Ryujinx.Graphics.OpenGL/Effects/Shaders/ffx_fsr1.h b/src/Ryujinx.Graphics.OpenGL/Effects/Shaders/ffx_fsr1.h
new file mode 100644
index 00000000..4e0b3d54
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Effects/Shaders/ffx_fsr1.h
@@ -0,0 +1,1199 @@
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//
+//
+//                    AMD FidelityFX SUPER RESOLUTION [FSR 1] ::: SPATIAL SCALING & EXTRAS - v1.20210629
+//
+//
+//------------------------------------------------------------------------------------------------------------------------------
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//------------------------------------------------------------------------------------------------------------------------------
+// FidelityFX Super Resolution Sample
+//
+// Copyright (c) 2021 Advanced Micro Devices, Inc. All rights reserved.
+// Permission is hereby granted, free of charge, to any person obtaining a copy
+// of this software and associated documentation files(the "Software"), to deal
+// in the Software without restriction, including without limitation the rights
+// to use, copy, modify, merge, publish, distribute, sublicense, and / or sell
+// copies of the Software, and to permit persons to whom the Software is
+// furnished to do so, subject to the following conditions :
+// The above copyright notice and this permission notice shall be included in
+// all copies or substantial portions of the Software.
+// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.IN NO EVENT SHALL THE
+// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+// THE SOFTWARE.
+//------------------------------------------------------------------------------------------------------------------------------
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//------------------------------------------------------------------------------------------------------------------------------
+// ABOUT
+// =====
+// FSR is a collection of algorithms relating to generating a higher resolution image.
+// This specific header focuses on single-image non-temporal image scaling, and related tools.
+// 
+// The core functions are EASU and RCAS:
+//  [EASU] Edge Adaptive Spatial Upsampling ....... 1x to 4x area range spatial scaling, clamped adaptive elliptical filter.
+//  [RCAS] Robust Contrast Adaptive Sharpening .... A non-scaling variation on CAS.
+// RCAS needs to be applied after EASU as a separate pass.
+// 
+// Optional utility functions are:
+//  [LFGA] Linear Film Grain Applicator ........... Tool to apply film grain after scaling.
+//  [SRTM] Simple Reversible Tone-Mapper .......... Linear HDR {0 to FP16_MAX} to {0 to 1} and back.
+//  [TEPD] Temporal Energy Preserving Dither ...... Temporally energy preserving dithered {0 to 1} linear to gamma 2.0 conversion.
+// See each individual sub-section for inline documentation.
+//------------------------------------------------------------------------------------------------------------------------------
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//------------------------------------------------------------------------------------------------------------------------------
+// FUNCTION PERMUTATIONS
+// =====================
+// *F() ..... Single item computation with 32-bit.
+// *H() ..... Single item computation with 16-bit, with packing (aka two 16-bit ops in parallel) when possible.
+// *Hx2() ... Processing two items in parallel with 16-bit, easier packing.
+//            Not all interfaces in this file have a *Hx2() form.
+//==============================================================================================================================
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//
+//                                        FSR - [EASU] EDGE ADAPTIVE SPATIAL UPSAMPLING
+//
+//------------------------------------------------------------------------------------------------------------------------------
+// EASU provides a high quality spatial-only scaling at relatively low cost.
+// Meaning EASU is appropiate for laptops and other low-end GPUs.
+// Quality from 1x to 4x area scaling is good.
+//------------------------------------------------------------------------------------------------------------------------------
+// The scalar uses a modified fast approximation to the standard lanczos(size=2) kernel.
+// EASU runs in a single pass, so it applies a directionally and anisotropically adaptive radial lanczos.
+// This is also kept as simple as possible to have minimum runtime.
+//------------------------------------------------------------------------------------------------------------------------------
+// The lanzcos filter has negative lobes, so by itself it will introduce ringing.
+// To remove all ringing, the algorithm uses the nearest 2x2 input texels as a neighborhood,
+// and limits output to the minimum and maximum of that neighborhood.
+//------------------------------------------------------------------------------------------------------------------------------
+// Input image requirements:
+// 
+// Color needs to be encoded as 3 channel[red, green, blue](e.g.XYZ not supported)
+// Each channel needs to be in the range[0, 1]
+// Any color primaries are supported
+// Display / tonemapping curve needs to be as if presenting to sRGB display or similar(e.g.Gamma 2.0)
+// There should be no banding in the input
+// There should be no high amplitude noise in the input
+// There should be no noise in the input that is not at input pixel granularity
+// For performance purposes, use 32bpp formats
+//------------------------------------------------------------------------------------------------------------------------------
+// Best to apply EASU at the end of the frame after tonemapping 
+// but before film grain or composite of the UI.
+//------------------------------------------------------------------------------------------------------------------------------
+// Example of including this header for D3D HLSL :
+// 
+//  #define A_GPU 1
+//  #define A_HLSL 1
+//  #define A_HALF 1
+//  #include "ffx_a.h"
+//  #define FSR_EASU_H 1
+//  #define FSR_RCAS_H 1
+//  //declare input callbacks
+//  #include "ffx_fsr1.h"
+// 
+// Example of including this header for Vulkan GLSL :
+// 
+//  #define A_GPU 1
+//  #define A_GLSL 1
+//  #define A_HALF 1
+//  #include "ffx_a.h"
+//  #define FSR_EASU_H 1
+//  #define FSR_RCAS_H 1
+//  //declare input callbacks
+//  #include "ffx_fsr1.h"
+// 
+// Example of including this header for Vulkan HLSL :
+// 
+//  #define A_GPU 1
+//  #define A_HLSL 1
+//  #define A_HLSL_6_2 1
+//  #define A_NO_16_BIT_CAST 1
+//  #define A_HALF 1
+//  #include "ffx_a.h"
+//  #define FSR_EASU_H 1
+//  #define FSR_RCAS_H 1
+//  //declare input callbacks
+//  #include "ffx_fsr1.h"
+// 
+//  Example of declaring the required input callbacks for GLSL :
+//  The callbacks need to gather4 for each color channel using the specified texture coordinate 'p'.
+//  EASU uses gather4 to reduce position computation logic and for free Arrays of Structures to Structures of Arrays conversion.
+// 
+//  AH4 FsrEasuRH(AF2 p){return AH4(textureGather(sampler2D(tex,sam),p,0));}
+//  AH4 FsrEasuGH(AF2 p){return AH4(textureGather(sampler2D(tex,sam),p,1));}
+//  AH4 FsrEasuBH(AF2 p){return AH4(textureGather(sampler2D(tex,sam),p,2));}
+//  ...
+//  The FsrEasuCon function needs to be called from the CPU or GPU to set up constants.
+//  The difference in viewport and input image size is there to support Dynamic Resolution Scaling.
+//  To use FsrEasuCon() on the CPU, define A_CPU before including ffx_a and ffx_fsr1.
+//  Including a GPU example here, the 'con0' through 'con3' values would be stored out to a constant buffer.
+//  AU4 con0,con1,con2,con3;
+//  FsrEasuCon(con0,con1,con2,con3,
+//    1920.0,1080.0,  // Viewport size (top left aligned) in the input image which is to be scaled.
+//    3840.0,2160.0,  // The size of the input image.
+//    2560.0,1440.0); // The output resolution.
+//==============================================================================================================================
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                      CONSTANT SETUP
+//==============================================================================================================================
+// Call to setup required constant values (works on CPU or GPU).
+A_STATIC void FsrEasuCon(
+outAU4 con0,
+outAU4 con1,
+outAU4 con2,
+outAU4 con3,
+// This the rendered image resolution being upscaled
+AF1 inputViewportInPixelsX,
+AF1 inputViewportInPixelsY,
+// This is the resolution of the resource containing the input image (useful for dynamic resolution)
+AF1 inputSizeInPixelsX,
+AF1 inputSizeInPixelsY,
+// This is the display resolution which the input image gets upscaled to
+AF1 outputSizeInPixelsX,
+AF1 outputSizeInPixelsY){
+ // Output integer position to a pixel position in viewport.
+ con0[0]=AU1_AF1(inputViewportInPixelsX*ARcpF1(outputSizeInPixelsX));
+ con0[1]=AU1_AF1(inputViewportInPixelsY*ARcpF1(outputSizeInPixelsY));
+ con0[2]=AU1_AF1(AF1_(0.5)*inputViewportInPixelsX*ARcpF1(outputSizeInPixelsX)-AF1_(0.5));
+ con0[3]=AU1_AF1(AF1_(0.5)*inputViewportInPixelsY*ARcpF1(outputSizeInPixelsY)-AF1_(0.5));
+ // Viewport pixel position to normalized image space.
+ // This is used to get upper-left of 'F' tap.
+ con1[0]=AU1_AF1(ARcpF1(inputSizeInPixelsX));
+ con1[1]=AU1_AF1(ARcpF1(inputSizeInPixelsY));
+ // Centers of gather4, first offset from upper-left of 'F'.
+ //      +---+---+
+ //      |   |   |
+ //      +--(0)--+
+ //      | b | c |
+ //  +---F---+---+---+
+ //  | e | f | g | h |
+ //  +--(1)--+--(2)--+
+ //  | i | j | k | l |
+ //  +---+---+---+---+
+ //      | n | o |
+ //      +--(3)--+
+ //      |   |   |
+ //      +---+---+
+ con1[2]=AU1_AF1(AF1_( 1.0)*ARcpF1(inputSizeInPixelsX));
+ con1[3]=AU1_AF1(AF1_(-1.0)*ARcpF1(inputSizeInPixelsY));
+ // These are from (0) instead of 'F'.
+ con2[0]=AU1_AF1(AF1_(-1.0)*ARcpF1(inputSizeInPixelsX));
+ con2[1]=AU1_AF1(AF1_( 2.0)*ARcpF1(inputSizeInPixelsY));
+ con2[2]=AU1_AF1(AF1_( 1.0)*ARcpF1(inputSizeInPixelsX));
+ con2[3]=AU1_AF1(AF1_( 2.0)*ARcpF1(inputSizeInPixelsY));
+ con3[0]=AU1_AF1(AF1_( 0.0)*ARcpF1(inputSizeInPixelsX));
+ con3[1]=AU1_AF1(AF1_( 4.0)*ARcpF1(inputSizeInPixelsY));
+ con3[2]=con3[3]=0;}
+
+//If the an offset into the input image resource
+A_STATIC void FsrEasuConOffset(
+    outAU4 con0,
+    outAU4 con1,
+    outAU4 con2,
+    outAU4 con3,
+    // This the rendered image resolution being upscaled
+    AF1 inputViewportInPixelsX,
+    AF1 inputViewportInPixelsY,
+    // This is the resolution of the resource containing the input image (useful for dynamic resolution)
+    AF1 inputSizeInPixelsX,
+    AF1 inputSizeInPixelsY,
+    // This is the display resolution which the input image gets upscaled to
+    AF1 outputSizeInPixelsX,
+    AF1 outputSizeInPixelsY,
+    // This is the input image offset into the resource containing it (useful for dynamic resolution)
+    AF1 inputOffsetInPixelsX,
+    AF1 inputOffsetInPixelsY) {
+    FsrEasuCon(con0, con1, con2, con3, inputViewportInPixelsX, inputViewportInPixelsY, inputSizeInPixelsX, inputSizeInPixelsY, outputSizeInPixelsX, outputSizeInPixelsY);
+    con0[2] = AU1_AF1(AF1_(0.5) * inputViewportInPixelsX * ARcpF1(outputSizeInPixelsX) - AF1_(0.5) + inputOffsetInPixelsX);
+    con0[3] = AU1_AF1(AF1_(0.5) * inputViewportInPixelsY * ARcpF1(outputSizeInPixelsY) - AF1_(0.5) + inputOffsetInPixelsY);
+}
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                   NON-PACKED 32-BIT VERSION
+//==============================================================================================================================
+#if defined(A_GPU)&&defined(FSR_EASU_F)
+ // Input callback prototypes, need to be implemented by calling shader
+ AF4 FsrEasuRF(AF2 p);
+ AF4 FsrEasuGF(AF2 p);
+ AF4 FsrEasuBF(AF2 p);
+//------------------------------------------------------------------------------------------------------------------------------
+ // Filtering for a given tap for the scalar.
+ void FsrEasuTapF(
+ inout AF3 aC, // Accumulated color, with negative lobe.
+ inout AF1 aW, // Accumulated weight.
+ AF2 off, // Pixel offset from resolve position to tap.
+ AF2 dir, // Gradient direction.
+ AF2 len, // Length.
+ AF1 lob, // Negative lobe strength.
+ AF1 clp, // Clipping point.
+ AF3 c){ // Tap color.
+  // Rotate offset by direction.
+  AF2 v;
+  v.x=(off.x*( dir.x))+(off.y*dir.y);
+  v.y=(off.x*(-dir.y))+(off.y*dir.x);
+  // Anisotropy.
+  v*=len;
+  // Compute distance^2.
+  AF1 d2=v.x*v.x+v.y*v.y;
+  // Limit to the window as at corner, 2 taps can easily be outside.
+  d2=min(d2,clp);
+  // Approximation of lancos2 without sin() or rcp(), or sqrt() to get x.
+  //  (25/16 * (2/5 * x^2 - 1)^2 - (25/16 - 1)) * (1/4 * x^2 - 1)^2
+  //  |_______________________________________|   |_______________|
+  //                   base                             window
+  // The general form of the 'base' is,
+  //  (a*(b*x^2-1)^2-(a-1))
+  // Where 'a=1/(2*b-b^2)' and 'b' moves around the negative lobe.
+  AF1 wB=AF1_(2.0/5.0)*d2+AF1_(-1.0);
+  AF1 wA=lob*d2+AF1_(-1.0);
+  wB*=wB;
+  wA*=wA;
+  wB=AF1_(25.0/16.0)*wB+AF1_(-(25.0/16.0-1.0));
+  AF1 w=wB*wA;
+  // Do weighted average.
+  aC+=c*w;aW+=w;}
+//------------------------------------------------------------------------------------------------------------------------------
+ // Accumulate direction and length.
+ void FsrEasuSetF(
+ inout AF2 dir,
+ inout AF1 len,
+ AF2 pp,
+ AP1 biS,AP1 biT,AP1 biU,AP1 biV,
+ AF1 lA,AF1 lB,AF1 lC,AF1 lD,AF1 lE){
+  // Compute bilinear weight, branches factor out as predicates are compiler time immediates.
+  //  s t
+  //  u v
+  AF1 w = AF1_(0.0);
+  if(biS)w=(AF1_(1.0)-pp.x)*(AF1_(1.0)-pp.y);
+  if(biT)w=           pp.x *(AF1_(1.0)-pp.y);
+  if(biU)w=(AF1_(1.0)-pp.x)*           pp.y ;
+  if(biV)w=           pp.x *           pp.y ;
+  // Direction is the '+' diff.
+  //    a
+  //  b c d
+  //    e
+  // Then takes magnitude from abs average of both sides of 'c'.
+  // Length converts gradient reversal to 0, smoothly to non-reversal at 1, shaped, then adding horz and vert terms.
+  AF1 dc=lD-lC;
+  AF1 cb=lC-lB;
+  AF1 lenX=max(abs(dc),abs(cb));
+  lenX=APrxLoRcpF1(lenX);
+  AF1 dirX=lD-lB;
+  dir.x+=dirX*w;
+  lenX=ASatF1(abs(dirX)*lenX);
+  lenX*=lenX;
+  len+=lenX*w;
+  // Repeat for the y axis.
+  AF1 ec=lE-lC;
+  AF1 ca=lC-lA;
+  AF1 lenY=max(abs(ec),abs(ca));
+  lenY=APrxLoRcpF1(lenY);
+  AF1 dirY=lE-lA;
+  dir.y+=dirY*w;
+  lenY=ASatF1(abs(dirY)*lenY);
+  lenY*=lenY;
+  len+=lenY*w;}
+//------------------------------------------------------------------------------------------------------------------------------
+ void FsrEasuF(
+ out AF3 pix,
+ AU2 ip, // Integer pixel position in output.
+ AU4 con0, // Constants generated by FsrEasuCon().
+ AU4 con1,
+ AU4 con2,
+ AU4 con3){
+//------------------------------------------------------------------------------------------------------------------------------
+  // Get position of 'f'.
+  AF2 pp=AF2(ip)*AF2_AU2(con0.xy)+AF2_AU2(con0.zw);
+  AF2 fp=floor(pp);
+  pp-=fp;
+//------------------------------------------------------------------------------------------------------------------------------
+  // 12-tap kernel.
+  //    b c
+  //  e f g h
+  //  i j k l
+  //    n o
+  // Gather 4 ordering.
+  //  a b
+  //  r g
+  // For packed FP16, need either {rg} or {ab} so using the following setup for gather in all versions,
+  //    a b    <- unused (z)
+  //    r g
+  //  a b a b
+  //  r g r g
+  //    a b
+  //    r g    <- unused (z)
+  // Allowing dead-code removal to remove the 'z's.
+  AF2 p0=fp*AF2_AU2(con1.xy)+AF2_AU2(con1.zw);
+  // These are from p0 to avoid pulling two constants on pre-Navi hardware.
+  AF2 p1=p0+AF2_AU2(con2.xy);
+  AF2 p2=p0+AF2_AU2(con2.zw);
+  AF2 p3=p0+AF2_AU2(con3.xy);
+  AF4 bczzR=FsrEasuRF(p0);
+  AF4 bczzG=FsrEasuGF(p0);
+  AF4 bczzB=FsrEasuBF(p0);
+  AF4 ijfeR=FsrEasuRF(p1);
+  AF4 ijfeG=FsrEasuGF(p1);
+  AF4 ijfeB=FsrEasuBF(p1);
+  AF4 klhgR=FsrEasuRF(p2);
+  AF4 klhgG=FsrEasuGF(p2);
+  AF4 klhgB=FsrEasuBF(p2);
+  AF4 zzonR=FsrEasuRF(p3);
+  AF4 zzonG=FsrEasuGF(p3);
+  AF4 zzonB=FsrEasuBF(p3);
+//------------------------------------------------------------------------------------------------------------------------------
+  // Simplest multi-channel approximate luma possible (luma times 2, in 2 FMA/MAD).
+  AF4 bczzL=bczzB*AF4_(0.5)+(bczzR*AF4_(0.5)+bczzG);
+  AF4 ijfeL=ijfeB*AF4_(0.5)+(ijfeR*AF4_(0.5)+ijfeG);
+  AF4 klhgL=klhgB*AF4_(0.5)+(klhgR*AF4_(0.5)+klhgG);
+  AF4 zzonL=zzonB*AF4_(0.5)+(zzonR*AF4_(0.5)+zzonG);
+  // Rename.
+  AF1 bL=bczzL.x;
+  AF1 cL=bczzL.y;
+  AF1 iL=ijfeL.x;
+  AF1 jL=ijfeL.y;
+  AF1 fL=ijfeL.z;
+  AF1 eL=ijfeL.w;
+  AF1 kL=klhgL.x;
+  AF1 lL=klhgL.y;
+  AF1 hL=klhgL.z;
+  AF1 gL=klhgL.w;
+  AF1 oL=zzonL.z;
+  AF1 nL=zzonL.w;
+  // Accumulate for bilinear interpolation.
+  AF2 dir=AF2_(0.0);
+  AF1 len=AF1_(0.0);
+  FsrEasuSetF(dir,len,pp,true, false,false,false,bL,eL,fL,gL,jL);
+  FsrEasuSetF(dir,len,pp,false,true ,false,false,cL,fL,gL,hL,kL);
+  FsrEasuSetF(dir,len,pp,false,false,true ,false,fL,iL,jL,kL,nL);
+  FsrEasuSetF(dir,len,pp,false,false,false,true ,gL,jL,kL,lL,oL);
+//------------------------------------------------------------------------------------------------------------------------------
+  // Normalize with approximation, and cleanup close to zero.
+  AF2 dir2=dir*dir;
+  AF1 dirR=dir2.x+dir2.y;
+  AP1 zro=dirR<AF1_(1.0/32768.0);
+  dirR=APrxLoRsqF1(dirR);
+  dirR=zro?AF1_(1.0):dirR;
+  dir.x=zro?AF1_(1.0):dir.x;
+  dir*=AF2_(dirR);
+  // Transform from {0 to 2} to {0 to 1} range, and shape with square.
+  len=len*AF1_(0.5);
+  len*=len;
+  // Stretch kernel {1.0 vert|horz, to sqrt(2.0) on diagonal}.
+  AF1 stretch=(dir.x*dir.x+dir.y*dir.y)*APrxLoRcpF1(max(abs(dir.x),abs(dir.y)));
+  // Anisotropic length after rotation,
+  //  x := 1.0 lerp to 'stretch' on edges
+  //  y := 1.0 lerp to 2x on edges
+  AF2 len2=AF2(AF1_(1.0)+(stretch-AF1_(1.0))*len,AF1_(1.0)+AF1_(-0.5)*len);
+  // Based on the amount of 'edge',
+  // the window shifts from +/-{sqrt(2.0) to slightly beyond 2.0}.
+  AF1 lob=AF1_(0.5)+AF1_((1.0/4.0-0.04)-0.5)*len;
+  // Set distance^2 clipping point to the end of the adjustable window.
+  AF1 clp=APrxLoRcpF1(lob);
+//------------------------------------------------------------------------------------------------------------------------------
+  // Accumulation mixed with min/max of 4 nearest.
+  //    b c
+  //  e f g h
+  //  i j k l
+  //    n o
+  AF3 min4=min(AMin3F3(AF3(ijfeR.z,ijfeG.z,ijfeB.z),AF3(klhgR.w,klhgG.w,klhgB.w),AF3(ijfeR.y,ijfeG.y,ijfeB.y)),
+               AF3(klhgR.x,klhgG.x,klhgB.x));
+  AF3 max4=max(AMax3F3(AF3(ijfeR.z,ijfeG.z,ijfeB.z),AF3(klhgR.w,klhgG.w,klhgB.w),AF3(ijfeR.y,ijfeG.y,ijfeB.y)),
+               AF3(klhgR.x,klhgG.x,klhgB.x));
+  // Accumulation.
+  AF3 aC=AF3_(0.0);
+  AF1 aW=AF1_(0.0);
+  FsrEasuTapF(aC,aW,AF2( 0.0,-1.0)-pp,dir,len2,lob,clp,AF3(bczzR.x,bczzG.x,bczzB.x)); // b
+  FsrEasuTapF(aC,aW,AF2( 1.0,-1.0)-pp,dir,len2,lob,clp,AF3(bczzR.y,bczzG.y,bczzB.y)); // c
+  FsrEasuTapF(aC,aW,AF2(-1.0, 1.0)-pp,dir,len2,lob,clp,AF3(ijfeR.x,ijfeG.x,ijfeB.x)); // i
+  FsrEasuTapF(aC,aW,AF2( 0.0, 1.0)-pp,dir,len2,lob,clp,AF3(ijfeR.y,ijfeG.y,ijfeB.y)); // j
+  FsrEasuTapF(aC,aW,AF2( 0.0, 0.0)-pp,dir,len2,lob,clp,AF3(ijfeR.z,ijfeG.z,ijfeB.z)); // f
+  FsrEasuTapF(aC,aW,AF2(-1.0, 0.0)-pp,dir,len2,lob,clp,AF3(ijfeR.w,ijfeG.w,ijfeB.w)); // e
+  FsrEasuTapF(aC,aW,AF2( 1.0, 1.0)-pp,dir,len2,lob,clp,AF3(klhgR.x,klhgG.x,klhgB.x)); // k
+  FsrEasuTapF(aC,aW,AF2( 2.0, 1.0)-pp,dir,len2,lob,clp,AF3(klhgR.y,klhgG.y,klhgB.y)); // l
+  FsrEasuTapF(aC,aW,AF2( 2.0, 0.0)-pp,dir,len2,lob,clp,AF3(klhgR.z,klhgG.z,klhgB.z)); // h
+  FsrEasuTapF(aC,aW,AF2( 1.0, 0.0)-pp,dir,len2,lob,clp,AF3(klhgR.w,klhgG.w,klhgB.w)); // g
+  FsrEasuTapF(aC,aW,AF2( 1.0, 2.0)-pp,dir,len2,lob,clp,AF3(zzonR.z,zzonG.z,zzonB.z)); // o
+  FsrEasuTapF(aC,aW,AF2( 0.0, 2.0)-pp,dir,len2,lob,clp,AF3(zzonR.w,zzonG.w,zzonB.w)); // n
+//------------------------------------------------------------------------------------------------------------------------------
+  // Normalize and dering.
+  pix=min(max4,max(min4,aC*AF3_(ARcpF1(aW))));}
+#endif
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                    PACKED 16-BIT VERSION
+//==============================================================================================================================
+#if defined(A_GPU)&&defined(A_HALF)&&defined(FSR_EASU_H)
+// Input callback prototypes, need to be implemented by calling shader
+ AH4 FsrEasuRH(AF2 p);
+ AH4 FsrEasuGH(AF2 p);
+ AH4 FsrEasuBH(AF2 p);
+//------------------------------------------------------------------------------------------------------------------------------
+ // This runs 2 taps in parallel.
+ void FsrEasuTapH(
+ inout AH2 aCR,inout AH2 aCG,inout AH2 aCB,
+ inout AH2 aW,
+ AH2 offX,AH2 offY,
+ AH2 dir,
+ AH2 len,
+ AH1 lob,
+ AH1 clp,
+ AH2 cR,AH2 cG,AH2 cB){
+  AH2 vX,vY;
+  vX=offX*  dir.xx +offY*dir.yy;
+  vY=offX*(-dir.yy)+offY*dir.xx;
+  vX*=len.x;vY*=len.y;
+  AH2 d2=vX*vX+vY*vY;
+  d2=min(d2,AH2_(clp));
+  AH2 wB=AH2_(2.0/5.0)*d2+AH2_(-1.0);
+  AH2 wA=AH2_(lob)*d2+AH2_(-1.0);
+  wB*=wB;
+  wA*=wA;
+  wB=AH2_(25.0/16.0)*wB+AH2_(-(25.0/16.0-1.0));
+  AH2 w=wB*wA;
+  aCR+=cR*w;aCG+=cG*w;aCB+=cB*w;aW+=w;}
+//------------------------------------------------------------------------------------------------------------------------------
+ // This runs 2 taps in parallel.
+ void FsrEasuSetH(
+ inout AH2 dirPX,inout AH2 dirPY,
+ inout AH2 lenP,
+ AH2 pp,
+ AP1 biST,AP1 biUV,
+ AH2 lA,AH2 lB,AH2 lC,AH2 lD,AH2 lE){
+  AH2 w = AH2_(0.0);
+  if(biST)w=(AH2(1.0,0.0)+AH2(-pp.x,pp.x))*AH2_(AH1_(1.0)-pp.y);
+  if(biUV)w=(AH2(1.0,0.0)+AH2(-pp.x,pp.x))*AH2_(          pp.y);
+  // ABS is not free in the packed FP16 path.
+  AH2 dc=lD-lC;
+  AH2 cb=lC-lB;
+  AH2 lenX=max(abs(dc),abs(cb));
+  lenX=ARcpH2(lenX);
+  AH2 dirX=lD-lB;
+  dirPX+=dirX*w;
+  lenX=ASatH2(abs(dirX)*lenX);
+  lenX*=lenX;
+  lenP+=lenX*w;
+  AH2 ec=lE-lC;
+  AH2 ca=lC-lA;
+  AH2 lenY=max(abs(ec),abs(ca));
+  lenY=ARcpH2(lenY);
+  AH2 dirY=lE-lA;
+  dirPY+=dirY*w;
+  lenY=ASatH2(abs(dirY)*lenY);
+  lenY*=lenY;
+  lenP+=lenY*w;}
+//------------------------------------------------------------------------------------------------------------------------------
+ void FsrEasuH(
+ out AH3 pix,
+ AU2 ip,
+ AU4 con0,
+ AU4 con1,
+ AU4 con2,
+ AU4 con3){
+//------------------------------------------------------------------------------------------------------------------------------
+  AF2 pp=AF2(ip)*AF2_AU2(con0.xy)+AF2_AU2(con0.zw);
+  AF2 fp=floor(pp);
+  pp-=fp;
+  AH2 ppp=AH2(pp);
+//------------------------------------------------------------------------------------------------------------------------------
+  AF2 p0=fp*AF2_AU2(con1.xy)+AF2_AU2(con1.zw);
+  AF2 p1=p0+AF2_AU2(con2.xy);
+  AF2 p2=p0+AF2_AU2(con2.zw);
+  AF2 p3=p0+AF2_AU2(con3.xy);
+  AH4 bczzR=FsrEasuRH(p0);
+  AH4 bczzG=FsrEasuGH(p0);
+  AH4 bczzB=FsrEasuBH(p0);
+  AH4 ijfeR=FsrEasuRH(p1);
+  AH4 ijfeG=FsrEasuGH(p1);
+  AH4 ijfeB=FsrEasuBH(p1);
+  AH4 klhgR=FsrEasuRH(p2);
+  AH4 klhgG=FsrEasuGH(p2);
+  AH4 klhgB=FsrEasuBH(p2);
+  AH4 zzonR=FsrEasuRH(p3);
+  AH4 zzonG=FsrEasuGH(p3);
+  AH4 zzonB=FsrEasuBH(p3);
+//------------------------------------------------------------------------------------------------------------------------------
+  AH4 bczzL=bczzB*AH4_(0.5)+(bczzR*AH4_(0.5)+bczzG);
+  AH4 ijfeL=ijfeB*AH4_(0.5)+(ijfeR*AH4_(0.5)+ijfeG);
+  AH4 klhgL=klhgB*AH4_(0.5)+(klhgR*AH4_(0.5)+klhgG);
+  AH4 zzonL=zzonB*AH4_(0.5)+(zzonR*AH4_(0.5)+zzonG);
+  AH1 bL=bczzL.x;
+  AH1 cL=bczzL.y;
+  AH1 iL=ijfeL.x;
+  AH1 jL=ijfeL.y;
+  AH1 fL=ijfeL.z;
+  AH1 eL=ijfeL.w;
+  AH1 kL=klhgL.x;
+  AH1 lL=klhgL.y;
+  AH1 hL=klhgL.z;
+  AH1 gL=klhgL.w;
+  AH1 oL=zzonL.z;
+  AH1 nL=zzonL.w;
+  // This part is different, accumulating 2 taps in parallel.
+  AH2 dirPX=AH2_(0.0);
+  AH2 dirPY=AH2_(0.0);
+  AH2 lenP=AH2_(0.0);
+  FsrEasuSetH(dirPX,dirPY,lenP,ppp,true, false,AH2(bL,cL),AH2(eL,fL),AH2(fL,gL),AH2(gL,hL),AH2(jL,kL));
+  FsrEasuSetH(dirPX,dirPY,lenP,ppp,false,true ,AH2(fL,gL),AH2(iL,jL),AH2(jL,kL),AH2(kL,lL),AH2(nL,oL));
+  AH2 dir=AH2(dirPX.r+dirPX.g,dirPY.r+dirPY.g);
+  AH1 len=lenP.r+lenP.g;
+//------------------------------------------------------------------------------------------------------------------------------
+  AH2 dir2=dir*dir;
+  AH1 dirR=dir2.x+dir2.y;
+  AP1 zro=dirR<AH1_(1.0/32768.0);
+  dirR=APrxLoRsqH1(dirR);
+  dirR=zro?AH1_(1.0):dirR;
+  dir.x=zro?AH1_(1.0):dir.x;
+  dir*=AH2_(dirR);
+  len=len*AH1_(0.5);
+  len*=len;
+  AH1 stretch=(dir.x*dir.x+dir.y*dir.y)*APrxLoRcpH1(max(abs(dir.x),abs(dir.y)));
+  AH2 len2=AH2(AH1_(1.0)+(stretch-AH1_(1.0))*len,AH1_(1.0)+AH1_(-0.5)*len);
+  AH1 lob=AH1_(0.5)+AH1_((1.0/4.0-0.04)-0.5)*len;
+  AH1 clp=APrxLoRcpH1(lob);
+//------------------------------------------------------------------------------------------------------------------------------
+  // FP16 is different, using packed trick to do min and max in same operation.
+  AH2 bothR=max(max(AH2(-ijfeR.z,ijfeR.z),AH2(-klhgR.w,klhgR.w)),max(AH2(-ijfeR.y,ijfeR.y),AH2(-klhgR.x,klhgR.x)));
+  AH2 bothG=max(max(AH2(-ijfeG.z,ijfeG.z),AH2(-klhgG.w,klhgG.w)),max(AH2(-ijfeG.y,ijfeG.y),AH2(-klhgG.x,klhgG.x)));
+  AH2 bothB=max(max(AH2(-ijfeB.z,ijfeB.z),AH2(-klhgB.w,klhgB.w)),max(AH2(-ijfeB.y,ijfeB.y),AH2(-klhgB.x,klhgB.x)));
+  // This part is different for FP16, working pairs of taps at a time.
+  AH2 pR=AH2_(0.0);
+  AH2 pG=AH2_(0.0);
+  AH2 pB=AH2_(0.0);
+  AH2 pW=AH2_(0.0);
+  FsrEasuTapH(pR,pG,pB,pW,AH2( 0.0, 1.0)-ppp.xx,AH2(-1.0,-1.0)-ppp.yy,dir,len2,lob,clp,bczzR.xy,bczzG.xy,bczzB.xy);
+  FsrEasuTapH(pR,pG,pB,pW,AH2(-1.0, 0.0)-ppp.xx,AH2( 1.0, 1.0)-ppp.yy,dir,len2,lob,clp,ijfeR.xy,ijfeG.xy,ijfeB.xy);
+  FsrEasuTapH(pR,pG,pB,pW,AH2( 0.0,-1.0)-ppp.xx,AH2( 0.0, 0.0)-ppp.yy,dir,len2,lob,clp,ijfeR.zw,ijfeG.zw,ijfeB.zw);
+  FsrEasuTapH(pR,pG,pB,pW,AH2( 1.0, 2.0)-ppp.xx,AH2( 1.0, 1.0)-ppp.yy,dir,len2,lob,clp,klhgR.xy,klhgG.xy,klhgB.xy);
+  FsrEasuTapH(pR,pG,pB,pW,AH2( 2.0, 1.0)-ppp.xx,AH2( 0.0, 0.0)-ppp.yy,dir,len2,lob,clp,klhgR.zw,klhgG.zw,klhgB.zw);
+  FsrEasuTapH(pR,pG,pB,pW,AH2( 1.0, 0.0)-ppp.xx,AH2( 2.0, 2.0)-ppp.yy,dir,len2,lob,clp,zzonR.zw,zzonG.zw,zzonB.zw);
+  AH3 aC=AH3(pR.x+pR.y,pG.x+pG.y,pB.x+pB.y);
+  AH1 aW=pW.x+pW.y;
+//------------------------------------------------------------------------------------------------------------------------------
+  // Slightly different for FP16 version due to combined min and max.
+  pix=min(AH3(bothR.y,bothG.y,bothB.y),max(-AH3(bothR.x,bothG.x,bothB.x),aC*AH3_(ARcpH1(aW))));}
+#endif
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//
+//                                      FSR - [RCAS] ROBUST CONTRAST ADAPTIVE SHARPENING
+//
+//------------------------------------------------------------------------------------------------------------------------------
+// CAS uses a simplified mechanism to convert local contrast into a variable amount of sharpness.
+// RCAS uses a more exact mechanism, solving for the maximum local sharpness possible before clipping.
+// RCAS also has a built in process to limit sharpening of what it detects as possible noise.
+// RCAS sharper does not support scaling, as it should be applied after EASU scaling.
+// Pass EASU output straight into RCAS, no color conversions necessary.
+//------------------------------------------------------------------------------------------------------------------------------
+// RCAS is based on the following logic.
+// RCAS uses a 5 tap filter in a cross pattern (same as CAS),
+//    w                n
+//  w 1 w  for taps  w m e 
+//    w                s
+// Where 'w' is the negative lobe weight.
+//  output = (w*(n+e+w+s)+m)/(4*w+1)
+// RCAS solves for 'w' by seeing where the signal might clip out of the {0 to 1} input range,
+//  0 == (w*(n+e+w+s)+m)/(4*w+1) -> w = -m/(n+e+w+s)
+//  1 == (w*(n+e+w+s)+m)/(4*w+1) -> w = (1-m)/(n+e+w+s-4*1)
+// Then chooses the 'w' which results in no clipping, limits 'w', and multiplies by the 'sharp' amount.
+// This solution above has issues with MSAA input as the steps along the gradient cause edge detection issues.
+// So RCAS uses 4x the maximum and 4x the minimum (depending on equation)in place of the individual taps.
+// As well as switching from 'm' to either the minimum or maximum (depending on side), to help in energy conservation.
+// This stabilizes RCAS.
+// RCAS does a simple highpass which is normalized against the local contrast then shaped,
+//       0.25
+//  0.25  -1  0.25
+//       0.25
+// This is used as a noise detection filter, to reduce the effect of RCAS on grain, and focus on real edges.
+//
+//  GLSL example for the required callbacks :
+// 
+//  AH4 FsrRcasLoadH(ASW2 p){return AH4(imageLoad(imgSrc,ASU2(p)));}
+//  void FsrRcasInputH(inout AH1 r,inout AH1 g,inout AH1 b)
+//  {
+//    //do any simple input color conversions here or leave empty if none needed
+//  }
+//  
+//  FsrRcasCon need to be called from the CPU or GPU to set up constants.
+//  Including a GPU example here, the 'con' value would be stored out to a constant buffer.
+// 
+//  AU4 con;
+//  FsrRcasCon(con,
+//   0.0); // The scale is {0.0 := maximum sharpness, to N>0, where N is the number of stops (halving) of the reduction of sharpness}.
+// ---------------
+// RCAS sharpening supports a CAS-like pass-through alpha via,
+//  #define FSR_RCAS_PASSTHROUGH_ALPHA 1
+// RCAS also supports a define to enable a more expensive path to avoid some sharpening of noise.
+// Would suggest it is better to apply film grain after RCAS sharpening (and after scaling) instead of using this define,
+//  #define FSR_RCAS_DENOISE 1
+//==============================================================================================================================
+// This is set at the limit of providing unnatural results for sharpening.
+#define FSR_RCAS_LIMIT (0.25-(1.0/16.0))
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                      CONSTANT SETUP
+//==============================================================================================================================
+// Call to setup required constant values (works on CPU or GPU).
+A_STATIC void FsrRcasCon(
+outAU4 con,
+// The scale is {0.0 := maximum, to N>0, where N is the number of stops (halving) of the reduction of sharpness}.
+AF1 sharpness){
+ // Transform from stops to linear value.
+ sharpness=AExp2F1(-sharpness);
+ varAF2(hSharp)=initAF2(sharpness,sharpness);
+ con[0]=AU1_AF1(sharpness);
+ con[1]=AU1_AH2_AF2(hSharp);
+ con[2]=0;
+ con[3]=0;}
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                   NON-PACKED 32-BIT VERSION
+//==============================================================================================================================
+#if defined(A_GPU)&&defined(FSR_RCAS_F)
+ // Input callback prototypes that need to be implemented by calling shader
+ AF4 FsrRcasLoadF(ASU2 p);
+ void FsrRcasInputF(inout AF1 r,inout AF1 g,inout AF1 b);
+//------------------------------------------------------------------------------------------------------------------------------
+ void FsrRcasF(
+ out AF1 pixR, // Output values, non-vector so port between RcasFilter() and RcasFilterH() is easy.
+ out AF1 pixG,
+ out AF1 pixB,
+ #ifdef FSR_RCAS_PASSTHROUGH_ALPHA
+  out AF1 pixA,
+ #endif
+ AU2 ip, // Integer pixel position in output.
+ AU4 con){ // Constant generated by RcasSetup().
+  // Algorithm uses minimal 3x3 pixel neighborhood.
+  //    b 
+  //  d e f
+  //    h
+  ASU2 sp=ASU2(ip);
+  AF3 b=FsrRcasLoadF(sp+ASU2( 0,-1)).rgb;
+  AF3 d=FsrRcasLoadF(sp+ASU2(-1, 0)).rgb;
+  #ifdef FSR_RCAS_PASSTHROUGH_ALPHA
+   AF4 ee=FsrRcasLoadF(sp);
+   AF3 e=ee.rgb;pixA=ee.a;
+  #else
+   AF3 e=FsrRcasLoadF(sp).rgb;
+  #endif
+  AF3 f=FsrRcasLoadF(sp+ASU2( 1, 0)).rgb;
+  AF3 h=FsrRcasLoadF(sp+ASU2( 0, 1)).rgb;
+  // Rename (32-bit) or regroup (16-bit).
+  AF1 bR=b.r;
+  AF1 bG=b.g;
+  AF1 bB=b.b;
+  AF1 dR=d.r;
+  AF1 dG=d.g;
+  AF1 dB=d.b;
+  AF1 eR=e.r;
+  AF1 eG=e.g;
+  AF1 eB=e.b;
+  AF1 fR=f.r;
+  AF1 fG=f.g;
+  AF1 fB=f.b;
+  AF1 hR=h.r;
+  AF1 hG=h.g;
+  AF1 hB=h.b;
+  // Run optional input transform.
+  FsrRcasInputF(bR,bG,bB);
+  FsrRcasInputF(dR,dG,dB);
+  FsrRcasInputF(eR,eG,eB);
+  FsrRcasInputF(fR,fG,fB);
+  FsrRcasInputF(hR,hG,hB);
+  // Luma times 2.
+  AF1 bL=bB*AF1_(0.5)+(bR*AF1_(0.5)+bG);
+  AF1 dL=dB*AF1_(0.5)+(dR*AF1_(0.5)+dG);
+  AF1 eL=eB*AF1_(0.5)+(eR*AF1_(0.5)+eG);
+  AF1 fL=fB*AF1_(0.5)+(fR*AF1_(0.5)+fG);
+  AF1 hL=hB*AF1_(0.5)+(hR*AF1_(0.5)+hG);
+  // Noise detection.
+  AF1 nz=AF1_(0.25)*bL+AF1_(0.25)*dL+AF1_(0.25)*fL+AF1_(0.25)*hL-eL;
+  nz=ASatF1(abs(nz)*APrxMedRcpF1(AMax3F1(AMax3F1(bL,dL,eL),fL,hL)-AMin3F1(AMin3F1(bL,dL,eL),fL,hL)));
+  nz=AF1_(-0.5)*nz+AF1_(1.0);
+  // Min and max of ring.
+  AF1 mn4R=min(AMin3F1(bR,dR,fR),hR);
+  AF1 mn4G=min(AMin3F1(bG,dG,fG),hG);
+  AF1 mn4B=min(AMin3F1(bB,dB,fB),hB);
+  AF1 mx4R=max(AMax3F1(bR,dR,fR),hR);
+  AF1 mx4G=max(AMax3F1(bG,dG,fG),hG);
+  AF1 mx4B=max(AMax3F1(bB,dB,fB),hB);
+  // Immediate constants for peak range.
+  AF2 peakC=AF2(1.0,-1.0*4.0);
+  // Limiters, these need to be high precision RCPs.
+  AF1 hitMinR=min(mn4R,eR)*ARcpF1(AF1_(4.0)*mx4R);
+  AF1 hitMinG=min(mn4G,eG)*ARcpF1(AF1_(4.0)*mx4G);
+  AF1 hitMinB=min(mn4B,eB)*ARcpF1(AF1_(4.0)*mx4B);
+  AF1 hitMaxR=(peakC.x-max(mx4R,eR))*ARcpF1(AF1_(4.0)*mn4R+peakC.y);
+  AF1 hitMaxG=(peakC.x-max(mx4G,eG))*ARcpF1(AF1_(4.0)*mn4G+peakC.y);
+  AF1 hitMaxB=(peakC.x-max(mx4B,eB))*ARcpF1(AF1_(4.0)*mn4B+peakC.y);
+  AF1 lobeR=max(-hitMinR,hitMaxR);
+  AF1 lobeG=max(-hitMinG,hitMaxG);
+  AF1 lobeB=max(-hitMinB,hitMaxB);
+  AF1 lobe=max(AF1_(-FSR_RCAS_LIMIT),min(AMax3F1(lobeR,lobeG,lobeB),AF1_(0.0)))*AF1_AU1(con.x);
+  // Apply noise removal.
+  #ifdef FSR_RCAS_DENOISE
+   lobe*=nz;
+  #endif
+  // Resolve, which needs the medium precision rcp approximation to avoid visible tonality changes.
+  AF1 rcpL=APrxMedRcpF1(AF1_(4.0)*lobe+AF1_(1.0));
+  pixR=(lobe*bR+lobe*dR+lobe*hR+lobe*fR+eR)*rcpL;
+  pixG=(lobe*bG+lobe*dG+lobe*hG+lobe*fG+eG)*rcpL;
+  pixB=(lobe*bB+lobe*dB+lobe*hB+lobe*fB+eB)*rcpL;
+  return;} 
+#endif
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                  NON-PACKED 16-BIT VERSION
+//==============================================================================================================================
+#if defined(A_GPU)&&defined(A_HALF)&&defined(FSR_RCAS_H)
+ // Input callback prototypes that need to be implemented by calling shader
+ AH4 FsrRcasLoadH(ASW2 p);
+ void FsrRcasInputH(inout AH1 r,inout AH1 g,inout AH1 b);
+//------------------------------------------------------------------------------------------------------------------------------
+ void FsrRcasH(
+ out AH1 pixR, // Output values, non-vector so port between RcasFilter() and RcasFilterH() is easy.
+ out AH1 pixG,
+ out AH1 pixB,
+ #ifdef FSR_RCAS_PASSTHROUGH_ALPHA
+  out AH1 pixA,
+ #endif
+ AU2 ip, // Integer pixel position in output.
+ AU4 con){ // Constant generated by RcasSetup().
+  // Sharpening algorithm uses minimal 3x3 pixel neighborhood.
+  //    b 
+  //  d e f
+  //    h
+  ASW2 sp=ASW2(ip);
+  AH3 b=FsrRcasLoadH(sp+ASW2( 0,-1)).rgb;
+  AH3 d=FsrRcasLoadH(sp+ASW2(-1, 0)).rgb;
+  #ifdef FSR_RCAS_PASSTHROUGH_ALPHA
+   AH4 ee=FsrRcasLoadH(sp);
+   AH3 e=ee.rgb;pixA=ee.a;
+  #else
+   AH3 e=FsrRcasLoadH(sp).rgb;
+  #endif
+  AH3 f=FsrRcasLoadH(sp+ASW2( 1, 0)).rgb;
+  AH3 h=FsrRcasLoadH(sp+ASW2( 0, 1)).rgb;
+  // Rename (32-bit) or regroup (16-bit).
+  AH1 bR=b.r;
+  AH1 bG=b.g;
+  AH1 bB=b.b;
+  AH1 dR=d.r;
+  AH1 dG=d.g;
+  AH1 dB=d.b;
+  AH1 eR=e.r;
+  AH1 eG=e.g;
+  AH1 eB=e.b;
+  AH1 fR=f.r;
+  AH1 fG=f.g;
+  AH1 fB=f.b;
+  AH1 hR=h.r;
+  AH1 hG=h.g;
+  AH1 hB=h.b;
+  // Run optional input transform.
+  FsrRcasInputH(bR,bG,bB);
+  FsrRcasInputH(dR,dG,dB);
+  FsrRcasInputH(eR,eG,eB);
+  FsrRcasInputH(fR,fG,fB);
+  FsrRcasInputH(hR,hG,hB);
+  // Luma times 2.
+  AH1 bL=bB*AH1_(0.5)+(bR*AH1_(0.5)+bG);
+  AH1 dL=dB*AH1_(0.5)+(dR*AH1_(0.5)+dG);
+  AH1 eL=eB*AH1_(0.5)+(eR*AH1_(0.5)+eG);
+  AH1 fL=fB*AH1_(0.5)+(fR*AH1_(0.5)+fG);
+  AH1 hL=hB*AH1_(0.5)+(hR*AH1_(0.5)+hG);
+  // Noise detection.
+  AH1 nz=AH1_(0.25)*bL+AH1_(0.25)*dL+AH1_(0.25)*fL+AH1_(0.25)*hL-eL;
+  nz=ASatH1(abs(nz)*APrxMedRcpH1(AMax3H1(AMax3H1(bL,dL,eL),fL,hL)-AMin3H1(AMin3H1(bL,dL,eL),fL,hL)));
+  nz=AH1_(-0.5)*nz+AH1_(1.0);
+  // Min and max of ring.
+  AH1 mn4R=min(AMin3H1(bR,dR,fR),hR);
+  AH1 mn4G=min(AMin3H1(bG,dG,fG),hG);
+  AH1 mn4B=min(AMin3H1(bB,dB,fB),hB);
+  AH1 mx4R=max(AMax3H1(bR,dR,fR),hR);
+  AH1 mx4G=max(AMax3H1(bG,dG,fG),hG);
+  AH1 mx4B=max(AMax3H1(bB,dB,fB),hB);
+  // Immediate constants for peak range.
+  AH2 peakC=AH2(1.0,-1.0*4.0);
+  // Limiters, these need to be high precision RCPs.
+  AH1 hitMinR=min(mn4R,eR)*ARcpH1(AH1_(4.0)*mx4R);
+  AH1 hitMinG=min(mn4G,eG)*ARcpH1(AH1_(4.0)*mx4G);
+  AH1 hitMinB=min(mn4B,eB)*ARcpH1(AH1_(4.0)*mx4B);
+  AH1 hitMaxR=(peakC.x-max(mx4R,eR))*ARcpH1(AH1_(4.0)*mn4R+peakC.y);
+  AH1 hitMaxG=(peakC.x-max(mx4G,eG))*ARcpH1(AH1_(4.0)*mn4G+peakC.y);
+  AH1 hitMaxB=(peakC.x-max(mx4B,eB))*ARcpH1(AH1_(4.0)*mn4B+peakC.y);
+  AH1 lobeR=max(-hitMinR,hitMaxR);
+  AH1 lobeG=max(-hitMinG,hitMaxG);
+  AH1 lobeB=max(-hitMinB,hitMaxB);
+  AH1 lobe=max(AH1_(-FSR_RCAS_LIMIT),min(AMax3H1(lobeR,lobeG,lobeB),AH1_(0.0)))*AH2_AU1(con.y).x;
+  // Apply noise removal.
+  #ifdef FSR_RCAS_DENOISE
+   lobe*=nz;
+  #endif
+  // Resolve, which needs the medium precision rcp approximation to avoid visible tonality changes.
+  AH1 rcpL=APrxMedRcpH1(AH1_(4.0)*lobe+AH1_(1.0));
+  pixR=(lobe*bR+lobe*dR+lobe*hR+lobe*fR+eR)*rcpL;
+  pixG=(lobe*bG+lobe*dG+lobe*hG+lobe*fG+eG)*rcpL;
+  pixB=(lobe*bB+lobe*dB+lobe*hB+lobe*fB+eB)*rcpL;}
+#endif
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                     PACKED 16-BIT VERSION
+//==============================================================================================================================
+#if defined(A_GPU)&&defined(A_HALF)&&defined(FSR_RCAS_HX2)
+ // Input callback prototypes that need to be implemented by the calling shader
+ AH4 FsrRcasLoadHx2(ASW2 p);
+ void FsrRcasInputHx2(inout AH2 r,inout AH2 g,inout AH2 b);
+//------------------------------------------------------------------------------------------------------------------------------
+ // Can be used to convert from packed Structures of Arrays to Arrays of Structures for store.
+ void FsrRcasDepackHx2(out AH4 pix0,out AH4 pix1,AH2 pixR,AH2 pixG,AH2 pixB){
+  #ifdef A_HLSL
+   // Invoke a slower path for DX only, since it won't allow uninitialized values.
+   pix0.a=pix1.a=0.0;
+  #endif
+  pix0.rgb=AH3(pixR.x,pixG.x,pixB.x);
+  pix1.rgb=AH3(pixR.y,pixG.y,pixB.y);}
+//------------------------------------------------------------------------------------------------------------------------------
+ void FsrRcasHx2(
+ // Output values are for 2 8x8 tiles in a 16x8 region.
+ //  pix<R,G,B>.x =  left 8x8 tile
+ //  pix<R,G,B>.y = right 8x8 tile
+ // This enables later processing to easily be packed as well.
+ out AH2 pixR,
+ out AH2 pixG,
+ out AH2 pixB,
+ #ifdef FSR_RCAS_PASSTHROUGH_ALPHA
+  out AH2 pixA,
+ #endif
+ AU2 ip, // Integer pixel position in output.
+ AU4 con){ // Constant generated by RcasSetup().
+  // No scaling algorithm uses minimal 3x3 pixel neighborhood.
+  ASW2 sp0=ASW2(ip);
+  AH3 b0=FsrRcasLoadHx2(sp0+ASW2( 0,-1)).rgb;
+  AH3 d0=FsrRcasLoadHx2(sp0+ASW2(-1, 0)).rgb;
+  #ifdef FSR_RCAS_PASSTHROUGH_ALPHA
+   AH4 ee0=FsrRcasLoadHx2(sp0);
+   AH3 e0=ee0.rgb;pixA.r=ee0.a;
+  #else
+   AH3 e0=FsrRcasLoadHx2(sp0).rgb;
+  #endif
+  AH3 f0=FsrRcasLoadHx2(sp0+ASW2( 1, 0)).rgb;
+  AH3 h0=FsrRcasLoadHx2(sp0+ASW2( 0, 1)).rgb;
+  ASW2 sp1=sp0+ASW2(8,0);
+  AH3 b1=FsrRcasLoadHx2(sp1+ASW2( 0,-1)).rgb;
+  AH3 d1=FsrRcasLoadHx2(sp1+ASW2(-1, 0)).rgb;
+  #ifdef FSR_RCAS_PASSTHROUGH_ALPHA
+   AH4 ee1=FsrRcasLoadHx2(sp1);
+   AH3 e1=ee1.rgb;pixA.g=ee1.a;
+  #else
+   AH3 e1=FsrRcasLoadHx2(sp1).rgb;
+  #endif
+  AH3 f1=FsrRcasLoadHx2(sp1+ASW2( 1, 0)).rgb;
+  AH3 h1=FsrRcasLoadHx2(sp1+ASW2( 0, 1)).rgb;
+  // Arrays of Structures to Structures of Arrays conversion.
+  AH2 bR=AH2(b0.r,b1.r);
+  AH2 bG=AH2(b0.g,b1.g);
+  AH2 bB=AH2(b0.b,b1.b);
+  AH2 dR=AH2(d0.r,d1.r);
+  AH2 dG=AH2(d0.g,d1.g);
+  AH2 dB=AH2(d0.b,d1.b);
+  AH2 eR=AH2(e0.r,e1.r);
+  AH2 eG=AH2(e0.g,e1.g);
+  AH2 eB=AH2(e0.b,e1.b);
+  AH2 fR=AH2(f0.r,f1.r);
+  AH2 fG=AH2(f0.g,f1.g);
+  AH2 fB=AH2(f0.b,f1.b);
+  AH2 hR=AH2(h0.r,h1.r);
+  AH2 hG=AH2(h0.g,h1.g);
+  AH2 hB=AH2(h0.b,h1.b);
+  // Run optional input transform.
+  FsrRcasInputHx2(bR,bG,bB);
+  FsrRcasInputHx2(dR,dG,dB);
+  FsrRcasInputHx2(eR,eG,eB);
+  FsrRcasInputHx2(fR,fG,fB);
+  FsrRcasInputHx2(hR,hG,hB);
+  // Luma times 2.
+  AH2 bL=bB*AH2_(0.5)+(bR*AH2_(0.5)+bG);
+  AH2 dL=dB*AH2_(0.5)+(dR*AH2_(0.5)+dG);
+  AH2 eL=eB*AH2_(0.5)+(eR*AH2_(0.5)+eG);
+  AH2 fL=fB*AH2_(0.5)+(fR*AH2_(0.5)+fG);
+  AH2 hL=hB*AH2_(0.5)+(hR*AH2_(0.5)+hG);
+  // Noise detection.
+  AH2 nz=AH2_(0.25)*bL+AH2_(0.25)*dL+AH2_(0.25)*fL+AH2_(0.25)*hL-eL;
+  nz=ASatH2(abs(nz)*APrxMedRcpH2(AMax3H2(AMax3H2(bL,dL,eL),fL,hL)-AMin3H2(AMin3H2(bL,dL,eL),fL,hL)));
+  nz=AH2_(-0.5)*nz+AH2_(1.0);
+  // Min and max of ring.
+  AH2 mn4R=min(AMin3H2(bR,dR,fR),hR);
+  AH2 mn4G=min(AMin3H2(bG,dG,fG),hG);
+  AH2 mn4B=min(AMin3H2(bB,dB,fB),hB);
+  AH2 mx4R=max(AMax3H2(bR,dR,fR),hR);
+  AH2 mx4G=max(AMax3H2(bG,dG,fG),hG);
+  AH2 mx4B=max(AMax3H2(bB,dB,fB),hB);
+  // Immediate constants for peak range.
+  AH2 peakC=AH2(1.0,-1.0*4.0);
+  // Limiters, these need to be high precision RCPs.
+  AH2 hitMinR=min(mn4R,eR)*ARcpH2(AH2_(4.0)*mx4R);
+  AH2 hitMinG=min(mn4G,eG)*ARcpH2(AH2_(4.0)*mx4G);
+  AH2 hitMinB=min(mn4B,eB)*ARcpH2(AH2_(4.0)*mx4B);
+  AH2 hitMaxR=(peakC.x-max(mx4R,eR))*ARcpH2(AH2_(4.0)*mn4R+peakC.y);
+  AH2 hitMaxG=(peakC.x-max(mx4G,eG))*ARcpH2(AH2_(4.0)*mn4G+peakC.y);
+  AH2 hitMaxB=(peakC.x-max(mx4B,eB))*ARcpH2(AH2_(4.0)*mn4B+peakC.y);
+  AH2 lobeR=max(-hitMinR,hitMaxR);
+  AH2 lobeG=max(-hitMinG,hitMaxG);
+  AH2 lobeB=max(-hitMinB,hitMaxB);
+  AH2 lobe=max(AH2_(-FSR_RCAS_LIMIT),min(AMax3H2(lobeR,lobeG,lobeB),AH2_(0.0)))*AH2_(AH2_AU1(con.y).x);
+  // Apply noise removal.
+  #ifdef FSR_RCAS_DENOISE
+   lobe*=nz;
+  #endif
+  // Resolve, which needs the medium precision rcp approximation to avoid visible tonality changes.
+  AH2 rcpL=APrxMedRcpH2(AH2_(4.0)*lobe+AH2_(1.0));
+  pixR=(lobe*bR+lobe*dR+lobe*hR+lobe*fR+eR)*rcpL;
+  pixG=(lobe*bG+lobe*dG+lobe*hG+lobe*fG+eG)*rcpL;
+  pixB=(lobe*bB+lobe*dB+lobe*hB+lobe*fB+eB)*rcpL;}
+#endif
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//
+//                                          FSR - [LFGA] LINEAR FILM GRAIN APPLICATOR
+//
+//------------------------------------------------------------------------------------------------------------------------------
+// Adding output-resolution film grain after scaling is a good way to mask both rendering and scaling artifacts.
+// Suggest using tiled blue noise as film grain input, with peak noise frequency set for a specific look and feel.
+// The 'Lfga*()' functions provide a convenient way to introduce grain.
+// These functions limit grain based on distance to signal limits.
+// This is done so that the grain is temporally energy preserving, and thus won't modify image tonality.
+// Grain application should be done in a linear colorspace.
+// The grain should be temporally changing, but have a temporal sum per pixel that adds to zero (non-biased).
+//------------------------------------------------------------------------------------------------------------------------------
+// Usage,
+//   FsrLfga*(
+//    color, // In/out linear colorspace color {0 to 1} ranged.
+//    grain, // Per pixel grain texture value {-0.5 to 0.5} ranged, input is 3-channel to support colored grain.
+//    amount); // Amount of grain (0 to 1} ranged.
+//------------------------------------------------------------------------------------------------------------------------------
+// Example if grain texture is monochrome: 'FsrLfgaF(color,AF3_(grain),amount)'
+//==============================================================================================================================
+#if defined(A_GPU)
+ // Maximum grain is the minimum distance to the signal limit.
+ void FsrLfgaF(inout AF3 c,AF3 t,AF1 a){c+=(t*AF3_(a))*min(AF3_(1.0)-c,c);}
+#endif
+//==============================================================================================================================
+#if defined(A_GPU)&&defined(A_HALF)
+ // Half precision version (slower).
+ void FsrLfgaH(inout AH3 c,AH3 t,AH1 a){c+=(t*AH3_(a))*min(AH3_(1.0)-c,c);}
+//------------------------------------------------------------------------------------------------------------------------------
+ // Packed half precision version (faster).
+ void FsrLfgaHx2(inout AH2 cR,inout AH2 cG,inout AH2 cB,AH2 tR,AH2 tG,AH2 tB,AH1 a){
+  cR+=(tR*AH2_(a))*min(AH2_(1.0)-cR,cR);cG+=(tG*AH2_(a))*min(AH2_(1.0)-cG,cG);cB+=(tB*AH2_(a))*min(AH2_(1.0)-cB,cB);}
+#endif
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//
+//                                          FSR - [SRTM] SIMPLE REVERSIBLE TONE-MAPPER
+//
+//------------------------------------------------------------------------------------------------------------------------------
+// This provides a way to take linear HDR color {0 to FP16_MAX} and convert it into a temporary {0 to 1} ranged post-tonemapped linear.
+// The tonemapper preserves RGB ratio, which helps maintain HDR color bleed during filtering.
+//------------------------------------------------------------------------------------------------------------------------------
+// Reversible tonemapper usage,
+//  FsrSrtm*(color); // {0 to FP16_MAX} converted to {0 to 1}.
+//  FsrSrtmInv*(color); // {0 to 1} converted into {0 to 32768, output peak safe for FP16}.
+//==============================================================================================================================
+#if defined(A_GPU)
+ void FsrSrtmF(inout AF3 c){c*=AF3_(ARcpF1(AMax3F1(c.r,c.g,c.b)+AF1_(1.0)));}
+ // The extra max solves the c=1.0 case (which is a /0).
+ void FsrSrtmInvF(inout AF3 c){c*=AF3_(ARcpF1(max(AF1_(1.0/32768.0),AF1_(1.0)-AMax3F1(c.r,c.g,c.b))));}
+#endif
+//==============================================================================================================================
+#if defined(A_GPU)&&defined(A_HALF)
+ void FsrSrtmH(inout AH3 c){c*=AH3_(ARcpH1(AMax3H1(c.r,c.g,c.b)+AH1_(1.0)));}
+ void FsrSrtmInvH(inout AH3 c){c*=AH3_(ARcpH1(max(AH1_(1.0/32768.0),AH1_(1.0)-AMax3H1(c.r,c.g,c.b))));}
+//------------------------------------------------------------------------------------------------------------------------------
+ void FsrSrtmHx2(inout AH2 cR,inout AH2 cG,inout AH2 cB){
+  AH2 rcp=ARcpH2(AMax3H2(cR,cG,cB)+AH2_(1.0));cR*=rcp;cG*=rcp;cB*=rcp;}
+ void FsrSrtmInvHx2(inout AH2 cR,inout AH2 cG,inout AH2 cB){
+  AH2 rcp=ARcpH2(max(AH2_(1.0/32768.0),AH2_(1.0)-AMax3H2(cR,cG,cB)));cR*=rcp;cG*=rcp;cB*=rcp;}
+#endif
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//
+//                                       FSR - [TEPD] TEMPORAL ENERGY PRESERVING DITHER
+//
+//------------------------------------------------------------------------------------------------------------------------------
+// Temporally energy preserving dithered {0 to 1} linear to gamma 2.0 conversion.
+// Gamma 2.0 is used so that the conversion back to linear is just to square the color.
+// The conversion comes in 8-bit and 10-bit modes, designed for output to 8-bit UNORM or 10:10:10:2 respectively.
+// Given good non-biased temporal blue noise as dither input,
+// the output dither will temporally conserve energy.
+// This is done by choosing the linear nearest step point instead of perceptual nearest.
+// See code below for details.
+//------------------------------------------------------------------------------------------------------------------------------
+// DX SPEC RULES FOR FLOAT->UNORM 8-BIT CONVERSION
+// ===============================================
+// - Output is 'uint(floor(saturate(n)*255.0+0.5))'.
+// - Thus rounding is to nearest.
+// - NaN gets converted to zero.
+// - INF is clamped to {0.0 to 1.0}.
+//==============================================================================================================================
+#if defined(A_GPU)
+ // Hand tuned integer position to dither value, with more values than simple checkerboard.
+ // Only 32-bit has enough precision for this compddation.
+ // Output is {0 to <1}.
+ AF1 FsrTepdDitF(AU2 p,AU1 f){
+  AF1 x=AF1_(p.x+f);
+  AF1 y=AF1_(p.y);
+  // The 1.61803 golden ratio.
+  AF1 a=AF1_((1.0+sqrt(5.0))/2.0);
+  // Number designed to provide a good visual pattern.
+  AF1 b=AF1_(1.0/3.69);
+  x=x*a+(y*b);
+  return AFractF1(x);}
+//------------------------------------------------------------------------------------------------------------------------------
+ // This version is 8-bit gamma 2.0.
+ // The 'c' input is {0 to 1}.
+ // Output is {0 to 1} ready for image store.
+ void FsrTepdC8F(inout AF3 c,AF1 dit){
+  AF3 n=sqrt(c);
+  n=floor(n*AF3_(255.0))*AF3_(1.0/255.0);
+  AF3 a=n*n;
+  AF3 b=n+AF3_(1.0/255.0);b=b*b;
+  // Ratio of 'a' to 'b' required to produce 'c'.
+  // APrxLoRcpF1() won't work here (at least for very high dynamic ranges).
+  // APrxMedRcpF1() is an IADD,FMA,MUL.
+  AF3 r=(c-b)*APrxMedRcpF3(a-b);
+  // Use the ratio as a cutoff to choose 'a' or 'b'.
+  // AGtZeroF1() is a MUL.
+  c=ASatF3(n+AGtZeroF3(AF3_(dit)-r)*AF3_(1.0/255.0));}
+//------------------------------------------------------------------------------------------------------------------------------
+ // This version is 10-bit gamma 2.0.
+ // The 'c' input is {0 to 1}.
+ // Output is {0 to 1} ready for image store.
+ void FsrTepdC10F(inout AF3 c,AF1 dit){
+  AF3 n=sqrt(c);
+  n=floor(n*AF3_(1023.0))*AF3_(1.0/1023.0);
+  AF3 a=n*n;
+  AF3 b=n+AF3_(1.0/1023.0);b=b*b;
+  AF3 r=(c-b)*APrxMedRcpF3(a-b);
+  c=ASatF3(n+AGtZeroF3(AF3_(dit)-r)*AF3_(1.0/1023.0));}
+#endif
+//==============================================================================================================================
+#if defined(A_GPU)&&defined(A_HALF)
+ AH1 FsrTepdDitH(AU2 p,AU1 f){
+  AF1 x=AF1_(p.x+f);
+  AF1 y=AF1_(p.y);
+  AF1 a=AF1_((1.0+sqrt(5.0))/2.0);
+  AF1 b=AF1_(1.0/3.69);
+  x=x*a+(y*b);
+  return AH1(AFractF1(x));}
+//------------------------------------------------------------------------------------------------------------------------------
+ void FsrTepdC8H(inout AH3 c,AH1 dit){
+  AH3 n=sqrt(c);
+  n=floor(n*AH3_(255.0))*AH3_(1.0/255.0);
+  AH3 a=n*n;
+  AH3 b=n+AH3_(1.0/255.0);b=b*b;
+  AH3 r=(c-b)*APrxMedRcpH3(a-b);
+  c=ASatH3(n+AGtZeroH3(AH3_(dit)-r)*AH3_(1.0/255.0));}
+//------------------------------------------------------------------------------------------------------------------------------
+ void FsrTepdC10H(inout AH3 c,AH1 dit){
+  AH3 n=sqrt(c);
+  n=floor(n*AH3_(1023.0))*AH3_(1.0/1023.0);
+  AH3 a=n*n;
+  AH3 b=n+AH3_(1.0/1023.0);b=b*b;
+  AH3 r=(c-b)*APrxMedRcpH3(a-b);
+  c=ASatH3(n+AGtZeroH3(AH3_(dit)-r)*AH3_(1.0/1023.0));}
+//==============================================================================================================================
+ // This computes dither for positions 'p' and 'p+{8,0}'.
+ AH2 FsrTepdDitHx2(AU2 p,AU1 f){
+  AF2 x;
+  x.x=AF1_(p.x+f);
+  x.y=x.x+AF1_(8.0);
+  AF1 y=AF1_(p.y);
+  AF1 a=AF1_((1.0+sqrt(5.0))/2.0);
+  AF1 b=AF1_(1.0/3.69);
+  x=x*AF2_(a)+AF2_(y*b);
+  return AH2(AFractF2(x));}
+//------------------------------------------------------------------------------------------------------------------------------
+ void FsrTepdC8Hx2(inout AH2 cR,inout AH2 cG,inout AH2 cB,AH2 dit){
+  AH2 nR=sqrt(cR);
+  AH2 nG=sqrt(cG);
+  AH2 nB=sqrt(cB);
+  nR=floor(nR*AH2_(255.0))*AH2_(1.0/255.0);
+  nG=floor(nG*AH2_(255.0))*AH2_(1.0/255.0);
+  nB=floor(nB*AH2_(255.0))*AH2_(1.0/255.0);
+  AH2 aR=nR*nR;
+  AH2 aG=nG*nG;
+  AH2 aB=nB*nB;
+  AH2 bR=nR+AH2_(1.0/255.0);bR=bR*bR;
+  AH2 bG=nG+AH2_(1.0/255.0);bG=bG*bG;
+  AH2 bB=nB+AH2_(1.0/255.0);bB=bB*bB;
+  AH2 rR=(cR-bR)*APrxMedRcpH2(aR-bR);
+  AH2 rG=(cG-bG)*APrxMedRcpH2(aG-bG);
+  AH2 rB=(cB-bB)*APrxMedRcpH2(aB-bB);
+  cR=ASatH2(nR+AGtZeroH2(dit-rR)*AH2_(1.0/255.0));
+  cG=ASatH2(nG+AGtZeroH2(dit-rG)*AH2_(1.0/255.0));
+  cB=ASatH2(nB+AGtZeroH2(dit-rB)*AH2_(1.0/255.0));}
+//------------------------------------------------------------------------------------------------------------------------------
+ void FsrTepdC10Hx2(inout AH2 cR,inout AH2 cG,inout AH2 cB,AH2 dit){
+  AH2 nR=sqrt(cR);
+  AH2 nG=sqrt(cG);
+  AH2 nB=sqrt(cB);
+  nR=floor(nR*AH2_(1023.0))*AH2_(1.0/1023.0);
+  nG=floor(nG*AH2_(1023.0))*AH2_(1.0/1023.0);
+  nB=floor(nB*AH2_(1023.0))*AH2_(1.0/1023.0);
+  AH2 aR=nR*nR;
+  AH2 aG=nG*nG;
+  AH2 aB=nB*nB;
+  AH2 bR=nR+AH2_(1.0/1023.0);bR=bR*bR;
+  AH2 bG=nG+AH2_(1.0/1023.0);bG=bG*bG;
+  AH2 bB=nB+AH2_(1.0/1023.0);bB=bB*bB;
+  AH2 rR=(cR-bR)*APrxMedRcpH2(aR-bR);
+  AH2 rG=(cG-bG)*APrxMedRcpH2(aG-bG);
+  AH2 rB=(cB-bB)*APrxMedRcpH2(aB-bB);
+  cR=ASatH2(nR+AGtZeroH2(dit-rR)*AH2_(1.0/1023.0));
+  cG=ASatH2(nG+AGtZeroH2(dit-rG)*AH2_(1.0/1023.0));
+  cB=ASatH2(nB+AGtZeroH2(dit-rB)*AH2_(1.0/1023.0));}
+#endif
diff --git a/src/Ryujinx.Graphics.OpenGL/Effects/Shaders/fsr_scaling.glsl b/src/Ryujinx.Graphics.OpenGL/Effects/Shaders/fsr_scaling.glsl
new file mode 100644
index 00000000..8e8755db
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Effects/Shaders/fsr_scaling.glsl
@@ -0,0 +1,88 @@
+#version 430 core
+precision mediump float;
+layout (local_size_x = 64) in;
+layout(rgba8, binding = 0, location=0) uniform image2D imgOutput;
+layout( location=1 ) uniform sampler2D Source;
+layout( location=2 ) uniform float srcX0;
+layout( location=3 ) uniform float srcX1;
+layout( location=4 ) uniform float srcY0;
+layout( location=5 ) uniform float srcY1;
+layout( location=6 ) uniform float dstX0;
+layout( location=7 ) uniform float dstX1;
+layout( location=8 ) uniform float dstY0;
+layout( location=9 ) uniform float dstY1;
+layout( location=10 ) uniform float scaleX;
+layout( location=11 ) uniform float scaleY;
+
+#define A_GPU 1
+#define A_GLSL 1
+#include "ffx_a.h"
+
+#define FSR_EASU_F 1
+AU4 con0, con1, con2, con3;
+float srcW, srcH, dstW, dstH;
+vec2 bLeft, tRight;
+
+AF2 translate(AF2 pos) {
+    return AF2(pos.x * scaleX, pos.y * scaleY);
+}
+
+void setBounds(vec2 bottomLeft, vec2 topRight) {
+    bLeft = bottomLeft;
+    tRight = topRight;
+}
+
+AF2 translateDest(AF2 pos) {
+    AF2 translatedPos = AF2(pos.x, pos.y);
+    translatedPos.x = dstX1 < dstX0 ? dstX1 - translatedPos.x : translatedPos.x;
+    translatedPos.y = dstY0 > dstY1 ? dstY0 + dstY1 - translatedPos.y - 1: translatedPos.y;
+    return translatedPos;
+}
+
+AF4 FsrEasuRF(AF2 p) { AF4 res = textureGather(Source, translate(p), 0); return res; }
+AF4 FsrEasuGF(AF2 p) { AF4 res = textureGather(Source, translate(p), 1); return res; }
+AF4 FsrEasuBF(AF2 p) { AF4 res = textureGather(Source, translate(p), 2); return res; }
+
+#include "ffx_fsr1.h"
+
+float insideBox(vec2 v) {
+    vec2 s = step(bLeft, v) - step(tRight, v);
+    return s.x * s.y;   
+}
+
+void CurrFilter(AU2 pos)
+{
+    if((insideBox(vec2(pos.x, pos.y))) == 0) {
+        imageStore(imgOutput, ASU2(pos.x, pos.y), AF4(0,0,0,1));
+       return;
+    }
+    AF3 c;
+    FsrEasuF(c, AU2(pos.x - bLeft.x, pos.y - bLeft.y), con0, con1, con2, con3);
+    imageStore(imgOutput, ASU2(translateDest(pos)), AF4(c, 1));
+}
+
+void main() {
+    srcW = abs(srcX1 - srcX0);
+    srcH = abs(srcY1 - srcY0);
+    dstW = abs(dstX1 - dstX0);
+    dstH = abs(dstY1 - dstY0);
+
+    AU2 gxy = ARmp8x8(gl_LocalInvocationID.x) + AU2(gl_WorkGroupID.x << 4u, gl_WorkGroupID.y << 4u);
+
+    setBounds(vec2(dstX0 < dstX1 ? dstX0 : dstX1, dstY0 < dstY1 ? dstY0 : dstY1),
+        vec2(dstX1 > dstX0 ? dstX1 : dstX0, dstY1 > dstY0 ? dstY1 : dstY0));
+
+    // Upscaling
+    FsrEasuCon(con0, con1, con2, con3,
+        srcW, srcH,  // Viewport size (top left aligned) in the input image which is to be scaled.
+        srcW, srcH,  // The size of the input image.
+        dstW, dstH); // The output resolution.
+
+    CurrFilter(gxy);
+    gxy.x += 8u;
+    CurrFilter(gxy);
+    gxy.y += 8u;
+    CurrFilter(gxy);
+    gxy.x -= 8u;
+    CurrFilter(gxy);
+}
\ No newline at end of file
diff --git a/src/Ryujinx.Graphics.OpenGL/Effects/Shaders/fsr_sharpening.glsl b/src/Ryujinx.Graphics.OpenGL/Effects/Shaders/fsr_sharpening.glsl
new file mode 100644
index 00000000..d3b98729
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Effects/Shaders/fsr_sharpening.glsl
@@ -0,0 +1,37 @@
+#version 430 core
+precision mediump float;
+layout (local_size_x = 64) in;
+layout(rgba8, binding = 0, location=0) uniform image2D imgOutput;
+layout( location=1 ) uniform sampler2D source;
+layout( location=2 ) uniform float sharpening;
+
+#define A_GPU 1
+#define A_GLSL 1
+#include "ffx_a.h"
+
+#define FSR_RCAS_F 1
+AU4 con0;
+
+AF4 FsrRcasLoadF(ASU2 p) { return AF4(texelFetch(source, p, 0)); }
+void FsrRcasInputF(inout AF1 r, inout AF1 g, inout AF1 b) {}
+
+#include "ffx_fsr1.h"
+
+void CurrFilter(AU2 pos)
+{
+    AF3 c;
+    FsrRcasF(c.r, c.g, c.b, pos, con0);
+    imageStore(imgOutput, ASU2(pos), AF4(c, 1));
+}
+
+void main() {
+    FsrRcasCon(con0, sharpening);
+    AU2 gxy = ARmp8x8(gl_LocalInvocationID.x) + AU2(gl_WorkGroupID.x << 4u, gl_WorkGroupID.y << 4u);
+    CurrFilter(gxy);
+    gxy.x += 8u;
+    CurrFilter(gxy);
+    gxy.y += 8u;
+    CurrFilter(gxy);
+    gxy.x -= 8u;
+    CurrFilter(gxy);
+}
\ No newline at end of file
diff --git a/src/Ryujinx.Graphics.OpenGL/Effects/Shaders/fxaa.glsl b/src/Ryujinx.Graphics.OpenGL/Effects/Shaders/fxaa.glsl
new file mode 100644
index 00000000..8bdcbca6
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Effects/Shaders/fxaa.glsl
@@ -0,0 +1,1174 @@
+/*============================================================================
+
+
+                    NVIDIA FXAA 3.11 by TIMOTHY LOTTES
+
+
+------------------------------------------------------------------------------
+COPYRIGHT (C) 2010, 2011 NVIDIA CORPORATION. ALL RIGHTS RESERVED.
+------------------------------------------------------------------------------
+TO THE MAXIMUM EXTENT PERMITTED BY APPLICABLE LAW, THIS SOFTWARE IS PROVIDED
+*AS IS* AND NVIDIA AND ITS SUPPLIERS DISCLAIM ALL WARRANTIES, EITHER EXPRESS
+OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, IMPLIED WARRANTIES OF
+MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. IN NO EVENT SHALL NVIDIA
+OR ITS SUPPLIERS BE LIABLE FOR ANY SPECIAL, INCIDENTAL, INDIRECT, OR
+CONSEQUENTIAL DAMAGES WHATSOEVER (INCLUDING, WITHOUT LIMITATION, DAMAGES FOR
+LOSS OF BUSINESS PROFITS, BUSINESS INTERRUPTION, LOSS OF BUSINESS INFORMATION,
+OR ANY OTHER PECUNIARY LOSS) ARISING OUT OF THE USE OF OR INABILITY TO USE
+THIS SOFTWARE, EVEN IF NVIDIA HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH
+DAMAGES.
+
+------------------------------------------------------------------------------
+                           INTEGRATION CHECKLIST
+------------------------------------------------------------------------------
+(1.)
+In the shader source, setup defines for the desired configuration.
+When providing multiple shaders (for different presets),
+simply setup the defines differently in multiple files.
+Example,
+
+  #define FXAA_PC 1
+  #define FXAA_HLSL_5 1
+  #define FXAA_QUALITY__PRESET 12
+
+Or,
+
+  #define FXAA_360 1
+  
+Or,
+
+  #define FXAA_PS3 1
+  
+Etc.
+
+(2.)
+Then include this file,
+
+  #include "Fxaa3_11.h"
+
+(3.)
+Then call the FXAA pixel shader from within your desired shader.
+Look at the FXAA Quality FxaaPixelShader() for docs on inputs.
+As for FXAA 3.11 all inputs for all shaders are the same 
+to enable easy porting between platforms.
+
+  return FxaaPixelShader(...);
+
+(4.)
+Insure pass prior to FXAA outputs RGBL (see next section).
+Or use,
+
+  #define FXAA_GREEN_AS_LUMA 1
+
+(5.)
+Setup engine to provide the following constants
+which are used in the FxaaPixelShader() inputs,
+
+  FxaaFloat2 fxaaQualityRcpFrame,
+  FxaaFloat4 fxaaConsoleRcpFrameOpt,
+  FxaaFloat4 fxaaConsoleRcpFrameOpt2,
+  FxaaFloat4 fxaaConsole360RcpFrameOpt2,
+  FxaaFloat fxaaQualitySubpix,
+  FxaaFloat fxaaQualityEdgeThreshold,
+  FxaaFloat fxaaQualityEdgeThresholdMin,
+  FxaaFloat fxaaConsoleEdgeSharpness,
+  FxaaFloat fxaaConsoleEdgeThreshold,
+  FxaaFloat fxaaConsoleEdgeThresholdMin,
+  FxaaFloat4 fxaaConsole360ConstDir
+
+Look at the FXAA Quality FxaaPixelShader() for docs on inputs.
+
+(6.)
+Have FXAA vertex shader run as a full screen triangle,
+and output "pos" and "fxaaConsolePosPos" 
+such that inputs in the pixel shader provide,
+
+  // {xy} = center of pixel
+  FxaaFloat2 pos,
+
+  // {xy__} = upper left of pixel
+  // {__zw} = lower right of pixel
+  FxaaFloat4 fxaaConsolePosPos,
+
+(7.)
+Insure the texture sampler(s) used by FXAA are set to bilinear filtering.
+
+
+------------------------------------------------------------------------------
+                    INTEGRATION - RGBL AND COLORSPACE
+------------------------------------------------------------------------------
+FXAA3 requires RGBL as input unless the following is set, 
+
+  #define FXAA_GREEN_AS_LUMA 1
+
+In which case the engine uses green in place of luma,
+and requires RGB input is in a non-linear colorspace.
+
+RGB should be LDR (low dynamic range).
+Specifically do FXAA after tonemapping.
+
+RGB data as returned by a texture fetch can be non-linear,
+or linear when FXAA_GREEN_AS_LUMA is not set.
+Note an "sRGB format" texture counts as linear,
+because the result of a texture fetch is linear data.
+Regular "RGBA8" textures in the sRGB colorspace are non-linear.
+
+If FXAA_GREEN_AS_LUMA is not set,
+luma must be stored in the alpha channel prior to running FXAA.
+This luma should be in a perceptual space (could be gamma 2.0).
+Example pass before FXAA where output is gamma 2.0 encoded,
+
+  color.rgb = ToneMap(color.rgb); // linear color output
+  color.rgb = sqrt(color.rgb);    // gamma 2.0 color output
+  return color;
+
+To use FXAA,
+
+  color.rgb = ToneMap(color.rgb);  // linear color output
+  color.rgb = sqrt(color.rgb);     // gamma 2.0 color output
+  color.a = dot(color.rgb, FxaaFloat3(0.299, 0.587, 0.114)); // compute luma
+  return color;
+
+Another example where output is linear encoded,
+say for instance writing to an sRGB formated render target,
+where the render target does the conversion back to sRGB after blending,
+
+  color.rgb = ToneMap(color.rgb); // linear color output
+  return color;
+
+To use FXAA,
+
+  color.rgb = ToneMap(color.rgb); // linear color output
+  color.a = sqrt(dot(color.rgb, FxaaFloat3(0.299, 0.587, 0.114))); // compute luma
+  return color;
+
+Getting luma correct is required for the algorithm to work correctly.
+
+
+------------------------------------------------------------------------------
+                          BEING LINEARLY CORRECT?
+------------------------------------------------------------------------------
+Applying FXAA to a framebuffer with linear RGB color will look worse.
+This is very counter intuitive, but happends to be true in this case.
+The reason is because dithering artifacts will be more visiable 
+in a linear colorspace.
+
+
+------------------------------------------------------------------------------
+                             COMPLEX INTEGRATION
+------------------------------------------------------------------------------
+Q. What if the engine is blending into RGB before wanting to run FXAA?
+
+A. In the last opaque pass prior to FXAA,
+   have the pass write out luma into alpha.
+   Then blend into RGB only.
+   FXAA should be able to run ok
+   assuming the blending pass did not any add aliasing.
+   This should be the common case for particles and common blending passes.
+
+A. Or use FXAA_GREEN_AS_LUMA.
+
+============================================================================*/
+
+#version 430 core
+
+layout(local_size_x = 16, local_size_y = 16) in;
+layout(rgba8, binding = 0) uniform image2D imgOutput;
+
+uniform sampler2D inputTexture;
+layout(location=0) uniform vec2 invResolution;
+
+#define FXAA_QUALITY__PRESET 12
+#define FXAA_GREEN_AS_LUMA 1
+#define FXAA_PC 1
+#define FXAA_GLSL_130 1
+
+
+/*============================================================================
+
+                             INTEGRATION KNOBS
+
+/*==========================================================================*/
+#ifndef FXAA_PC
+    //
+    // FXAA Quality
+    // The high quality PC algorithm.
+    //
+    #define FXAA_PC 0
+#endif
+/*--------------------------------------------------------------------------*/
+#ifndef FXAA_GLSL_120
+    #define FXAA_GLSL_120 0
+#endif
+/*--------------------------------------------------------------------------*/
+#ifndef FXAA_GLSL_130
+    #define FXAA_GLSL_130 0
+#endif
+/*==========================================================================*/
+#ifndef FXAA_GREEN_AS_LUMA
+    //
+    // For those using non-linear color,
+    // and either not able to get luma in alpha, or not wanting to,
+    // this enables FXAA to run using green as a proxy for luma.
+    // So with this enabled, no need to pack luma in alpha.
+    //
+    // This will turn off AA on anything which lacks some amount of green.
+    // Pure red and blue or combination of only R and B, will get no AA.
+    //
+    // Might want to lower the settings for both,
+    //    fxaaConsoleEdgeThresholdMin
+    //    fxaaQualityEdgeThresholdMin
+    // In order to insure AA does not get turned off on colors 
+    // which contain a minor amount of green.
+    //
+    // 1 = On.
+    // 0 = Off.
+    //
+    #define FXAA_GREEN_AS_LUMA 0
+#endif
+/*--------------------------------------------------------------------------*/
+#ifndef FXAA_EARLY_EXIT
+    //
+    // Controls algorithm's early exit path.
+    // On PS3 turning this ON adds 2 cycles to the shader.
+    // On 360 turning this OFF adds 10ths of a millisecond to the shader.
+    // Turning this off on console will result in a more blurry image.
+    // So this defaults to on.
+    //
+    // 1 = On.
+    // 0 = Off.
+    //
+    #define FXAA_EARLY_EXIT 1
+#endif
+/*--------------------------------------------------------------------------*/
+#ifndef FXAA_DISCARD
+    //
+    // Only valid for PC OpenGL currently.
+    // Probably will not work when FXAA_GREEN_AS_LUMA = 1.
+    //
+    // 1 = Use discard on pixels which don't need AA.
+    //     For APIs which enable concurrent TEX+ROP from same surface.
+    // 0 = Return unchanged color on pixels which don't need AA.
+    //
+    #define FXAA_DISCARD 0
+#endif
+/*--------------------------------------------------------------------------*/
+#ifndef FXAA_FAST_PIXEL_OFFSET
+    //
+    // Used for GLSL 120 only.
+    //
+    // 1 = GL API supports fast pixel offsets
+    // 0 = do not use fast pixel offsets
+    //
+    #ifdef GL_EXT_gpu_shader4
+        #define FXAA_FAST_PIXEL_OFFSET 1
+    #endif
+    #ifdef GL_NV_gpu_shader5
+        #define FXAA_FAST_PIXEL_OFFSET 1
+    #endif
+    #ifdef GL_ARB_gpu_shader5
+        #define FXAA_FAST_PIXEL_OFFSET 1
+    #endif
+    #ifndef FXAA_FAST_PIXEL_OFFSET
+        #define FXAA_FAST_PIXEL_OFFSET 0
+    #endif
+#endif
+/*--------------------------------------------------------------------------*/
+#ifndef FXAA_GATHER4_ALPHA
+    //
+    // 1 = API supports gather4 on alpha channel.
+    // 0 = API does not support gather4 on alpha channel.
+    //
+    #if (FXAA_HLSL_5 == 1)
+        #define FXAA_GATHER4_ALPHA 1
+    #endif
+    #ifdef GL_ARB_gpu_shader5
+        #define FXAA_GATHER4_ALPHA 1
+    #endif
+    #ifdef GL_NV_gpu_shader5
+        #define FXAA_GATHER4_ALPHA 1
+    #endif
+    #ifndef FXAA_GATHER4_ALPHA
+        #define FXAA_GATHER4_ALPHA 0
+    #endif
+#endif
+
+/*============================================================================
+                        FXAA QUALITY - TUNING KNOBS
+------------------------------------------------------------------------------
+NOTE the other tuning knobs are now in the shader function inputs!
+============================================================================*/
+#ifndef FXAA_QUALITY__PRESET
+    //
+    // Choose the quality preset.
+    // This needs to be compiled into the shader as it effects code.
+    // Best option to include multiple presets is to 
+    // in each shader define the preset, then include this file.
+    // 
+    // OPTIONS
+    // -----------------------------------------------------------------------
+    // 10 to 15 - default medium dither (10=fastest, 15=highest quality)
+    // 20 to 29 - less dither, more expensive (20=fastest, 29=highest quality)
+    // 39       - no dither, very expensive 
+    //
+    // NOTES
+    // -----------------------------------------------------------------------
+    // 12 = slightly faster then FXAA 3.9 and higher edge quality (default)
+    // 13 = about same speed as FXAA 3.9 and better than 12
+    // 23 = closest to FXAA 3.9 visually and performance wise
+    //  _ = the lowest digit is directly related to performance
+    // _  = the highest digit is directly related to style
+    // 
+    #define FXAA_QUALITY__PRESET 12
+#endif
+
+
+/*============================================================================
+
+                           FXAA QUALITY - PRESETS
+
+============================================================================*/
+
+/*============================================================================
+                     FXAA QUALITY - MEDIUM DITHER PRESETS
+============================================================================*/
+#if (FXAA_QUALITY__PRESET == 10)
+    #define FXAA_QUALITY__PS 3
+    #define FXAA_QUALITY__P0 1.5
+    #define FXAA_QUALITY__P1 3.0
+    #define FXAA_QUALITY__P2 12.0
+#endif
+/*--------------------------------------------------------------------------*/
+#if (FXAA_QUALITY__PRESET == 11)
+    #define FXAA_QUALITY__PS 4
+    #define FXAA_QUALITY__P0 1.0
+    #define FXAA_QUALITY__P1 1.5
+    #define FXAA_QUALITY__P2 3.0
+    #define FXAA_QUALITY__P3 12.0
+#endif
+/*--------------------------------------------------------------------------*/
+#if (FXAA_QUALITY__PRESET == 12)
+    #define FXAA_QUALITY__PS 5
+    #define FXAA_QUALITY__P0 1.0
+    #define FXAA_QUALITY__P1 1.5
+    #define FXAA_QUALITY__P2 2.0
+    #define FXAA_QUALITY__P3 4.0
+    #define FXAA_QUALITY__P4 12.0
+#endif
+/*--------------------------------------------------------------------------*/
+#if (FXAA_QUALITY__PRESET == 13)
+    #define FXAA_QUALITY__PS 6
+    #define FXAA_QUALITY__P0 1.0
+    #define FXAA_QUALITY__P1 1.5
+    #define FXAA_QUALITY__P2 2.0
+    #define FXAA_QUALITY__P3 2.0
+    #define FXAA_QUALITY__P4 4.0
+    #define FXAA_QUALITY__P5 12.0
+#endif
+/*--------------------------------------------------------------------------*/
+#if (FXAA_QUALITY__PRESET == 14)
+    #define FXAA_QUALITY__PS 7
+    #define FXAA_QUALITY__P0 1.0
+    #define FXAA_QUALITY__P1 1.5
+    #define FXAA_QUALITY__P2 2.0
+    #define FXAA_QUALITY__P3 2.0
+    #define FXAA_QUALITY__P4 2.0
+    #define FXAA_QUALITY__P5 4.0
+    #define FXAA_QUALITY__P6 12.0
+#endif
+/*--------------------------------------------------------------------------*/
+#if (FXAA_QUALITY__PRESET == 15)
+    #define FXAA_QUALITY__PS 8
+    #define FXAA_QUALITY__P0 1.0
+    #define FXAA_QUALITY__P1 1.5
+    #define FXAA_QUALITY__P2 2.0
+    #define FXAA_QUALITY__P3 2.0
+    #define FXAA_QUALITY__P4 2.0
+    #define FXAA_QUALITY__P5 2.0
+    #define FXAA_QUALITY__P6 4.0
+    #define FXAA_QUALITY__P7 12.0
+#endif
+
+/*============================================================================
+                     FXAA QUALITY - LOW DITHER PRESETS
+============================================================================*/
+#if (FXAA_QUALITY__PRESET == 20)
+    #define FXAA_QUALITY__PS 3
+    #define FXAA_QUALITY__P0 1.5
+    #define FXAA_QUALITY__P1 2.0
+    #define FXAA_QUALITY__P2 8.0
+#endif
+/*--------------------------------------------------------------------------*/
+#if (FXAA_QUALITY__PRESET == 21)
+    #define FXAA_QUALITY__PS 4
+    #define FXAA_QUALITY__P0 1.0
+    #define FXAA_QUALITY__P1 1.5
+    #define FXAA_QUALITY__P2 2.0
+    #define FXAA_QUALITY__P3 8.0
+#endif
+/*--------------------------------------------------------------------------*/
+#if (FXAA_QUALITY__PRESET == 22)
+    #define FXAA_QUALITY__PS 5
+    #define FXAA_QUALITY__P0 1.0
+    #define FXAA_QUALITY__P1 1.5
+    #define FXAA_QUALITY__P2 2.0
+    #define FXAA_QUALITY__P3 2.0
+    #define FXAA_QUALITY__P4 8.0
+#endif
+/*--------------------------------------------------------------------------*/
+#if (FXAA_QUALITY__PRESET == 23)
+    #define FXAA_QUALITY__PS 6
+    #define FXAA_QUALITY__P0 1.0
+    #define FXAA_QUALITY__P1 1.5
+    #define FXAA_QUALITY__P2 2.0
+    #define FXAA_QUALITY__P3 2.0
+    #define FXAA_QUALITY__P4 2.0
+    #define FXAA_QUALITY__P5 8.0
+#endif
+/*--------------------------------------------------------------------------*/
+#if (FXAA_QUALITY__PRESET == 24)
+    #define FXAA_QUALITY__PS 7
+    #define FXAA_QUALITY__P0 1.0
+    #define FXAA_QUALITY__P1 1.5
+    #define FXAA_QUALITY__P2 2.0
+    #define FXAA_QUALITY__P3 2.0
+    #define FXAA_QUALITY__P4 2.0
+    #define FXAA_QUALITY__P5 3.0
+    #define FXAA_QUALITY__P6 8.0
+#endif
+/*--------------------------------------------------------------------------*/
+#if (FXAA_QUALITY__PRESET == 25)
+    #define FXAA_QUALITY__PS 8
+    #define FXAA_QUALITY__P0 1.0
+    #define FXAA_QUALITY__P1 1.5
+    #define FXAA_QUALITY__P2 2.0
+    #define FXAA_QUALITY__P3 2.0
+    #define FXAA_QUALITY__P4 2.0
+    #define FXAA_QUALITY__P5 2.0
+    #define FXAA_QUALITY__P6 4.0
+    #define FXAA_QUALITY__P7 8.0
+#endif
+/*--------------------------------------------------------------------------*/
+#if (FXAA_QUALITY__PRESET == 26)
+    #define FXAA_QUALITY__PS 9
+    #define FXAA_QUALITY__P0 1.0
+    #define FXAA_QUALITY__P1 1.5
+    #define FXAA_QUALITY__P2 2.0
+    #define FXAA_QUALITY__P3 2.0
+    #define FXAA_QUALITY__P4 2.0
+    #define FXAA_QUALITY__P5 2.0
+    #define FXAA_QUALITY__P6 2.0
+    #define FXAA_QUALITY__P7 4.0
+    #define FXAA_QUALITY__P8 8.0
+#endif
+/*--------------------------------------------------------------------------*/
+#if (FXAA_QUALITY__PRESET == 27)
+    #define FXAA_QUALITY__PS 10
+    #define FXAA_QUALITY__P0 1.0
+    #define FXAA_QUALITY__P1 1.5
+    #define FXAA_QUALITY__P2 2.0
+    #define FXAA_QUALITY__P3 2.0
+    #define FXAA_QUALITY__P4 2.0
+    #define FXAA_QUALITY__P5 2.0
+    #define FXAA_QUALITY__P6 2.0
+    #define FXAA_QUALITY__P7 2.0
+    #define FXAA_QUALITY__P8 4.0
+    #define FXAA_QUALITY__P9 8.0
+#endif
+/*--------------------------------------------------------------------------*/
+#if (FXAA_QUALITY__PRESET == 28)
+    #define FXAA_QUALITY__PS 11
+    #define FXAA_QUALITY__P0 1.0
+    #define FXAA_QUALITY__P1 1.5
+    #define FXAA_QUALITY__P2 2.0
+    #define FXAA_QUALITY__P3 2.0
+    #define FXAA_QUALITY__P4 2.0
+    #define FXAA_QUALITY__P5 2.0
+    #define FXAA_QUALITY__P6 2.0
+    #define FXAA_QUALITY__P7 2.0
+    #define FXAA_QUALITY__P8 2.0
+    #define FXAA_QUALITY__P9 4.0
+    #define FXAA_QUALITY__P10 8.0
+#endif
+/*--------------------------------------------------------------------------*/
+#if (FXAA_QUALITY__PRESET == 29)
+    #define FXAA_QUALITY__PS 12
+    #define FXAA_QUALITY__P0 1.0
+    #define FXAA_QUALITY__P1 1.5
+    #define FXAA_QUALITY__P2 2.0
+    #define FXAA_QUALITY__P3 2.0
+    #define FXAA_QUALITY__P4 2.0
+    #define FXAA_QUALITY__P5 2.0
+    #define FXAA_QUALITY__P6 2.0
+    #define FXAA_QUALITY__P7 2.0
+    #define FXAA_QUALITY__P8 2.0
+    #define FXAA_QUALITY__P9 2.0
+    #define FXAA_QUALITY__P10 4.0
+    #define FXAA_QUALITY__P11 8.0
+#endif
+
+/*============================================================================
+                     FXAA QUALITY - EXTREME QUALITY
+============================================================================*/
+#if (FXAA_QUALITY__PRESET == 39)
+    #define FXAA_QUALITY__PS 12
+    #define FXAA_QUALITY__P0 1.0
+    #define FXAA_QUALITY__P1 1.0
+    #define FXAA_QUALITY__P2 1.0
+    #define FXAA_QUALITY__P3 1.0
+    #define FXAA_QUALITY__P4 1.0
+    #define FXAA_QUALITY__P5 1.5
+    #define FXAA_QUALITY__P6 2.0
+    #define FXAA_QUALITY__P7 2.0
+    #define FXAA_QUALITY__P8 2.0
+    #define FXAA_QUALITY__P9 2.0
+    #define FXAA_QUALITY__P10 4.0
+    #define FXAA_QUALITY__P11 8.0
+#endif
+
+
+
+/*============================================================================
+
+                                API PORTING
+
+============================================================================*/
+#if (FXAA_GLSL_120 == 1) || (FXAA_GLSL_130 == 1)
+    #define FxaaBool bool
+    #define FxaaDiscard discard
+    #define FxaaFloat float
+    #define FxaaFloat2 vec2
+    #define FxaaFloat3 vec3
+    #define FxaaFloat4 vec4
+    #define FxaaHalf float
+    #define FxaaHalf2 vec2
+    #define FxaaHalf3 vec3
+    #define FxaaHalf4 vec4
+    #define FxaaInt2 ivec2
+    #define FxaaSat(x) clamp(x, 0.0, 1.0)
+    #define FxaaTex sampler2D
+#else
+    #define FxaaBool bool
+    #define FxaaDiscard clip(-1)
+    #define FxaaFloat float
+    #define FxaaFloat2 float2
+    #define FxaaFloat3 float3
+    #define FxaaFloat4 float4
+    #define FxaaHalf half
+    #define FxaaHalf2 half2
+    #define FxaaHalf3 half3
+    #define FxaaHalf4 half4
+    #define FxaaSat(x) saturate(x)
+#endif
+/*--------------------------------------------------------------------------*/
+#if (FXAA_GLSL_120 == 1)
+    // Requires,
+    //  #version 120
+    // And at least,
+    //  #extension GL_EXT_gpu_shader4 : enable
+    //  (or set FXAA_FAST_PIXEL_OFFSET 1 to work like DX9)
+    #define FxaaTexTop(t, p) texture2DLod(t, p, 0.0)
+    #if (FXAA_FAST_PIXEL_OFFSET == 1)
+        #define FxaaTexOff(t, p, o, r) texture2DLodOffset(t, p, 0.0, o)
+    #else
+        #define FxaaTexOff(t, p, o, r) texture2DLod(t, p + (o * r), 0.0)
+    #endif
+    #if (FXAA_GATHER4_ALPHA == 1)
+        // use #extension GL_ARB_gpu_shader5 : enable
+        #define FxaaTexAlpha4(t, p) textureGather(t, p, 3)
+        #define FxaaTexOffAlpha4(t, p, o) textureGatherOffset(t, p, o, 3)
+        #define FxaaTexGreen4(t, p) textureGather(t, p, 1)
+        #define FxaaTexOffGreen4(t, p, o) textureGatherOffset(t, p, o, 1)
+    #endif
+#endif
+/*--------------------------------------------------------------------------*/
+#if (FXAA_GLSL_130 == 1)
+    // Requires "#version 130" or better
+    #define FxaaTexTop(t, p) textureLod(t, p, 0.0)
+    #define FxaaTexOff(t, p, o, r) textureLodOffset(t, p, 0.0, o)
+    #if (FXAA_GATHER4_ALPHA == 1)
+        // use #extension GL_ARB_gpu_shader5 : enable
+        #define FxaaTexAlpha4(t, p) textureGather(t, p, 3)
+        #define FxaaTexOffAlpha4(t, p, o) textureGatherOffset(t, p, o, 3)
+        #define FxaaTexGreen4(t, p) textureGather(t, p, 1)
+        #define FxaaTexOffGreen4(t, p, o) textureGatherOffset(t, p, o, 1)
+    #endif
+#endif
+
+
+/*============================================================================
+                   GREEN AS LUMA OPTION SUPPORT FUNCTION
+============================================================================*/
+#if (FXAA_GREEN_AS_LUMA == 0)
+    FxaaFloat FxaaLuma(FxaaFloat4 rgba) { return rgba.w; }
+#else
+    FxaaFloat FxaaLuma(FxaaFloat4 rgba) { return rgba.y; }
+#endif    
+
+
+
+
+/*============================================================================
+
+                             FXAA3 QUALITY - PC
+
+============================================================================*/
+#if (FXAA_PC == 1)
+/*--------------------------------------------------------------------------*/
+FxaaFloat4 FxaaPixelShader(
+    //
+    // Use noperspective interpolation here (turn off perspective interpolation).
+    // {xy} = center of pixel
+    FxaaFloat2 pos,
+    //
+    // Used only for FXAA Console, and not used on the 360 version.
+    // Use noperspective interpolation here (turn off perspective interpolation).
+    // {xy__} = upper left of pixel
+    // {__zw} = lower right of pixel
+    FxaaFloat4 fxaaConsolePosPos,
+    //
+    // Input color texture.
+    // {rgb_} = color in linear or perceptual color space
+    // if (FXAA_GREEN_AS_LUMA == 0)
+    //     {___a} = luma in perceptual color space (not linear)
+    FxaaTex tex,
+    //
+    // Only used on the optimized 360 version of FXAA Console.
+    // For everything but 360, just use the same input here as for "tex".
+    // For 360, same texture, just alias with a 2nd sampler.
+    // This sampler needs to have an exponent bias of -1.
+    FxaaTex fxaaConsole360TexExpBiasNegOne,
+    //
+    // Only used on the optimized 360 version of FXAA Console.
+    // For everything but 360, just use the same input here as for "tex".
+    // For 360, same texture, just alias with a 3nd sampler.
+    // This sampler needs to have an exponent bias of -2.
+    FxaaTex fxaaConsole360TexExpBiasNegTwo,
+    //
+    // Only used on FXAA Quality.
+    // This must be from a constant/uniform.
+    // {x_} = 1.0/screenWidthInPixels
+    // {_y} = 1.0/screenHeightInPixels
+    FxaaFloat2 fxaaQualityRcpFrame,
+    //
+    // Only used on FXAA Console.
+    // This must be from a constant/uniform.
+    // This effects sub-pixel AA quality and inversely sharpness.
+    //   Where N ranges between,
+    //     N = 0.50 (default)
+    //     N = 0.33 (sharper)
+    // {x___} = -N/screenWidthInPixels  
+    // {_y__} = -N/screenHeightInPixels
+    // {__z_} =  N/screenWidthInPixels  
+    // {___w} =  N/screenHeightInPixels 
+    FxaaFloat4 fxaaConsoleRcpFrameOpt,
+    //
+    // Only used on FXAA Console.
+    // Not used on 360, but used on PS3 and PC.
+    // This must be from a constant/uniform.
+    // {x___} = -2.0/screenWidthInPixels  
+    // {_y__} = -2.0/screenHeightInPixels
+    // {__z_} =  2.0/screenWidthInPixels  
+    // {___w} =  2.0/screenHeightInPixels 
+    FxaaFloat4 fxaaConsoleRcpFrameOpt2,
+    //
+    // Only used on FXAA Console.
+    // Only used on 360 in place of fxaaConsoleRcpFrameOpt2.
+    // This must be from a constant/uniform.
+    // {x___} =  8.0/screenWidthInPixels  
+    // {_y__} =  8.0/screenHeightInPixels
+    // {__z_} = -4.0/screenWidthInPixels  
+    // {___w} = -4.0/screenHeightInPixels 
+    FxaaFloat4 fxaaConsole360RcpFrameOpt2,
+    //
+    // Only used on FXAA Quality.
+    // This used to be the FXAA_QUALITY__SUBPIX define.
+    // It is here now to allow easier tuning.
+    // Choose the amount of sub-pixel aliasing removal.
+    // This can effect sharpness.
+    //   1.00 - upper limit (softer)
+    //   0.75 - default amount of filtering
+    //   0.50 - lower limit (sharper, less sub-pixel aliasing removal)
+    //   0.25 - almost off
+    //   0.00 - completely off
+    FxaaFloat fxaaQualitySubpix,
+    //
+    // Only used on FXAA Quality.
+    // This used to be the FXAA_QUALITY__EDGE_THRESHOLD define.
+    // It is here now to allow easier tuning.
+    // The minimum amount of local contrast required to apply algorithm.
+    //   0.333 - too little (faster)
+    //   0.250 - low quality
+    //   0.166 - default
+    //   0.125 - high quality 
+    //   0.063 - overkill (slower)
+    FxaaFloat fxaaQualityEdgeThreshold,
+    //
+    // Only used on FXAA Quality.
+    // This used to be the FXAA_QUALITY__EDGE_THRESHOLD_MIN define.
+    // It is here now to allow easier tuning.
+    // Trims the algorithm from processing darks.
+    //   0.0833 - upper limit (default, the start of visible unfiltered edges)
+    //   0.0625 - high quality (faster)
+    //   0.0312 - visible limit (slower)
+    // Special notes when using FXAA_GREEN_AS_LUMA,
+    //   Likely want to set this to zero.
+    //   As colors that are mostly not-green
+    //   will appear very dark in the green channel!
+    //   Tune by looking at mostly non-green content,
+    //   then start at zero and increase until aliasing is a problem.
+    FxaaFloat fxaaQualityEdgeThresholdMin,
+    // 
+    // Only used on FXAA Console.
+    // This used to be the FXAA_CONSOLE__EDGE_SHARPNESS define.
+    // It is here now to allow easier tuning.
+    // This does not effect PS3, as this needs to be compiled in.
+    //   Use FXAA_CONSOLE__PS3_EDGE_SHARPNESS for PS3.
+    //   Due to the PS3 being ALU bound,
+    //   there are only three safe values here: 2 and 4 and 8.
+    //   These options use the shaders ability to a free *|/ by 2|4|8.
+    // For all other platforms can be a non-power of two.
+    //   8.0 is sharper (default!!!)
+    //   4.0 is softer
+    //   2.0 is really soft (good only for vector graphics inputs)
+    FxaaFloat fxaaConsoleEdgeSharpness,
+    //
+    // Only used on FXAA Console.
+    // This used to be the FXAA_CONSOLE__EDGE_THRESHOLD define.
+    // It is here now to allow easier tuning.
+    // This does not effect PS3, as this needs to be compiled in.
+    //   Use FXAA_CONSOLE__PS3_EDGE_THRESHOLD for PS3.
+    //   Due to the PS3 being ALU bound,
+    //   there are only two safe values here: 1/4 and 1/8.
+    //   These options use the shaders ability to a free *|/ by 2|4|8.
+    // The console setting has a different mapping than the quality setting.
+    // Other platforms can use other values.
+    //   0.125 leaves less aliasing, but is softer (default!!!)
+    //   0.25 leaves more aliasing, and is sharper
+    FxaaFloat fxaaConsoleEdgeThreshold,
+    //
+    // Only used on FXAA Console.
+    // This used to be the FXAA_CONSOLE__EDGE_THRESHOLD_MIN define.
+    // It is here now to allow easier tuning.
+    // Trims the algorithm from processing darks.
+    // The console setting has a different mapping than the quality setting.
+    // This only applies when FXAA_EARLY_EXIT is 1.
+    // This does not apply to PS3, 
+    // PS3 was simplified to avoid more shader instructions.
+    //   0.06 - faster but more aliasing in darks
+    //   0.05 - default
+    //   0.04 - slower and less aliasing in darks
+    // Special notes when using FXAA_GREEN_AS_LUMA,
+    //   Likely want to set this to zero.
+    //   As colors that are mostly not-green
+    //   will appear very dark in the green channel!
+    //   Tune by looking at mostly non-green content,
+    //   then start at zero and increase until aliasing is a problem.
+    FxaaFloat fxaaConsoleEdgeThresholdMin,
+    //    
+    // Extra constants for 360 FXAA Console only.
+    // Use zeros or anything else for other platforms.
+    // These must be in physical constant registers and NOT immedates.
+    // Immedates will result in compiler un-optimizing.
+    // {xyzw} = float4(1.0, -1.0, 0.25, -0.25)
+    FxaaFloat4 fxaaConsole360ConstDir
+) {
+/*--------------------------------------------------------------------------*/
+    FxaaFloat2 posM;
+    posM.x = pos.x;
+    posM.y = pos.y;
+    #if (FXAA_GATHER4_ALPHA == 1)
+        #if (FXAA_DISCARD == 0)
+            FxaaFloat4 rgbyM = FxaaTexTop(tex, posM);
+            #if (FXAA_GREEN_AS_LUMA == 0)
+                #define lumaM rgbyM.w
+            #else
+                #define lumaM rgbyM.y
+            #endif
+        #endif
+        #if (FXAA_GREEN_AS_LUMA == 0)
+            FxaaFloat4 luma4A = FxaaTexAlpha4(tex, posM);
+            FxaaFloat4 luma4B = FxaaTexOffAlpha4(tex, posM, FxaaInt2(-1, -1));
+        #else
+            FxaaFloat4 luma4A = FxaaTexGreen4(tex, posM);
+            FxaaFloat4 luma4B = FxaaTexOffGreen4(tex, posM, FxaaInt2(-1, -1));
+        #endif
+        #if (FXAA_DISCARD == 1)
+            #define lumaM luma4A.w
+        #endif
+        #define lumaE luma4A.z
+        #define lumaS luma4A.x
+        #define lumaSE luma4A.y
+        #define lumaNW luma4B.w
+        #define lumaN luma4B.z
+        #define lumaW luma4B.x
+    #else
+        FxaaFloat4 rgbyM = FxaaTexTop(tex, posM);
+        #if (FXAA_GREEN_AS_LUMA == 0)
+            #define lumaM rgbyM.w
+        #else
+            #define lumaM rgbyM.y
+        #endif
+        FxaaFloat lumaS = FxaaLuma(FxaaTexOff(tex, posM, FxaaInt2( 0, 1), fxaaQualityRcpFrame.xy));
+        FxaaFloat lumaE = FxaaLuma(FxaaTexOff(tex, posM, FxaaInt2( 1, 0), fxaaQualityRcpFrame.xy));
+        FxaaFloat lumaN = FxaaLuma(FxaaTexOff(tex, posM, FxaaInt2( 0,-1), fxaaQualityRcpFrame.xy));
+        FxaaFloat lumaW = FxaaLuma(FxaaTexOff(tex, posM, FxaaInt2(-1, 0), fxaaQualityRcpFrame.xy));
+    #endif
+/*--------------------------------------------------------------------------*/
+    FxaaFloat maxSM = max(lumaS, lumaM);
+    FxaaFloat minSM = min(lumaS, lumaM);
+    FxaaFloat maxESM = max(lumaE, maxSM);
+    FxaaFloat minESM = min(lumaE, minSM);
+    FxaaFloat maxWN = max(lumaN, lumaW);
+    FxaaFloat minWN = min(lumaN, lumaW);
+    FxaaFloat rangeMax = max(maxWN, maxESM);
+    FxaaFloat rangeMin = min(minWN, minESM);
+    FxaaFloat rangeMaxScaled = rangeMax * fxaaQualityEdgeThreshold;
+    FxaaFloat range = rangeMax - rangeMin;
+    FxaaFloat rangeMaxClamped = max(fxaaQualityEdgeThresholdMin, rangeMaxScaled);
+    FxaaBool earlyExit = range < rangeMaxClamped;
+/*--------------------------------------------------------------------------*/
+    if(earlyExit)
+        #if (FXAA_DISCARD == 1)
+            FxaaDiscard;
+        #else
+            return rgbyM;
+        #endif
+/*--------------------------------------------------------------------------*/
+    #if (FXAA_GATHER4_ALPHA == 0)
+        FxaaFloat lumaNW = FxaaLuma(FxaaTexOff(tex, posM, FxaaInt2(-1,-1), fxaaQualityRcpFrame.xy));
+        FxaaFloat lumaSE = FxaaLuma(FxaaTexOff(tex, posM, FxaaInt2( 1, 1), fxaaQualityRcpFrame.xy));
+        FxaaFloat lumaNE = FxaaLuma(FxaaTexOff(tex, posM, FxaaInt2( 1,-1), fxaaQualityRcpFrame.xy));
+        FxaaFloat lumaSW = FxaaLuma(FxaaTexOff(tex, posM, FxaaInt2(-1, 1), fxaaQualityRcpFrame.xy));
+    #else
+        FxaaFloat lumaNE = FxaaLuma(FxaaTexOff(tex, posM, FxaaInt2(1, -1), fxaaQualityRcpFrame.xy));
+        FxaaFloat lumaSW = FxaaLuma(FxaaTexOff(tex, posM, FxaaInt2(-1, 1), fxaaQualityRcpFrame.xy));
+    #endif
+/*--------------------------------------------------------------------------*/
+    FxaaFloat lumaNS = lumaN + lumaS;
+    FxaaFloat lumaWE = lumaW + lumaE;
+    FxaaFloat subpixRcpRange = 1.0/range;
+    FxaaFloat subpixNSWE = lumaNS + lumaWE;
+    FxaaFloat edgeHorz1 = (-2.0 * lumaM) + lumaNS;
+    FxaaFloat edgeVert1 = (-2.0 * lumaM) + lumaWE;
+/*--------------------------------------------------------------------------*/
+    FxaaFloat lumaNESE = lumaNE + lumaSE;
+    FxaaFloat lumaNWNE = lumaNW + lumaNE;
+    FxaaFloat edgeHorz2 = (-2.0 * lumaE) + lumaNESE;
+    FxaaFloat edgeVert2 = (-2.0 * lumaN) + lumaNWNE;
+/*--------------------------------------------------------------------------*/
+    FxaaFloat lumaNWSW = lumaNW + lumaSW;
+    FxaaFloat lumaSWSE = lumaSW + lumaSE;
+    FxaaFloat edgeHorz4 = (abs(edgeHorz1) * 2.0) + abs(edgeHorz2);
+    FxaaFloat edgeVert4 = (abs(edgeVert1) * 2.0) + abs(edgeVert2);
+    FxaaFloat edgeHorz3 = (-2.0 * lumaW) + lumaNWSW;
+    FxaaFloat edgeVert3 = (-2.0 * lumaS) + lumaSWSE;
+    FxaaFloat edgeHorz = abs(edgeHorz3) + edgeHorz4;
+    FxaaFloat edgeVert = abs(edgeVert3) + edgeVert4;
+/*--------------------------------------------------------------------------*/
+    FxaaFloat subpixNWSWNESE = lumaNWSW + lumaNESE;
+    FxaaFloat lengthSign = fxaaQualityRcpFrame.x;
+    FxaaBool horzSpan = edgeHorz >= edgeVert;
+    FxaaFloat subpixA = subpixNSWE * 2.0 + subpixNWSWNESE;
+/*--------------------------------------------------------------------------*/
+    if(!horzSpan) lumaN = lumaW;
+    if(!horzSpan) lumaS = lumaE;
+    if(horzSpan) lengthSign = fxaaQualityRcpFrame.y;
+    FxaaFloat subpixB = (subpixA * (1.0/12.0)) - lumaM;
+/*--------------------------------------------------------------------------*/
+    FxaaFloat gradientN = lumaN - lumaM;
+    FxaaFloat gradientS = lumaS - lumaM;
+    FxaaFloat lumaNN = lumaN + lumaM;
+    FxaaFloat lumaSS = lumaS + lumaM;
+    FxaaBool pairN = abs(gradientN) >= abs(gradientS);
+    FxaaFloat gradient = max(abs(gradientN), abs(gradientS));
+    if(pairN) lengthSign = -lengthSign;
+    FxaaFloat subpixC = FxaaSat(abs(subpixB) * subpixRcpRange);
+/*--------------------------------------------------------------------------*/
+    FxaaFloat2 posB;
+    posB.x = posM.x;
+    posB.y = posM.y;
+    FxaaFloat2 offNP;
+    offNP.x = (!horzSpan) ? 0.0 : fxaaQualityRcpFrame.x;
+    offNP.y = ( horzSpan) ? 0.0 : fxaaQualityRcpFrame.y;
+    if(!horzSpan) posB.x += lengthSign * 0.5;
+    if( horzSpan) posB.y += lengthSign * 0.5;
+/*--------------------------------------------------------------------------*/
+    FxaaFloat2 posN;
+    posN.x = posB.x - offNP.x * FXAA_QUALITY__P0;
+    posN.y = posB.y - offNP.y * FXAA_QUALITY__P0;
+    FxaaFloat2 posP;
+    posP.x = posB.x + offNP.x * FXAA_QUALITY__P0;
+    posP.y = posB.y + offNP.y * FXAA_QUALITY__P0;
+    FxaaFloat subpixD = ((-2.0)*subpixC) + 3.0;
+    FxaaFloat lumaEndN = FxaaLuma(FxaaTexTop(tex, posN));
+    FxaaFloat subpixE = subpixC * subpixC;
+    FxaaFloat lumaEndP = FxaaLuma(FxaaTexTop(tex, posP));
+/*--------------------------------------------------------------------------*/
+    if(!pairN) lumaNN = lumaSS;
+    FxaaFloat gradientScaled = gradient * 1.0/4.0;
+    FxaaFloat lumaMM = lumaM - lumaNN * 0.5;
+    FxaaFloat subpixF = subpixD * subpixE;
+    FxaaBool lumaMLTZero = lumaMM < 0.0;
+/*--------------------------------------------------------------------------*/
+    lumaEndN -= lumaNN * 0.5;
+    lumaEndP -= lumaNN * 0.5;
+    FxaaBool doneN = abs(lumaEndN) >= gradientScaled;
+    FxaaBool doneP = abs(lumaEndP) >= gradientScaled;
+    if(!doneN) posN.x -= offNP.x * FXAA_QUALITY__P1;
+    if(!doneN) posN.y -= offNP.y * FXAA_QUALITY__P1;
+    FxaaBool doneNP = (!doneN) || (!doneP);
+    if(!doneP) posP.x += offNP.x * FXAA_QUALITY__P1;
+    if(!doneP) posP.y += offNP.y * FXAA_QUALITY__P1;
+/*--------------------------------------------------------------------------*/
+    if(doneNP) {
+        if(!doneN) lumaEndN = FxaaLuma(FxaaTexTop(tex, posN.xy));
+        if(!doneP) lumaEndP = FxaaLuma(FxaaTexTop(tex, posP.xy));
+        if(!doneN) lumaEndN = lumaEndN - lumaNN * 0.5;
+        if(!doneP) lumaEndP = lumaEndP - lumaNN * 0.5;
+        doneN = abs(lumaEndN) >= gradientScaled;
+        doneP = abs(lumaEndP) >= gradientScaled;
+        if(!doneN) posN.x -= offNP.x * FXAA_QUALITY__P2;
+        if(!doneN) posN.y -= offNP.y * FXAA_QUALITY__P2;
+        doneNP = (!doneN) || (!doneP);
+        if(!doneP) posP.x += offNP.x * FXAA_QUALITY__P2;
+        if(!doneP) posP.y += offNP.y * FXAA_QUALITY__P2;
+/*--------------------------------------------------------------------------*/
+        #if (FXAA_QUALITY__PS > 3)
+        if(doneNP) {
+            if(!doneN) lumaEndN = FxaaLuma(FxaaTexTop(tex, posN.xy));
+            if(!doneP) lumaEndP = FxaaLuma(FxaaTexTop(tex, posP.xy));
+            if(!doneN) lumaEndN = lumaEndN - lumaNN * 0.5;
+            if(!doneP) lumaEndP = lumaEndP - lumaNN * 0.5;
+            doneN = abs(lumaEndN) >= gradientScaled;
+            doneP = abs(lumaEndP) >= gradientScaled;
+            if(!doneN) posN.x -= offNP.x * FXAA_QUALITY__P3;
+            if(!doneN) posN.y -= offNP.y * FXAA_QUALITY__P3;
+            doneNP = (!doneN) || (!doneP);
+            if(!doneP) posP.x += offNP.x * FXAA_QUALITY__P3;
+            if(!doneP) posP.y += offNP.y * FXAA_QUALITY__P3;
+/*--------------------------------------------------------------------------*/
+            #if (FXAA_QUALITY__PS > 4)
+            if(doneNP) {
+                if(!doneN) lumaEndN = FxaaLuma(FxaaTexTop(tex, posN.xy));
+                if(!doneP) lumaEndP = FxaaLuma(FxaaTexTop(tex, posP.xy));
+                if(!doneN) lumaEndN = lumaEndN - lumaNN * 0.5;
+                if(!doneP) lumaEndP = lumaEndP - lumaNN * 0.5;
+                doneN = abs(lumaEndN) >= gradientScaled;
+                doneP = abs(lumaEndP) >= gradientScaled;
+                if(!doneN) posN.x -= offNP.x * FXAA_QUALITY__P4;
+                if(!doneN) posN.y -= offNP.y * FXAA_QUALITY__P4;
+                doneNP = (!doneN) || (!doneP);
+                if(!doneP) posP.x += offNP.x * FXAA_QUALITY__P4;
+                if(!doneP) posP.y += offNP.y * FXAA_QUALITY__P4;
+/*--------------------------------------------------------------------------*/
+                #if (FXAA_QUALITY__PS > 5)
+                if(doneNP) {
+                    if(!doneN) lumaEndN = FxaaLuma(FxaaTexTop(tex, posN.xy));
+                    if(!doneP) lumaEndP = FxaaLuma(FxaaTexTop(tex, posP.xy));
+                    if(!doneN) lumaEndN = lumaEndN - lumaNN * 0.5;
+                    if(!doneP) lumaEndP = lumaEndP - lumaNN * 0.5;
+                    doneN = abs(lumaEndN) >= gradientScaled;
+                    doneP = abs(lumaEndP) >= gradientScaled;
+                    if(!doneN) posN.x -= offNP.x * FXAA_QUALITY__P5;
+                    if(!doneN) posN.y -= offNP.y * FXAA_QUALITY__P5;
+                    doneNP = (!doneN) || (!doneP);
+                    if(!doneP) posP.x += offNP.x * FXAA_QUALITY__P5;
+                    if(!doneP) posP.y += offNP.y * FXAA_QUALITY__P5;
+/*--------------------------------------------------------------------------*/
+                    #if (FXAA_QUALITY__PS > 6)
+                    if(doneNP) {
+                        if(!doneN) lumaEndN = FxaaLuma(FxaaTexTop(tex, posN.xy));
+                        if(!doneP) lumaEndP = FxaaLuma(FxaaTexTop(tex, posP.xy));
+                        if(!doneN) lumaEndN = lumaEndN - lumaNN * 0.5;
+                        if(!doneP) lumaEndP = lumaEndP - lumaNN * 0.5;
+                        doneN = abs(lumaEndN) >= gradientScaled;
+                        doneP = abs(lumaEndP) >= gradientScaled;
+                        if(!doneN) posN.x -= offNP.x * FXAA_QUALITY__P6;
+                        if(!doneN) posN.y -= offNP.y * FXAA_QUALITY__P6;
+                        doneNP = (!doneN) || (!doneP);
+                        if(!doneP) posP.x += offNP.x * FXAA_QUALITY__P6;
+                        if(!doneP) posP.y += offNP.y * FXAA_QUALITY__P6;
+/*--------------------------------------------------------------------------*/
+                        #if (FXAA_QUALITY__PS > 7)
+                        if(doneNP) {
+                            if(!doneN) lumaEndN = FxaaLuma(FxaaTexTop(tex, posN.xy));
+                            if(!doneP) lumaEndP = FxaaLuma(FxaaTexTop(tex, posP.xy));
+                            if(!doneN) lumaEndN = lumaEndN - lumaNN * 0.5;
+                            if(!doneP) lumaEndP = lumaEndP - lumaNN * 0.5;
+                            doneN = abs(lumaEndN) >= gradientScaled;
+                            doneP = abs(lumaEndP) >= gradientScaled;
+                            if(!doneN) posN.x -= offNP.x * FXAA_QUALITY__P7;
+                            if(!doneN) posN.y -= offNP.y * FXAA_QUALITY__P7;
+                            doneNP = (!doneN) || (!doneP);
+                            if(!doneP) posP.x += offNP.x * FXAA_QUALITY__P7;
+                            if(!doneP) posP.y += offNP.y * FXAA_QUALITY__P7;
+/*--------------------------------------------------------------------------*/
+    #if (FXAA_QUALITY__PS > 8)
+    if(doneNP) {
+        if(!doneN) lumaEndN = FxaaLuma(FxaaTexTop(tex, posN.xy));
+        if(!doneP) lumaEndP = FxaaLuma(FxaaTexTop(tex, posP.xy));
+        if(!doneN) lumaEndN = lumaEndN - lumaNN * 0.5;
+        if(!doneP) lumaEndP = lumaEndP - lumaNN * 0.5;
+        doneN = abs(lumaEndN) >= gradientScaled;
+        doneP = abs(lumaEndP) >= gradientScaled;
+        if(!doneN) posN.x -= offNP.x * FXAA_QUALITY__P8;
+        if(!doneN) posN.y -= offNP.y * FXAA_QUALITY__P8;
+        doneNP = (!doneN) || (!doneP);
+        if(!doneP) posP.x += offNP.x * FXAA_QUALITY__P8;
+        if(!doneP) posP.y += offNP.y * FXAA_QUALITY__P8;
+/*--------------------------------------------------------------------------*/
+        #if (FXAA_QUALITY__PS > 9)
+        if(doneNP) {
+            if(!doneN) lumaEndN = FxaaLuma(FxaaTexTop(tex, posN.xy));
+            if(!doneP) lumaEndP = FxaaLuma(FxaaTexTop(tex, posP.xy));
+            if(!doneN) lumaEndN = lumaEndN - lumaNN * 0.5;
+            if(!doneP) lumaEndP = lumaEndP - lumaNN * 0.5;
+            doneN = abs(lumaEndN) >= gradientScaled;
+            doneP = abs(lumaEndP) >= gradientScaled;
+            if(!doneN) posN.x -= offNP.x * FXAA_QUALITY__P9;
+            if(!doneN) posN.y -= offNP.y * FXAA_QUALITY__P9;
+            doneNP = (!doneN) || (!doneP);
+            if(!doneP) posP.x += offNP.x * FXAA_QUALITY__P9;
+            if(!doneP) posP.y += offNP.y * FXAA_QUALITY__P9;
+/*--------------------------------------------------------------------------*/
+            #if (FXAA_QUALITY__PS > 10)
+            if(doneNP) {
+                if(!doneN) lumaEndN = FxaaLuma(FxaaTexTop(tex, posN.xy));
+                if(!doneP) lumaEndP = FxaaLuma(FxaaTexTop(tex, posP.xy));
+                if(!doneN) lumaEndN = lumaEndN - lumaNN * 0.5;
+                if(!doneP) lumaEndP = lumaEndP - lumaNN * 0.5;
+                doneN = abs(lumaEndN) >= gradientScaled;
+                doneP = abs(lumaEndP) >= gradientScaled;
+                if(!doneN) posN.x -= offNP.x * FXAA_QUALITY__P10;
+                if(!doneN) posN.y -= offNP.y * FXAA_QUALITY__P10;
+                doneNP = (!doneN) || (!doneP);
+                if(!doneP) posP.x += offNP.x * FXAA_QUALITY__P10;
+                if(!doneP) posP.y += offNP.y * FXAA_QUALITY__P10;
+/*--------------------------------------------------------------------------*/
+                #if (FXAA_QUALITY__PS > 11)
+                if(doneNP) {
+                    if(!doneN) lumaEndN = FxaaLuma(FxaaTexTop(tex, posN.xy));
+                    if(!doneP) lumaEndP = FxaaLuma(FxaaTexTop(tex, posP.xy));
+                    if(!doneN) lumaEndN = lumaEndN - lumaNN * 0.5;
+                    if(!doneP) lumaEndP = lumaEndP - lumaNN * 0.5;
+                    doneN = abs(lumaEndN) >= gradientScaled;
+                    doneP = abs(lumaEndP) >= gradientScaled;
+                    if(!doneN) posN.x -= offNP.x * FXAA_QUALITY__P11;
+                    if(!doneN) posN.y -= offNP.y * FXAA_QUALITY__P11;
+                    doneNP = (!doneN) || (!doneP);
+                    if(!doneP) posP.x += offNP.x * FXAA_QUALITY__P11;
+                    if(!doneP) posP.y += offNP.y * FXAA_QUALITY__P11;
+/*--------------------------------------------------------------------------*/
+                    #if (FXAA_QUALITY__PS > 12)
+                    if(doneNP) {
+                        if(!doneN) lumaEndN = FxaaLuma(FxaaTexTop(tex, posN.xy));
+                        if(!doneP) lumaEndP = FxaaLuma(FxaaTexTop(tex, posP.xy));
+                        if(!doneN) lumaEndN = lumaEndN - lumaNN * 0.5;
+                        if(!doneP) lumaEndP = lumaEndP - lumaNN * 0.5;
+                        doneN = abs(lumaEndN) >= gradientScaled;
+                        doneP = abs(lumaEndP) >= gradientScaled;
+                        if(!doneN) posN.x -= offNP.x * FXAA_QUALITY__P12;
+                        if(!doneN) posN.y -= offNP.y * FXAA_QUALITY__P12;
+                        doneNP = (!doneN) || (!doneP);
+                        if(!doneP) posP.x += offNP.x * FXAA_QUALITY__P12;
+                        if(!doneP) posP.y += offNP.y * FXAA_QUALITY__P12;
+/*--------------------------------------------------------------------------*/
+                    }
+                    #endif
+/*--------------------------------------------------------------------------*/
+                }
+                #endif
+/*--------------------------------------------------------------------------*/
+            }
+            #endif
+/*--------------------------------------------------------------------------*/
+        }
+        #endif
+/*--------------------------------------------------------------------------*/
+    }
+    #endif
+/*--------------------------------------------------------------------------*/
+                        }
+                        #endif
+/*--------------------------------------------------------------------------*/
+                    }
+                    #endif
+/*--------------------------------------------------------------------------*/
+                }
+                #endif
+/*--------------------------------------------------------------------------*/
+            }
+            #endif
+/*--------------------------------------------------------------------------*/
+        }
+        #endif
+/*--------------------------------------------------------------------------*/
+    }
+/*--------------------------------------------------------------------------*/
+    FxaaFloat dstN = posM.x - posN.x;
+    FxaaFloat dstP = posP.x - posM.x;
+    if(!horzSpan) dstN = posM.y - posN.y;
+    if(!horzSpan) dstP = posP.y - posM.y;
+/*--------------------------------------------------------------------------*/
+    FxaaBool goodSpanN = (lumaEndN < 0.0) != lumaMLTZero;
+    FxaaFloat spanLength = (dstP + dstN);
+    FxaaBool goodSpanP = (lumaEndP < 0.0) != lumaMLTZero;
+    FxaaFloat spanLengthRcp = 1.0/spanLength;
+/*--------------------------------------------------------------------------*/
+    FxaaBool directionN = dstN < dstP;
+    FxaaFloat dst = min(dstN, dstP);
+    FxaaBool goodSpan = directionN ? goodSpanN : goodSpanP;
+    FxaaFloat subpixG = subpixF * subpixF;
+    FxaaFloat pixelOffset = (dst * (-spanLengthRcp)) + 0.5;
+    FxaaFloat subpixH = subpixG * fxaaQualitySubpix;
+/*--------------------------------------------------------------------------*/
+    FxaaFloat pixelOffsetGood = goodSpan ? pixelOffset : 0.0;
+    FxaaFloat pixelOffsetSubpix = max(pixelOffsetGood, subpixH);
+    if(!horzSpan) posM.x += pixelOffsetSubpix * lengthSign;
+    if( horzSpan) posM.y += pixelOffsetSubpix * lengthSign;
+    #if (FXAA_DISCARD == 1)
+        return FxaaTexTop(tex, posM);
+    #else
+        return FxaaFloat4(FxaaTexTop(tex, posM).xyz, lumaM);
+    #endif
+}
+/*==========================================================================*/
+#endif
+
+vec4 mainImage(vec2 fragCoord)
+{
+    vec2 rcpFrame = 1./invResolution.xy;
+  	vec2 uv2 = fragCoord.xy / invResolution.xy;
+
+    float fxaaQualitySubpix = 0.75;  // [0..1], default 0.75
+    float fxaaQualityEdgeThreshold = 0.166;  // [0.125..0.33], default 0.166
+    float fxaaQualityEdgeThresholdMin = 0.02;//0.0625; // ?
+    vec4 dummy4 =  vec4(0.0,0.0,0.0,0.0);
+    float dummy1 = 0.0;
+
+    vec4 col = FxaaPixelShader(uv2, dummy4,
+                                    inputTexture, inputTexture, inputTexture,
+                                    rcpFrame, dummy4, dummy4, dummy4,
+                                    fxaaQualitySubpix, fxaaQualityEdgeThreshold,
+                                    fxaaQualityEdgeThresholdMin,
+                                    dummy1, dummy1, dummy1, dummy4);
+
+    vec4 fragColor = vec4( col.xyz, 1. );
+
+    return fragColor;
+}
+
+void main()
+{
+    ivec2 loc = ivec2(gl_GlobalInvocationID.x * 4, gl_GlobalInvocationID.y * 4);
+    for(int i = 0; i < 4; i++)
+    {
+        for(int j = 0; j < 4; j++)
+        {
+            ivec2 texelCoord = ivec2(loc.x + i, loc.y + j);
+            vec4 outColor = mainImage(texelCoord + vec2(0.5));
+            imageStore(imgOutput, texelCoord, outColor);
+        }
+    }
+}
diff --git a/src/Ryujinx.Graphics.OpenGL/Effects/Shaders/smaa.hlsl b/src/Ryujinx.Graphics.OpenGL/Effects/Shaders/smaa.hlsl
new file mode 100644
index 00000000..2201f78c
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Effects/Shaders/smaa.hlsl
@@ -0,0 +1,1361 @@
+/**
+ * Copyright (C) 2013 Jorge Jimenez (jorge@iryoku.com)
+ * Copyright (C) 2013 Jose I. Echevarria (joseignacioechevarria@gmail.com)
+ * Copyright (C) 2013 Belen Masia (bmasia@unizar.es)
+ * Copyright (C) 2013 Fernando Navarro (fernandn@microsoft.com)
+ * Copyright (C) 2013 Diego Gutierrez (diegog@unizar.es)
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * this software and associated documentation files (the "Software"), to deal in
+ * the Software without restriction, including without limitation the rights to
+ * use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies
+ * of the Software, and to permit persons to whom the Software is furnished to
+ * do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software. As clarification, there
+ * is no requirement that the copyright notice and permission be included in
+ * binary distributions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+
+/**
+ *                  _______  ___  ___       ___           ___
+ *                 /       ||   \/   |     /   \         /   \
+ *                |   (---- |  \  /  |    /  ^  \       /  ^  \
+ *                 \   \    |  |\/|  |   /  /_\  \     /  /_\  \
+ *              ----)   |   |  |  |  |  /  _____  \   /  _____  \
+ *             |_______/    |__|  |__| /__/     \__\ /__/     \__\
+ * 
+ *                               E N H A N C E D
+ *       S U B P I X E L   M O R P H O L O G I C A L   A N T I A L I A S I N G
+ *
+ *                         http://www.iryoku.com/smaa/
+ *
+ * Hi, welcome aboard!
+ * 
+ * Here you'll find instructions to get the shader up and running as fast as
+ * possible.
+ *
+ * IMPORTANTE NOTICE: when updating, remember to update both this file and the
+ * precomputed textures! They may change from version to version.
+ *
+ * The shader has three passes, chained together as follows:
+ *
+ *                           |input|------------------�
+ *                              v                     |
+ *                    [ SMAA*EdgeDetection ]          |
+ *                              v                     |
+ *                          |edgesTex|                |
+ *                              v                     |
+ *              [ SMAABlendingWeightCalculation ]     |
+ *                              v                     |
+ *                          |blendTex|                |
+ *                              v                     |
+ *                [ SMAANeighborhoodBlending ] <------�
+ *                              v
+ *                           |output|
+ *
+ * Note that each [pass] has its own vertex and pixel shader. Remember to use
+ * oversized triangles instead of quads to avoid overshading along the
+ * diagonal.
+ *
+ * You've three edge detection methods to choose from: luma, color or depth.
+ * They represent different quality/performance and anti-aliasing/sharpness
+ * tradeoffs, so our recommendation is for you to choose the one that best
+ * suits your particular scenario:
+ *
+ * - Depth edge detection is usually the fastest but it may miss some edges.
+ *
+ * - Luma edge detection is usually more expensive than depth edge detection,
+ *   but catches visible edges that depth edge detection can miss.
+ *
+ * - Color edge detection is usually the most expensive one but catches
+ *   chroma-only edges.
+ *
+ * For quickstarters: just use luma edge detection.
+ *
+ * The general advice is to not rush the integration process and ensure each
+ * step is done correctly (don't try to integrate SMAA T2x with predicated edge
+ * detection from the start!). Ok then, let's go!
+ *
+ *  1. The first step is to create two RGBA temporal render targets for holding
+ *     |edgesTex| and |blendTex|.
+ *
+ *     In DX10 or DX11, you can use a RG render target for the edges texture.
+ *     In the case of NVIDIA GPUs, using RG render targets seems to actually be
+ *     slower.
+ *
+ *     On the Xbox 360, you can use the same render target for resolving both
+ *     |edgesTex| and |blendTex|, as they aren't needed simultaneously.
+ *
+ *  2. Both temporal render targets |edgesTex| and |blendTex| must be cleared
+ *     each frame. Do not forget to clear the alpha channel!
+ *
+ *  3. The next step is loading the two supporting precalculated textures,
+ *     'areaTex' and 'searchTex'. You'll find them in the 'Textures' folder as
+ *     C++ headers, and also as regular DDS files. They'll be needed for the
+ *     'SMAABlendingWeightCalculation' pass.
+ *
+ *     If you use the C++ headers, be sure to load them in the format specified
+ *     inside of them.
+ *
+ *     You can also compress 'areaTex' and 'searchTex' using BC5 and BC4
+ *     respectively, if you have that option in your content processor pipeline.
+ *     When compressing then, you get a non-perceptible quality decrease, and a
+ *     marginal performance increase.
+ *
+ *  4. All samplers must be set to linear filtering and clamp.
+ *
+ *     After you get the technique working, remember that 64-bit inputs have
+ *     half-rate linear filtering on GCN.
+ *
+ *     If SMAA is applied to 64-bit color buffers, switching to point filtering
+ *     when accesing them will increase the performance. Search for
+ *     'SMAASamplePoint' to see which textures may benefit from point
+ *     filtering, and where (which is basically the color input in the edge
+ *     detection and resolve passes).
+ *
+ *  5. All texture reads and buffer writes must be non-sRGB, with the exception
+ *     of the input read and the output write in
+ *     'SMAANeighborhoodBlending' (and only in this pass!). If sRGB reads in
+ *     this last pass are not possible, the technique will work anyway, but
+ *     will perform antialiasing in gamma space.
+ *
+ *     IMPORTANT: for best results the input read for the color/luma edge 
+ *     detection should *NOT* be sRGB.
+ *
+ *  6. Before including SMAA.h you'll have to setup the render target metrics,
+ *     the target and any optional configuration defines. Optionally you can
+ *     use a preset.
+ *
+ *     You have the following targets available: 
+ *         SMAA_HLSL_3
+ *         SMAA_HLSL_4
+ *         SMAA_HLSL_4_1
+ *         SMAA_GLSL_3 *
+ *         SMAA_GLSL_4 *
+ *
+ *         * (See SMAA_INCLUDE_VS and SMAA_INCLUDE_PS below).
+ *
+ *     And four presets:
+ *         SMAA_PRESET_LOW          (%60 of the quality)
+ *         SMAA_PRESET_MEDIUM       (%80 of the quality)
+ *         SMAA_PRESET_HIGH         (%95 of the quality)
+ *         SMAA_PRESET_ULTRA        (%99 of the quality)
+ *
+ *     For example:
+ *         #define SMAA_RT_METRICS float4(1.0 / 1280.0, 1.0 / 720.0, 1280.0, 720.0)
+ *         #define SMAA_HLSL_4
+ *         #define SMAA_PRESET_HIGH
+ *         #include "SMAA.h"
+ *
+ *     Note that SMAA_RT_METRICS doesn't need to be a macro, it can be a
+ *     uniform variable. The code is designed to minimize the impact of not
+ *     using a constant value, but it is still better to hardcode it.
+ *
+ *     Depending on how you encoded 'areaTex' and 'searchTex', you may have to
+ *     add (and customize) the following defines before including SMAA.h:
+ *          #define SMAA_AREATEX_SELECT(sample) sample.rg
+ *          #define SMAA_SEARCHTEX_SELECT(sample) sample.r
+ *
+ *     If your engine is already using porting macros, you can define
+ *     SMAA_CUSTOM_SL, and define the porting functions by yourself.
+ *
+ *  7. Then, you'll have to setup the passes as indicated in the scheme above.
+ *     You can take a look into SMAA.fx, to see how we did it for our demo.
+ *     Checkout the function wrappers, you may want to copy-paste them!
+ *
+ *  8. It's recommended to validate the produced |edgesTex| and |blendTex|.
+ *     You can use a screenshot from your engine to compare the |edgesTex|
+ *     and |blendTex| produced inside of the engine with the results obtained
+ *     with the reference demo.
+ *
+ *  9. After you get the last pass to work, it's time to optimize. You'll have
+ *     to initialize a stencil buffer in the first pass (discard is already in
+ *     the code), then mask execution by using it the second pass. The last
+ *     pass should be executed in all pixels.
+ *
+ *
+ * After this point you can choose to enable predicated thresholding,
+ * temporal supersampling and motion blur integration:
+ *
+ * a) If you want to use predicated thresholding, take a look into
+ *    SMAA_PREDICATION; you'll need to pass an extra texture in the edge
+ *    detection pass.
+ *
+ * b) If you want to enable temporal supersampling (SMAA T2x):
+ *
+ * 1. The first step is to render using subpixel jitters. I won't go into
+ *    detail, but it's as simple as moving each vertex position in the
+ *    vertex shader, you can check how we do it in our DX10 demo.
+ *
+ * 2. Then, you must setup the temporal resolve. You may want to take a look
+ *    into SMAAResolve for resolving 2x modes. After you get it working, you'll
+ *    probably see ghosting everywhere. But fear not, you can enable the
+ *    CryENGINE temporal reprojection by setting the SMAA_REPROJECTION macro.
+ *    Check out SMAA_DECODE_VELOCITY if your velocity buffer is encoded.
+ *
+ * 3. The next step is to apply SMAA to each subpixel jittered frame, just as
+ *    done for 1x.
+ *
+ * 4. At this point you should already have something usable, but for best
+ *    results the proper area textures must be set depending on current jitter.
+ *    For this, the parameter 'subsampleIndices' of
+ *    'SMAABlendingWeightCalculationPS' must be set as follows, for our T2x
+ *    mode:
+ *
+ *    @SUBSAMPLE_INDICES
+ *
+ *    | S# |  Camera Jitter   |  subsampleIndices    |
+ *    +----+------------------+---------------------+
+ *    |  0 |  ( 0.25, -0.25)  |  float4(1, 1, 1, 0)  |
+ *    |  1 |  (-0.25,  0.25)  |  float4(2, 2, 2, 0)  |
+ *
+ *    These jitter positions assume a bottom-to-top y axis. S# stands for the
+ *    sample number.
+ *
+ * More information about temporal supersampling here:
+ *    http://iryoku.com/aacourse/downloads/13-Anti-Aliasing-Methods-in-CryENGINE-3.pdf
+ *
+ * c) If you want to enable spatial multisampling (SMAA S2x):
+ *
+ * 1. The scene must be rendered using MSAA 2x. The MSAA 2x buffer must be
+ *    created with:
+ *      - DX10:     see below (*)
+ *      - DX10.1:   D3D10_STANDARD_MULTISAMPLE_PATTERN or
+ *      - DX11:     D3D11_STANDARD_MULTISAMPLE_PATTERN
+ *
+ *    This allows to ensure that the subsample order matches the table in
+ *    @SUBSAMPLE_INDICES.
+ *
+ *    (*) In the case of DX10, we refer the reader to:
+ *      - SMAA::detectMSAAOrder and
+ *      - SMAA::msaaReorder
+ *
+ *    These functions allow to match the standard multisample patterns by
+ *    detecting the subsample order for a specific GPU, and reordering
+ *    them appropriately.
+ *
+ * 2. A shader must be run to output each subsample into a separate buffer
+ *    (DX10 is required). You can use SMAASeparate for this purpose, or just do
+ *    it in an existing pass (for example, in the tone mapping pass, which has
+ *    the advantage of feeding tone mapped subsamples to SMAA, which will yield
+ *    better results).
+ *
+ * 3. The full SMAA 1x pipeline must be run for each separated buffer, storing
+ *    the results in the final buffer. The second run should alpha blend with
+ *    the existing final buffer using a blending factor of 0.5.
+ *    'subsampleIndices' must be adjusted as in the SMAA T2x case (see point
+ *    b).
+ *
+ * d) If you want to enable temporal supersampling on top of SMAA S2x
+ *    (which actually is SMAA 4x):
+ *
+ * 1. SMAA 4x consists on temporally jittering SMAA S2x, so the first step is
+ *    to calculate SMAA S2x for current frame. In this case, 'subsampleIndices'
+ *    must be set as follows:
+ *
+ *    | F# | S# |   Camera Jitter    |    Net Jitter     |   subsampleIndices   |
+ *    +----+----+--------------------+-------------------+----------------------+
+ *    |  0 |  0 |  ( 0.125,  0.125)  |  ( 0.375, -0.125) |  float4(5, 3, 1, 3)  |
+ *    |  0 |  1 |  ( 0.125,  0.125)  |  (-0.125,  0.375) |  float4(4, 6, 2, 3)  |
+ *    +----+----+--------------------+-------------------+----------------------+
+ *    |  1 |  2 |  (-0.125, -0.125)  |  ( 0.125, -0.375) |  float4(3, 5, 1, 4)  |
+ *    |  1 |  3 |  (-0.125, -0.125)  |  (-0.375,  0.125) |  float4(6, 4, 2, 4)  |
+ *
+ *    These jitter positions assume a bottom-to-top y axis. F# stands for the
+ *    frame number. S# stands for the sample number.
+ *
+ * 2. After calculating SMAA S2x for current frame (with the new subsample
+ *    indices), previous frame must be reprojected as in SMAA T2x mode (see
+ *    point b).
+ *
+ * e) If motion blur is used, you may want to do the edge detection pass
+ *    together with motion blur. This has two advantages:
+ *
+ * 1. Pixels under heavy motion can be omitted from the edge detection process.
+ *    For these pixels we can just store "no edge", as motion blur will take
+ *    care of them.
+ * 2. The center pixel tap is reused.
+ *
+ * Note that in this case depth testing should be used instead of stenciling,
+ * as we have to write all the pixels in the motion blur pass.
+ *
+ * That's it!
+ */
+
+//-----------------------------------------------------------------------------
+// SMAA Presets
+
+/**
+ * Note that if you use one of these presets, the following configuration
+ * macros will be ignored if set in the "Configurable Defines" section.
+ */
+
+#if defined(SMAA_PRESET_LOW)
+#define SMAA_THRESHOLD 0.15
+#define SMAA_MAX_SEARCH_STEPS 4
+#define SMAA_DISABLE_DIAG_DETECTION
+#define SMAA_DISABLE_CORNER_DETECTION
+#elif defined(SMAA_PRESET_MEDIUM)
+#define SMAA_THRESHOLD 0.1
+#define SMAA_MAX_SEARCH_STEPS 8
+#define SMAA_DISABLE_DIAG_DETECTION
+#define SMAA_DISABLE_CORNER_DETECTION
+#elif defined(SMAA_PRESET_HIGH)
+#define SMAA_THRESHOLD 0.1
+#define SMAA_MAX_SEARCH_STEPS 16
+#define SMAA_MAX_SEARCH_STEPS_DIAG 8
+#define SMAA_CORNER_ROUNDING 25
+#elif defined(SMAA_PRESET_ULTRA)
+#define SMAA_THRESHOLD 0.05
+#define SMAA_MAX_SEARCH_STEPS 32
+#define SMAA_MAX_SEARCH_STEPS_DIAG 16
+#define SMAA_CORNER_ROUNDING 25
+#endif
+
+//-----------------------------------------------------------------------------
+// Configurable Defines
+
+/**
+ * SMAA_THRESHOLD specifies the threshold or sensitivity to edges.
+ * Lowering this value you will be able to detect more edges at the expense of
+ * performance. 
+ *
+ * Range: [0, 0.5]
+ *   0.1 is a reasonable value, and allows to catch most visible edges.
+ *   0.05 is a rather overkill value, that allows to catch 'em all.
+ *
+ *   If temporal supersampling is used, 0.2 could be a reasonable value, as low
+ *   contrast edges are properly filtered by just 2x.
+ */
+#ifndef SMAA_THRESHOLD
+#define SMAA_THRESHOLD 0.1
+#endif
+
+/**
+ * SMAA_DEPTH_THRESHOLD specifies the threshold for depth edge detection.
+ * 
+ * Range: depends on the depth range of the scene.
+ */
+#ifndef SMAA_DEPTH_THRESHOLD
+#define SMAA_DEPTH_THRESHOLD (0.1 * SMAA_THRESHOLD)
+#endif
+
+/**
+ * SMAA_MAX_SEARCH_STEPS specifies the maximum steps performed in the
+ * horizontal/vertical pattern searches, at each side of the pixel.
+ *
+ * In number of pixels, it's actually the double. So the maximum line length
+ * perfectly handled by, for example 16, is 64 (by perfectly, we meant that
+ * longer lines won't look as good, but still antialiased).
+ *
+ * Range: [0, 112]
+ */
+#ifndef SMAA_MAX_SEARCH_STEPS
+#define SMAA_MAX_SEARCH_STEPS 16
+#endif
+
+/**
+ * SMAA_MAX_SEARCH_STEPS_DIAG specifies the maximum steps performed in the
+ * diagonal pattern searches, at each side of the pixel. In this case we jump
+ * one pixel at time, instead of two.
+ *
+ * Range: [0, 20]
+ *
+ * On high-end machines it is cheap (between a 0.8x and 0.9x slower for 16 
+ * steps), but it can have a significant impact on older machines.
+ *
+ * Define SMAA_DISABLE_DIAG_DETECTION to disable diagonal processing.
+ */
+#ifndef SMAA_MAX_SEARCH_STEPS_DIAG
+#define SMAA_MAX_SEARCH_STEPS_DIAG 8
+#endif
+
+/**
+ * SMAA_CORNER_ROUNDING specifies how much sharp corners will be rounded.
+ *
+ * Range: [0, 100]
+ *
+ * Define SMAA_DISABLE_CORNER_DETECTION to disable corner processing.
+ */
+#ifndef SMAA_CORNER_ROUNDING
+#define SMAA_CORNER_ROUNDING 25
+#endif
+
+/**
+ * If there is an neighbor edge that has SMAA_LOCAL_CONTRAST_FACTOR times
+ * bigger contrast than current edge, current edge will be discarded.
+ *
+ * This allows to eliminate spurious crossing edges, and is based on the fact
+ * that, if there is too much contrast in a direction, that will hide
+ * perceptually contrast in the other neighbors.
+ */
+#ifndef SMAA_LOCAL_CONTRAST_ADAPTATION_FACTOR
+#define SMAA_LOCAL_CONTRAST_ADAPTATION_FACTOR 2.0
+#endif
+
+/**
+ * Predicated thresholding allows to better preserve texture details and to
+ * improve performance, by decreasing the number of detected edges using an
+ * additional buffer like the light accumulation buffer, object ids or even the
+ * depth buffer (the depth buffer usage may be limited to indoor or short range
+ * scenes).
+ *
+ * It locally decreases the luma or color threshold if an edge is found in an
+ * additional buffer (so the global threshold can be higher).
+ *
+ * This method was developed by Playstation EDGE MLAA team, and used in 
+ * Killzone 3, by using the light accumulation buffer. More information here:
+ *     http://iryoku.com/aacourse/downloads/06-MLAA-on-PS3.pptx 
+ */
+#ifndef SMAA_PREDICATION
+#define SMAA_PREDICATION 0
+#endif
+
+/**
+ * Threshold to be used in the additional predication buffer. 
+ *
+ * Range: depends on the input, so you'll have to find the magic number that
+ * works for you.
+ */
+#ifndef SMAA_PREDICATION_THRESHOLD
+#define SMAA_PREDICATION_THRESHOLD 0.01
+#endif
+
+/**
+ * How much to scale the global threshold used for luma or color edge
+ * detection when using predication.
+ *
+ * Range: [1, 5]
+ */
+#ifndef SMAA_PREDICATION_SCALE
+#define SMAA_PREDICATION_SCALE 2.0
+#endif
+
+/**
+ * How much to locally decrease the threshold.
+ *
+ * Range: [0, 1]
+ */
+#ifndef SMAA_PREDICATION_STRENGTH
+#define SMAA_PREDICATION_STRENGTH 0.4
+#endif
+
+/**
+ * Temporal reprojection allows to remove ghosting artifacts when using
+ * temporal supersampling. We use the CryEngine 3 method which also introduces
+ * velocity weighting. This feature is of extreme importance for totally
+ * removing ghosting. More information here:
+ *    http://iryoku.com/aacourse/downloads/13-Anti-Aliasing-Methods-in-CryENGINE-3.pdf
+ *
+ * Note that you'll need to setup a velocity buffer for enabling reprojection.
+ * For static geometry, saving the previous depth buffer is a viable
+ * alternative.
+ */
+#ifndef SMAA_REPROJECTION
+#define SMAA_REPROJECTION 0
+#endif
+
+/**
+ * SMAA_REPROJECTION_WEIGHT_SCALE controls the velocity weighting. It allows to
+ * remove ghosting trails behind the moving object, which are not removed by
+ * just using reprojection. Using low values will exhibit ghosting, while using
+ * high values will disable temporal supersampling under motion.
+ *
+ * Behind the scenes, velocity weighting removes temporal supersampling when
+ * the velocity of the subsamples differs (meaning they are different objects).
+ *
+ * Range: [0, 80]
+ */
+#ifndef SMAA_REPROJECTION_WEIGHT_SCALE
+#define SMAA_REPROJECTION_WEIGHT_SCALE 30.0
+#endif
+
+/**
+ * On some compilers, discard cannot be used in vertex shaders. Thus, they need
+ * to be compiled separately.
+ */
+#ifndef SMAA_INCLUDE_VS
+#define SMAA_INCLUDE_VS 1
+#endif
+#ifndef SMAA_INCLUDE_PS
+#define SMAA_INCLUDE_PS 1
+#endif
+
+//-----------------------------------------------------------------------------
+// Texture Access Defines
+
+#ifndef SMAA_AREATEX_SELECT
+#if defined(SMAA_HLSL_3)
+#define SMAA_AREATEX_SELECT(sample) sample.ra
+#else
+#define SMAA_AREATEX_SELECT(sample) sample.rg
+#endif
+#endif
+
+#ifndef SMAA_SEARCHTEX_SELECT
+#define SMAA_SEARCHTEX_SELECT(sample) sample.r
+#endif
+
+#ifndef SMAA_DECODE_VELOCITY
+#define SMAA_DECODE_VELOCITY(sample) sample.rg
+#endif
+
+//-----------------------------------------------------------------------------
+// Non-Configurable Defines
+
+#define SMAA_AREATEX_MAX_DISTANCE 16
+#define SMAA_AREATEX_MAX_DISTANCE_DIAG 20
+#define SMAA_AREATEX_PIXEL_SIZE (1.0 / float2(160.0, 560.0))
+#define SMAA_AREATEX_SUBTEX_SIZE (1.0 / 7.0)
+#define SMAA_SEARCHTEX_SIZE float2(66.0, 33.0)
+#define SMAA_SEARCHTEX_PACKED_SIZE float2(64.0, 16.0)
+#define SMAA_CORNER_ROUNDING_NORM (float(SMAA_CORNER_ROUNDING) / 100.0)
+
+//-----------------------------------------------------------------------------
+// Porting Functions
+
+#if defined(SMAA_HLSL_3)
+#define SMAATexture2D(tex) sampler2D tex
+#define SMAATexturePass2D(tex) tex
+#define SMAASampleLevelZero(tex, coord) tex2Dlod(tex, float4(coord, 0.0, 0.0))
+#define SMAASampleLevelZeroPoint(tex, coord) tex2Dlod(tex, float4(coord, 0.0, 0.0))
+#define SMAASampleLevelZeroOffset(tex, coord, offset) tex2Dlod(tex, float4(coord + offset * SMAA_RT_METRICS.xy, 0.0, 0.0))
+#define SMAASample(tex, coord) tex2D(tex, coord)
+#define SMAASamplePoint(tex, coord) tex2D(tex, coord)
+#define SMAASampleOffset(tex, coord, offset) tex2D(tex, coord + offset * SMAA_RT_METRICS.xy)
+#define SMAA_FLATTEN [flatten]
+#define SMAA_BRANCH [branch]
+#endif
+#if defined(SMAA_HLSL_4) || defined(SMAA_HLSL_4_1)
+SamplerState LinearSampler { Filter = MIN_MAG_LINEAR_MIP_POINT; AddressU = Clamp; AddressV = Clamp; };
+SamplerState PointSampler { Filter = MIN_MAG_MIP_POINT; AddressU = Clamp; AddressV = Clamp; };
+#define SMAATexture2D(tex) Texture2D tex
+#define SMAATexturePass2D(tex) tex
+#define SMAASampleLevelZero(tex, coord) tex.SampleLevel(LinearSampler, coord, 0)
+#define SMAASampleLevelZeroPoint(tex, coord) tex.SampleLevel(PointSampler, coord, 0)
+#define SMAASampleLevelZeroOffset(tex, coord, offset) tex.SampleLevel(LinearSampler, coord, 0, offset)
+#define SMAASample(tex, coord) tex.Sample(LinearSampler, coord)
+#define SMAASamplePoint(tex, coord) tex.Sample(PointSampler, coord)
+#define SMAASampleOffset(tex, coord, offset) tex.Sample(LinearSampler, coord, offset)
+#define SMAA_FLATTEN [flatten]
+#define SMAA_BRANCH [branch]
+#define SMAATexture2DMS2(tex) Texture2DMS<float4, 2> tex
+#define SMAALoad(tex, pos, sample) tex.Load(pos, sample)
+#if defined(SMAA_HLSL_4_1)
+#define SMAAGather(tex, coord) tex.Gather(LinearSampler, coord, 0)
+#endif
+#endif
+#if defined(SMAA_GLSL_3) || defined(SMAA_GLSL_4)
+#define SMAATexture2D(tex) sampler2D tex
+#define SMAATexturePass2D(tex) tex
+#define SMAASampleLevelZero(tex, coord) textureLod(tex, coord, 0.0)
+#define SMAASampleLevelZeroPoint(tex, coord) textureLod(tex, coord, 0.0)
+#define SMAASampleLevelZeroOffset(tex, coord, offset) textureLodOffset(tex, coord, 0.0, offset)
+#define SMAASample(tex, coord) texture(tex, coord)
+#define SMAASamplePoint(tex, coord) texture(tex, coord)
+#define SMAASampleOffset(tex, coord, offset) texture(tex, coord, offset)
+#define SMAA_FLATTEN
+#define SMAA_BRANCH
+#define lerp(a, b, t) mix(a, b, t)
+#define saturate(a) clamp(a, 0.0, 1.0)
+#if defined(SMAA_GLSL_4)
+#define mad(a, b, c) fma(a, b, c)
+#define SMAAGather(tex, coord) textureGather(tex, coord)
+#else
+#define mad(a, b, c) (a * b + c)
+#endif
+#define float2 vec2
+#define float3 vec3
+#define float4 vec4
+#define int2 ivec2
+#define int3 ivec3
+#define int4 ivec4
+#define bool2 bvec2
+#define bool3 bvec3
+#define bool4 bvec4
+#endif
+
+#if !defined(SMAA_HLSL_3) && !defined(SMAA_HLSL_4) && !defined(SMAA_HLSL_4_1) && !defined(SMAA_GLSL_3) && !defined(SMAA_GLSL_4) && !defined(SMAA_CUSTOM_SL)
+#error you must define the shading language: SMAA_HLSL_*, SMAA_GLSL_* or SMAA_CUSTOM_SL
+#endif
+
+//-----------------------------------------------------------------------------
+// Misc functions
+
+/**
+ * Gathers current pixel, and the top-left neighbors.
+ */
+float3 SMAAGatherNeighbours(float2 texcoord,
+                            float4 offset[3],
+                            SMAATexture2D(tex)) {
+    #ifdef SMAAGather
+    return SMAAGather(tex, texcoord + SMAA_RT_METRICS.xy * float2(-0.5, -0.5)).grb;
+    #else
+    float P = SMAASamplePoint(tex, texcoord).r;
+    float Pleft = SMAASamplePoint(tex, offset[0].xy).r;
+    float Ptop  = SMAASamplePoint(tex, offset[0].zw).r;
+    return float3(P, Pleft, Ptop);
+    #endif
+}
+
+/**
+ * Adjusts the threshold by means of predication.
+ */
+float2 SMAACalculatePredicatedThreshold(float2 texcoord,
+                                        float4 offset[3],
+                                        SMAATexture2D(predicationTex)) {
+    float3 neighbours = SMAAGatherNeighbours(texcoord, offset, SMAATexturePass2D(predicationTex));
+    float2 delta = abs(neighbours.xx - neighbours.yz);
+    float2 edges = step(SMAA_PREDICATION_THRESHOLD, delta);
+    return SMAA_PREDICATION_SCALE * SMAA_THRESHOLD * (1.0 - SMAA_PREDICATION_STRENGTH * edges);
+}
+
+/**
+ * Conditional move:
+ */
+void SMAAMovc(bool2 cond, inout float2 variable, float2 value) {
+    SMAA_FLATTEN if (cond.x) variable.x = value.x;
+    SMAA_FLATTEN if (cond.y) variable.y = value.y;
+}
+
+void SMAAMovc(bool4 cond, inout float4 variable, float4 value) {
+    SMAAMovc(cond.xy, variable.xy, value.xy);
+    SMAAMovc(cond.zw, variable.zw, value.zw);
+}
+
+
+#if SMAA_INCLUDE_VS
+//-----------------------------------------------------------------------------
+// Vertex Shaders
+
+/**
+ * Edge Detection Vertex Shader
+ */
+void SMAAEdgeDetectionVS(float2 texcoord,
+                         out float4 offset[3]) {
+    offset[0] = mad(SMAA_RT_METRICS.xyxy, float4(-1.0, 0.0, 0.0, -1.0), texcoord.xyxy);
+    offset[1] = mad(SMAA_RT_METRICS.xyxy, float4( 1.0, 0.0, 0.0,  1.0), texcoord.xyxy);
+    offset[2] = mad(SMAA_RT_METRICS.xyxy, float4(-2.0, 0.0, 0.0, -2.0), texcoord.xyxy);
+}
+
+/**
+ * Blend Weight Calculation Vertex Shader
+ */
+void SMAABlendingWeightCalculationVS(float2 texcoord,
+                                     out float2 pixcoord,
+                                     out float4 offset[3]) {
+    pixcoord = texcoord * SMAA_RT_METRICS.zw;
+
+    // We will use these offsets for the searches later on (see @PSEUDO_GATHER4):
+    offset[0] = mad(SMAA_RT_METRICS.xyxy, float4(-0.25, -0.125,  1.25, -0.125), texcoord.xyxy);
+    offset[1] = mad(SMAA_RT_METRICS.xyxy, float4(-0.125, -0.25, -0.125,  1.25), texcoord.xyxy);
+
+    // And these for the searches, they indicate the ends of the loops:
+    offset[2] = mad(SMAA_RT_METRICS.xxyy,
+                    float4(-2.0, 2.0, -2.0, 2.0) * float(SMAA_MAX_SEARCH_STEPS),
+                    float4(offset[0].xz, offset[1].yw));
+}
+
+/**
+ * Neighborhood Blending Vertex Shader
+ */
+void SMAANeighborhoodBlendingVS(float2 texcoord,
+                                out float4 offset) {
+    offset = mad(SMAA_RT_METRICS.xyxy, float4( 1.0, 0.0, 0.0,  1.0), texcoord.xyxy);
+}
+#endif // SMAA_INCLUDE_VS
+
+#if SMAA_INCLUDE_PS
+//-----------------------------------------------------------------------------
+// Edge Detection Pixel Shaders (First Pass)
+
+/**
+ * Luma Edge Detection
+ *
+ * IMPORTANT NOTICE: luma edge detection requires gamma-corrected colors, and
+ * thus 'colorTex' should be a non-sRGB texture.
+ */
+float2 SMAALumaEdgeDetectionPS(float2 texcoord,
+                               float4 offset[3],
+                               SMAATexture2D(colorTex)
+                               #if SMAA_PREDICATION
+                               , SMAATexture2D(predicationTex)
+                               #endif
+                               ) {
+    // Calculate the threshold:
+    #if SMAA_PREDICATION
+    float2 threshold = SMAACalculatePredicatedThreshold(texcoord, offset, SMAATexturePass2D(predicationTex));
+    #else
+    float2 threshold = float2(SMAA_THRESHOLD, SMAA_THRESHOLD);
+    #endif
+
+    // Calculate lumas:
+    float3 weights = float3(0.2126, 0.7152, 0.0722);
+    float L = dot(SMAASamplePoint(colorTex, texcoord).rgb, weights);
+
+    float Lleft = dot(SMAASamplePoint(colorTex, offset[0].xy).rgb, weights);
+    float Ltop  = dot(SMAASamplePoint(colorTex, offset[0].zw).rgb, weights);
+
+    // We do the usual threshold:
+    float4 delta;
+    delta.xy = abs(L - float2(Lleft, Ltop));
+    float2 edges = step(threshold, delta.xy);
+
+    // Then discard if there is no edge:
+    if (dot(edges, float2(1.0, 1.0)) == 0.0)
+        return float2(-2.0, -2.0);
+
+    // Calculate right and bottom deltas:
+    float Lright = dot(SMAASamplePoint(colorTex, offset[1].xy).rgb, weights);
+    float Lbottom  = dot(SMAASamplePoint(colorTex, offset[1].zw).rgb, weights);
+    delta.zw = abs(L - float2(Lright, Lbottom));
+
+    // Calculate the maximum delta in the direct neighborhood:
+    float2 maxDelta = max(delta.xy, delta.zw);
+
+    // Calculate left-left and top-top deltas:
+    float Lleftleft = dot(SMAASamplePoint(colorTex, offset[2].xy).rgb, weights);
+    float Ltoptop = dot(SMAASamplePoint(colorTex, offset[2].zw).rgb, weights);
+    delta.zw = abs(float2(Lleft, Ltop) - float2(Lleftleft, Ltoptop));
+
+    // Calculate the final maximum delta:
+    maxDelta = max(maxDelta.xy, delta.zw);
+    float finalDelta = max(maxDelta.x, maxDelta.y);
+
+    // Local contrast adaptation:
+    edges.xy *= step(finalDelta, SMAA_LOCAL_CONTRAST_ADAPTATION_FACTOR * delta.xy);
+
+    return edges;
+}
+
+/**
+ * Color Edge Detection
+ *
+ * IMPORTANT NOTICE: color edge detection requires gamma-corrected colors, and
+ * thus 'colorTex' should be a non-sRGB texture.
+ */
+float2 SMAAColorEdgeDetectionPS(float2 texcoord,
+                                float4 offset[3],
+                                SMAATexture2D(colorTex)
+                                #if SMAA_PREDICATION
+                                , SMAATexture2D(predicationTex)
+                                #endif
+                                ) {
+    // Calculate the threshold:
+    #if SMAA_PREDICATION
+    float2 threshold = SMAACalculatePredicatedThreshold(texcoord, offset, predicationTex);
+    #else
+    float2 threshold = float2(SMAA_THRESHOLD, SMAA_THRESHOLD);
+    #endif
+
+    // Calculate color deltas:
+    float4 delta;
+    float3 C = SMAASamplePoint(colorTex, texcoord).rgb;
+
+    float3 Cleft = SMAASamplePoint(colorTex, offset[0].xy).rgb;
+    float3 t = abs(C - Cleft);
+    delta.x = max(max(t.r, t.g), t.b);
+
+    float3 Ctop  = SMAASamplePoint(colorTex, offset[0].zw).rgb;
+    t = abs(C - Ctop);
+    delta.y = max(max(t.r, t.g), t.b);
+
+    // We do the usual threshold:
+    float2 edges = step(threshold, delta.xy);
+
+    // Then discard if there is no edge:
+    if (dot(edges, float2(1.0, 1.0)) == 0.0)
+        return float2(-2.0, -2.0);
+
+    // Calculate right and bottom deltas:
+    float3 Cright = SMAASamplePoint(colorTex, offset[1].xy).rgb;
+    t = abs(C - Cright);
+    delta.z = max(max(t.r, t.g), t.b);
+
+    float3 Cbottom  = SMAASamplePoint(colorTex, offset[1].zw).rgb;
+    t = abs(C - Cbottom);
+    delta.w = max(max(t.r, t.g), t.b);
+
+    // Calculate the maximum delta in the direct neighborhood:
+    float2 maxDelta = max(delta.xy, delta.zw);
+
+    // Calculate left-left and top-top deltas:
+    float3 Cleftleft  = SMAASamplePoint(colorTex, offset[2].xy).rgb;
+    t = abs(C - Cleftleft);
+    delta.z = max(max(t.r, t.g), t.b);
+
+    float3 Ctoptop = SMAASamplePoint(colorTex, offset[2].zw).rgb;
+    t = abs(C - Ctoptop);
+    delta.w = max(max(t.r, t.g), t.b);
+
+    // Calculate the final maximum delta:
+    maxDelta = max(maxDelta.xy, delta.zw);
+    float finalDelta = max(maxDelta.x, maxDelta.y);
+
+    // Local contrast adaptation:
+    edges.xy *= step(finalDelta, SMAA_LOCAL_CONTRAST_ADAPTATION_FACTOR * delta.xy);
+
+    return edges;
+}
+
+/**
+ * Depth Edge Detection
+ */
+float2 SMAADepthEdgeDetectionPS(float2 texcoord,
+                                float4 offset[3],
+                                SMAATexture2D(depthTex)) {
+    float3 neighbours = SMAAGatherNeighbours(texcoord, offset, SMAATexturePass2D(depthTex));
+    float2 delta = abs(neighbours.xx - float2(neighbours.y, neighbours.z));
+    float2 edges = step(SMAA_DEPTH_THRESHOLD, delta);
+
+    if (dot(edges, float2(1.0, 1.0)) == 0.0)
+        return float2(-2.0, -2.0);
+
+    return edges;
+}
+
+//-----------------------------------------------------------------------------
+// Diagonal Search Functions
+
+#if !defined(SMAA_DISABLE_DIAG_DETECTION)
+
+/**
+ * Allows to decode two binary values from a bilinear-filtered access.
+ */
+float2 SMAADecodeDiagBilinearAccess(float2 e) {
+    // Bilinear access for fetching 'e' have a 0.25 offset, and we are
+    // interested in the R and G edges:
+    //
+    // +---G---+-------+
+    // |   x o R   x   |
+    // +-------+-------+
+    //
+    // Then, if one of these edge is enabled:
+    //   Red:   (0.75 * X + 0.25 * 1) => 0.25 or 1.0
+    //   Green: (0.75 * 1 + 0.25 * X) => 0.75 or 1.0
+    //
+    // This function will unpack the values (mad + mul + round):
+    // wolframalpha.com: round(x * abs(5 * x - 5 * 0.75)) plot 0 to 1
+    e.r = e.r * abs(5.0 * e.r - 5.0 * 0.75);
+    return round(e);
+}
+
+float4 SMAADecodeDiagBilinearAccess(float4 e) {
+    e.rb = e.rb * abs(5.0 * e.rb - 5.0 * 0.75);
+    return round(e);
+}
+
+/**
+ * These functions allows to perform diagonal pattern searches.
+ */
+float2 SMAASearchDiag1(SMAATexture2D(edgesTex), float2 texcoord, float2 dir, out float2 e) {
+    float4 coord = float4(texcoord, -1.0, 1.0);
+    float3 t = float3(SMAA_RT_METRICS.xy, 1.0);
+    while (coord.z < float(SMAA_MAX_SEARCH_STEPS_DIAG - 1) &&
+           coord.w > 0.9) {
+        coord.xyz = mad(t, float3(dir, 1.0), coord.xyz);
+        e = SMAASampleLevelZero(edgesTex, coord.xy).rg;
+        coord.w = dot(e, float2(0.5, 0.5));
+    }
+    return coord.zw;
+}
+
+float2 SMAASearchDiag2(SMAATexture2D(edgesTex), float2 texcoord, float2 dir, out float2 e) {
+    float4 coord = float4(texcoord, -1.0, 1.0);
+    coord.x += 0.25 * SMAA_RT_METRICS.x; // See @SearchDiag2Optimization
+    float3 t = float3(SMAA_RT_METRICS.xy, 1.0);
+    while (coord.z < float(SMAA_MAX_SEARCH_STEPS_DIAG - 1) &&
+           coord.w > 0.9) {
+        coord.xyz = mad(t, float3(dir, 1.0), coord.xyz);
+
+        // @SearchDiag2Optimization
+        // Fetch both edges at once using bilinear filtering:
+        e = SMAASampleLevelZero(edgesTex, coord.xy).rg;
+        e = SMAADecodeDiagBilinearAccess(e);
+
+        // Non-optimized version:
+        // e.g = SMAASampleLevelZero(edgesTex, coord.xy).g;
+        // e.r = SMAASampleLevelZeroOffset(edgesTex, coord.xy, int2(1, 0)).r;
+
+        coord.w = dot(e, float2(0.5, 0.5));
+    }
+    return coord.zw;
+}
+
+/** 
+ * Similar to SMAAArea, this calculates the area corresponding to a certain
+ * diagonal distance and crossing edges 'e'.
+ */
+float2 SMAAAreaDiag(SMAATexture2D(areaTex), float2 dist, float2 e, float offset) {
+    float2 texcoord = mad(float2(SMAA_AREATEX_MAX_DISTANCE_DIAG, SMAA_AREATEX_MAX_DISTANCE_DIAG), e, dist);
+
+    // We do a scale and bias for mapping to texel space:
+    texcoord = mad(SMAA_AREATEX_PIXEL_SIZE, texcoord, 0.5 * SMAA_AREATEX_PIXEL_SIZE);
+
+    // Diagonal areas are on the second half of the texture:
+    texcoord.x += 0.5;
+
+    // Move to proper place, according to the subpixel offset:
+    texcoord.y += SMAA_AREATEX_SUBTEX_SIZE * offset;
+
+    // Do it!
+    return SMAA_AREATEX_SELECT(SMAASampleLevelZero(areaTex, texcoord));
+}
+
+/**
+ * This searches for diagonal patterns and returns the corresponding weights.
+ */
+float2 SMAACalculateDiagWeights(SMAATexture2D(edgesTex), SMAATexture2D(areaTex), float2 texcoord, float2 e, float4 subsampleIndices) {
+    float2 weights = float2(0.0, 0.0);
+
+    // Search for the line ends:
+    float4 d;
+    float2 end;
+    if (e.r > 0.0) {
+        d.xz = SMAASearchDiag1(SMAATexturePass2D(edgesTex), texcoord, float2(-1.0,  1.0), end);
+        d.x += float(end.y > 0.9);
+    } else
+        d.xz = float2(0.0, 0.0);
+    d.yw = SMAASearchDiag1(SMAATexturePass2D(edgesTex), texcoord, float2(1.0, -1.0), end);
+
+    SMAA_BRANCH
+    if (d.x + d.y > 2.0) { // d.x + d.y + 1 > 3
+        // Fetch the crossing edges:
+        float4 coords = mad(float4(-d.x + 0.25, d.x, d.y, -d.y - 0.25), SMAA_RT_METRICS.xyxy, texcoord.xyxy);
+        float4 c;
+        c.xy = SMAASampleLevelZeroOffset(edgesTex, coords.xy, int2(-1,  0)).rg;
+        c.zw = SMAASampleLevelZeroOffset(edgesTex, coords.zw, int2( 1,  0)).rg;
+        c.yxwz = SMAADecodeDiagBilinearAccess(c.xyzw);
+
+        // Non-optimized version:
+        // float4 coords = mad(float4(-d.x, d.x, d.y, -d.y), SMAA_RT_METRICS.xyxy, texcoord.xyxy);
+        // float4 c;
+        // c.x = SMAASampleLevelZeroOffset(edgesTex, coords.xy, int2(-1,  0)).g;
+        // c.y = SMAASampleLevelZeroOffset(edgesTex, coords.xy, int2( 0,  0)).r;
+        // c.z = SMAASampleLevelZeroOffset(edgesTex, coords.zw, int2( 1,  0)).g;
+        // c.w = SMAASampleLevelZeroOffset(edgesTex, coords.zw, int2( 1, -1)).r;
+
+        // Merge crossing edges at each side into a single value:
+        float2 cc = mad(float2(2.0, 2.0), c.xz, c.yw);
+
+        // Remove the crossing edge if we didn't found the end of the line:
+        SMAAMovc(bool2(step(0.9, d.zw)), cc, float2(0.0, 0.0));
+
+        // Fetch the areas for this line:
+        weights += SMAAAreaDiag(SMAATexturePass2D(areaTex), d.xy, cc, subsampleIndices.z);
+    }
+
+    // Search for the line ends:
+    d.xz = SMAASearchDiag2(SMAATexturePass2D(edgesTex), texcoord, float2(-1.0, -1.0), end);
+    if (SMAASampleLevelZeroOffset(edgesTex, texcoord, int2(1, 0)).r > 0.0) {
+        d.yw = SMAASearchDiag2(SMAATexturePass2D(edgesTex), texcoord, float2(1.0, 1.0), end);
+        d.y += float(end.y > 0.9);
+    } else
+        d.yw = float2(0.0, 0.0);
+
+    SMAA_BRANCH
+    if (d.x + d.y > 2.0) { // d.x + d.y + 1 > 3
+        // Fetch the crossing edges:
+        float4 coords = mad(float4(-d.x, -d.x, d.y, d.y), SMAA_RT_METRICS.xyxy, texcoord.xyxy);
+        float4 c;
+        c.x  = SMAASampleLevelZeroOffset(edgesTex, coords.xy, int2(-1,  0)).g;
+        c.y  = SMAASampleLevelZeroOffset(edgesTex, coords.xy, int2( 0, -1)).r;
+        c.zw = SMAASampleLevelZeroOffset(edgesTex, coords.zw, int2( 1,  0)).gr;
+        float2 cc = mad(float2(2.0, 2.0), c.xz, c.yw);
+
+        // Remove the crossing edge if we didn't found the end of the line:
+        SMAAMovc(bool2(step(0.9, d.zw)), cc, float2(0.0, 0.0));
+
+        // Fetch the areas for this line:
+        weights += SMAAAreaDiag(SMAATexturePass2D(areaTex), d.xy, cc, subsampleIndices.w).gr;
+    }
+
+    return weights;
+}
+#endif
+
+//-----------------------------------------------------------------------------
+// Horizontal/Vertical Search Functions
+
+/**
+ * This allows to determine how much length should we add in the last step
+ * of the searches. It takes the bilinearly interpolated edge (see 
+ * @PSEUDO_GATHER4), and adds 0, 1 or 2, depending on which edges and
+ * crossing edges are active.
+ */
+float SMAASearchLength(SMAATexture2D(searchTex), float2 e, float offset) {
+    // The texture is flipped vertically, with left and right cases taking half
+    // of the space horizontally:
+    float2 scale = SMAA_SEARCHTEX_SIZE * float2(0.5, -1.0);
+    float2 bias = SMAA_SEARCHTEX_SIZE * float2(offset, 1.0);
+
+    // Scale and bias to access texel centers:
+    scale += float2(-1.0,  1.0);
+    bias  += float2( 0.5, -0.5);
+
+    // Convert from pixel coordinates to texcoords:
+    // (We use SMAA_SEARCHTEX_PACKED_SIZE because the texture is cropped)
+    scale *= 1.0 / SMAA_SEARCHTEX_PACKED_SIZE;
+    bias *= 1.0 / SMAA_SEARCHTEX_PACKED_SIZE;
+
+    // Lookup the search texture:
+    return SMAA_SEARCHTEX_SELECT(SMAASampleLevelZero(searchTex, mad(scale, e, bias)));
+}
+
+/**
+ * Horizontal/vertical search functions for the 2nd pass.
+ */
+float SMAASearchXLeft(SMAATexture2D(edgesTex), SMAATexture2D(searchTex), float2 texcoord, float end) {
+    /**
+     * @PSEUDO_GATHER4
+     * This texcoord has been offset by (-0.25, -0.125) in the vertex shader to
+     * sample between edge, thus fetching four edges in a row.
+     * Sampling with different offsets in each direction allows to disambiguate
+     * which edges are active from the four fetched ones.
+     */
+    float2 e = float2(0.0, 1.0);
+    while (texcoord.x > end && 
+           e.g > 0.8281 && // Is there some edge not activated?
+           e.r == 0.0) { // Or is there a crossing edge that breaks the line?
+        e = SMAASampleLevelZero(edgesTex, texcoord).rg;
+        texcoord = mad(-float2(2.0, 0.0), SMAA_RT_METRICS.xy, texcoord);
+    }
+
+    float offset = mad(-(255.0 / 127.0), SMAASearchLength(SMAATexturePass2D(searchTex), e, 0.0), 3.25);
+    return mad(SMAA_RT_METRICS.x, offset, texcoord.x);
+
+    // Non-optimized version:
+    // We correct the previous (-0.25, -0.125) offset we applied:
+    // texcoord.x += 0.25 * SMAA_RT_METRICS.x;
+
+    // The searches are bias by 1, so adjust the coords accordingly:
+    // texcoord.x += SMAA_RT_METRICS.x;
+
+    // Disambiguate the length added by the last step:
+    // texcoord.x += 2.0 * SMAA_RT_METRICS.x; // Undo last step
+    // texcoord.x -= SMAA_RT_METRICS.x * (255.0 / 127.0) * SMAASearchLength(SMAATexturePass2D(searchTex), e, 0.0);
+    // return mad(SMAA_RT_METRICS.x, offset, texcoord.x);
+}
+
+float SMAASearchXRight(SMAATexture2D(edgesTex), SMAATexture2D(searchTex), float2 texcoord, float end) {
+    float2 e = float2(0.0, 1.0);
+    while (texcoord.x < end && 
+           e.g > 0.8281 && // Is there some edge not activated?
+           e.r == 0.0) { // Or is there a crossing edge that breaks the line?
+        e = SMAASampleLevelZero(edgesTex, texcoord).rg;
+        texcoord = mad(float2(2.0, 0.0), SMAA_RT_METRICS.xy, texcoord);
+    }
+    float offset = mad(-(255.0 / 127.0), SMAASearchLength(SMAATexturePass2D(searchTex), e, 0.5), 3.25);
+    return mad(-SMAA_RT_METRICS.x, offset, texcoord.x);
+}
+
+float SMAASearchYUp(SMAATexture2D(edgesTex), SMAATexture2D(searchTex), float2 texcoord, float end) {
+    float2 e = float2(1.0, 0.0);
+    while (texcoord.y > end && 
+           e.r > 0.8281 && // Is there some edge not activated?
+           e.g == 0.0) { // Or is there a crossing edge that breaks the line?
+        e = SMAASampleLevelZero(edgesTex, texcoord).rg;
+        texcoord = mad(-float2(0.0, 2.0), SMAA_RT_METRICS.xy, texcoord);
+    }
+    float offset = mad(-(255.0 / 127.0), SMAASearchLength(SMAATexturePass2D(searchTex), e.gr, 0.0), 3.25);
+    return mad(SMAA_RT_METRICS.y, offset, texcoord.y);
+}
+
+float SMAASearchYDown(SMAATexture2D(edgesTex), SMAATexture2D(searchTex), float2 texcoord, float end) {
+    float2 e = float2(1.0, 0.0);
+    while (texcoord.y < end && 
+           e.r > 0.8281 && // Is there some edge not activated?
+           e.g == 0.0) { // Or is there a crossing edge that breaks the line?
+        e = SMAASampleLevelZero(edgesTex, texcoord).rg;
+        texcoord = mad(float2(0.0, 2.0), SMAA_RT_METRICS.xy, texcoord);
+    }
+    float offset = mad(-(255.0 / 127.0), SMAASearchLength(SMAATexturePass2D(searchTex), e.gr, 0.5), 3.25);
+    return mad(-SMAA_RT_METRICS.y, offset, texcoord.y);
+}
+
+/** 
+ * Ok, we have the distance and both crossing edges. So, what are the areas
+ * at each side of current edge?
+ */
+float2 SMAAArea(SMAATexture2D(areaTex), float2 dist, float e1, float e2, float offset) {
+    // Rounding prevents precision errors of bilinear filtering:
+    float2 texcoord = mad(float2(SMAA_AREATEX_MAX_DISTANCE, SMAA_AREATEX_MAX_DISTANCE), round(4.0 * float2(e1, e2)), dist);
+    
+    // We do a scale and bias for mapping to texel space:
+    texcoord = mad(SMAA_AREATEX_PIXEL_SIZE, texcoord, 0.5 * SMAA_AREATEX_PIXEL_SIZE);
+
+    // Move to proper place, according to the subpixel offset:
+    texcoord.y = mad(SMAA_AREATEX_SUBTEX_SIZE, offset, texcoord.y);
+
+    // Do it!
+    return SMAA_AREATEX_SELECT(SMAASampleLevelZero(areaTex, texcoord));
+}
+
+//-----------------------------------------------------------------------------
+// Corner Detection Functions
+
+void SMAADetectHorizontalCornerPattern(SMAATexture2D(edgesTex), inout float2 weights, float4 texcoord, float2 d) {
+    #if !defined(SMAA_DISABLE_CORNER_DETECTION)
+    float2 leftRight = step(d.xy, d.yx);
+    float2 rounding = (1.0 - SMAA_CORNER_ROUNDING_NORM) * leftRight;
+
+    rounding /= leftRight.x + leftRight.y; // Reduce blending for pixels in the center of a line.
+
+    float2 factor = float2(1.0, 1.0);
+    factor.x -= rounding.x * SMAASampleLevelZeroOffset(edgesTex, texcoord.xy, int2(0,  1)).r;
+    factor.x -= rounding.y * SMAASampleLevelZeroOffset(edgesTex, texcoord.zw, int2(1,  1)).r;
+    factor.y -= rounding.x * SMAASampleLevelZeroOffset(edgesTex, texcoord.xy, int2(0, -2)).r;
+    factor.y -= rounding.y * SMAASampleLevelZeroOffset(edgesTex, texcoord.zw, int2(1, -2)).r;
+
+    weights *= saturate(factor);
+    #endif
+}
+
+void SMAADetectVerticalCornerPattern(SMAATexture2D(edgesTex), inout float2 weights, float4 texcoord, float2 d) {
+    #if !defined(SMAA_DISABLE_CORNER_DETECTION)
+    float2 leftRight = step(d.xy, d.yx);
+    float2 rounding = (1.0 - SMAA_CORNER_ROUNDING_NORM) * leftRight;
+
+    rounding /= leftRight.x + leftRight.y;
+
+    float2 factor = float2(1.0, 1.0);
+    factor.x -= rounding.x * SMAASampleLevelZeroOffset(edgesTex, texcoord.xy, int2( 1, 0)).g;
+    factor.x -= rounding.y * SMAASampleLevelZeroOffset(edgesTex, texcoord.zw, int2( 1, 1)).g;
+    factor.y -= rounding.x * SMAASampleLevelZeroOffset(edgesTex, texcoord.xy, int2(-2, 0)).g;
+    factor.y -= rounding.y * SMAASampleLevelZeroOffset(edgesTex, texcoord.zw, int2(-2, 1)).g;
+
+    weights *= saturate(factor);
+    #endif
+}
+
+//-----------------------------------------------------------------------------
+// Blending Weight Calculation Pixel Shader (Second Pass)
+
+float4 SMAABlendingWeightCalculationPS(float2 texcoord,
+                                       float2 pixcoord,
+                                       float4 offset[3],
+                                       SMAATexture2D(edgesTex),
+                                       SMAATexture2D(areaTex),
+                                       SMAATexture2D(searchTex),
+                                       float4 subsampleIndices) { // Just pass zero for SMAA 1x, see @SUBSAMPLE_INDICES.
+    float4 weights = float4(0.0, 0.0, 0.0, 0.0);
+
+    float2 e = SMAASample(edgesTex, texcoord).rg;
+
+    SMAA_BRANCH
+    if (e.g > 0.0) { // Edge at north
+        #if !defined(SMAA_DISABLE_DIAG_DETECTION)
+        // Diagonals have both north and west edges, so searching for them in
+        // one of the boundaries is enough.
+        weights.rg = SMAACalculateDiagWeights(SMAATexturePass2D(edgesTex), SMAATexturePass2D(areaTex), texcoord, e, subsampleIndices);
+
+        // We give priority to diagonals, so if we find a diagonal we skip 
+        // horizontal/vertical processing.
+        SMAA_BRANCH
+        if (weights.r == -weights.g) { // weights.r + weights.g == 0.0
+        #endif
+
+        float2 d;
+
+        // Find the distance to the left:
+        float3 coords;
+        coords.x = SMAASearchXLeft(SMAATexturePass2D(edgesTex), SMAATexturePass2D(searchTex), offset[0].xy, offset[2].x);
+        coords.y = offset[1].y; // offset[1].y = texcoord.y - 0.25 * SMAA_RT_METRICS.y (@CROSSING_OFFSET)
+        d.x = coords.x;
+
+        // Now fetch the left crossing edges, two at a time using bilinear
+        // filtering. Sampling at -0.25 (see @CROSSING_OFFSET) enables to
+        // discern what value each edge has:
+        float e1 = SMAASampleLevelZero(edgesTex, coords.xy).r;
+
+        // Find the distance to the right:
+        coords.z = SMAASearchXRight(SMAATexturePass2D(edgesTex), SMAATexturePass2D(searchTex), offset[0].zw, offset[2].y);
+        d.y = coords.z;
+
+        // We want the distances to be in pixel units (doing this here allow to
+        // better interleave arithmetic and memory accesses):
+        d = abs(round(mad(SMAA_RT_METRICS.zz, d, -pixcoord.xx)));
+
+        // SMAAArea below needs a sqrt, as the areas texture is compressed
+        // quadratically:
+        float2 sqrt_d = sqrt(d);
+
+        // Fetch the right crossing edges:
+        float e2 = SMAASampleLevelZeroOffset(edgesTex, coords.zy, int2(1, 0)).r;
+
+        // Ok, we know how this pattern looks like, now it is time for getting
+        // the actual area:
+        weights.rg = SMAAArea(SMAATexturePass2D(areaTex), sqrt_d, e1, e2, subsampleIndices.y);
+
+        // Fix corners:
+        coords.y = texcoord.y;
+        SMAADetectHorizontalCornerPattern(SMAATexturePass2D(edgesTex), weights.rg, coords.xyzy, d);
+
+        #if !defined(SMAA_DISABLE_DIAG_DETECTION)
+        } else
+            e.r = 0.0; // Skip vertical processing.
+        #endif
+    }
+
+    SMAA_BRANCH
+    if (e.r > 0.0) { // Edge at west
+        float2 d;
+
+        // Find the distance to the top:
+        float3 coords;
+        coords.y = SMAASearchYUp(SMAATexturePass2D(edgesTex), SMAATexturePass2D(searchTex), offset[1].xy, offset[2].z);
+        coords.x = offset[0].x; // offset[1].x = texcoord.x - 0.25 * SMAA_RT_METRICS.x;
+        d.x = coords.y;
+
+        // Fetch the top crossing edges:
+        float e1 = SMAASampleLevelZero(edgesTex, coords.xy).g;
+
+        // Find the distance to the bottom:
+        coords.z = SMAASearchYDown(SMAATexturePass2D(edgesTex), SMAATexturePass2D(searchTex), offset[1].zw, offset[2].w);
+        d.y = coords.z;
+
+        // We want the distances to be in pixel units:
+        d = abs(round(mad(SMAA_RT_METRICS.ww, d, -pixcoord.yy)));
+
+        // SMAAArea below needs a sqrt, as the areas texture is compressed 
+        // quadratically:
+        float2 sqrt_d = sqrt(d);
+
+        // Fetch the bottom crossing edges:
+        float e2 = SMAASampleLevelZeroOffset(edgesTex, coords.xz, int2(0, 1)).g;
+
+        // Get the area for this direction:
+        weights.ba = SMAAArea(SMAATexturePass2D(areaTex), sqrt_d, e1, e2, subsampleIndices.x);
+
+        // Fix corners:
+        coords.x = texcoord.x;
+        SMAADetectVerticalCornerPattern(SMAATexturePass2D(edgesTex), weights.ba, coords.xyxz, d);
+    }
+
+    return weights;
+}
+
+//-----------------------------------------------------------------------------
+// Neighborhood Blending Pixel Shader (Third Pass)
+
+float4 SMAANeighborhoodBlendingPS(float2 texcoord,
+                                  float4 offset,
+                                  SMAATexture2D(colorTex),
+                                  SMAATexture2D(blendTex)
+                                  #if SMAA_REPROJECTION
+                                  , SMAATexture2D(velocityTex)
+                                  #endif
+                                  ) {
+    // Fetch the blending weights for current pixel:
+    float4 a;
+    a.x = SMAASample(blendTex, offset.xy).a; // Right
+    a.y = SMAASample(blendTex, offset.zw).g; // Top
+    a.wz = SMAASample(blendTex, texcoord).xz; // Bottom / Left
+
+    // Is there any blending weight with a value greater than 0.0?
+    SMAA_BRANCH
+    if (dot(a, float4(1.0, 1.0, 1.0, 1.0)) < 1e-5) {
+        float4 color = SMAASampleLevelZero(colorTex, texcoord);
+
+        #if SMAA_REPROJECTION
+        float2 velocity = SMAA_DECODE_VELOCITY(SMAASampleLevelZero(velocityTex, texcoord));
+
+        // Pack velocity into the alpha channel:
+        color.a = sqrt(5.0 * length(velocity));
+        #endif
+
+        return color;
+    } else {
+        bool h = max(a.x, a.z) > max(a.y, a.w); // max(horizontal) > max(vertical)
+
+        // Calculate the blending offsets:
+        float4 blendingOffset = float4(0.0, a.y, 0.0, a.w);
+        float2 blendingWeight = a.yw;
+        SMAAMovc(bool4(h, h, h, h), blendingOffset, float4(a.x, 0.0, a.z, 0.0));
+        SMAAMovc(bool2(h, h), blendingWeight, a.xz);
+        blendingWeight /= dot(blendingWeight, float2(1.0, 1.0));
+
+        // Calculate the texture coordinates:
+        float4 blendingCoord = mad(blendingOffset, float4(SMAA_RT_METRICS.xy, -SMAA_RT_METRICS.xy), texcoord.xyxy);
+
+        // We exploit bilinear filtering to mix current pixel with the chosen
+        // neighbor:
+        float4 color = blendingWeight.x * SMAASampleLevelZero(colorTex, blendingCoord.xy);
+        color += blendingWeight.y * SMAASampleLevelZero(colorTex, blendingCoord.zw);
+
+        #if SMAA_REPROJECTION
+        // Antialias velocity for proper reprojection in a later stage:
+        float2 velocity = blendingWeight.x * SMAA_DECODE_VELOCITY(SMAASampleLevelZero(velocityTex, blendingCoord.xy));
+        velocity += blendingWeight.y * SMAA_DECODE_VELOCITY(SMAASampleLevelZero(velocityTex, blendingCoord.zw));
+
+        // Pack velocity into the alpha channel:
+        color.a = sqrt(5.0 * length(velocity));
+        #endif
+
+        return color;
+    }
+}
+
+//-----------------------------------------------------------------------------
+// Temporal Resolve Pixel Shader (Optional Pass)
+
+float4 SMAAResolvePS(float2 texcoord,
+                     SMAATexture2D(currentColorTex),
+                     SMAATexture2D(previousColorTex)
+                     #if SMAA_REPROJECTION
+                     , SMAATexture2D(velocityTex)
+                     #endif
+                     ) {
+    #if SMAA_REPROJECTION
+    // Velocity is assumed to be calculated for motion blur, so we need to
+    // inverse it for reprojection:
+    float2 velocity = -SMAA_DECODE_VELOCITY(SMAASamplePoint(velocityTex, texcoord).rg);
+
+    // Fetch current pixel:
+    float4 current = SMAASamplePoint(currentColorTex, texcoord);
+
+    // Reproject current coordinates and fetch previous pixel:
+    float4 previous = SMAASamplePoint(previousColorTex, texcoord + velocity);
+
+    // Attenuate the previous pixel if the velocity is different:
+    float delta = abs(current.a * current.a - previous.a * previous.a) / 5.0;
+    float weight = 0.5 * saturate(1.0 - sqrt(delta) * SMAA_REPROJECTION_WEIGHT_SCALE);
+
+    // Blend the pixels according to the calculated weight:
+    return lerp(current, previous, weight);
+    #else
+    // Just blend the pixels:
+    float4 current = SMAASamplePoint(currentColorTex, texcoord);
+    float4 previous = SMAASamplePoint(previousColorTex, texcoord);
+    return lerp(current, previous, 0.5);
+    #endif
+}
+
+//-----------------------------------------------------------------------------
+// Separate Multisamples Pixel Shader (Optional Pass)
+
+#ifdef SMAALoad
+void SMAASeparatePS(float4 position,
+                    float2 texcoord,
+                    out float4 target0,
+                    out float4 target1,
+                    SMAATexture2DMS2(colorTexMS)) {
+    int2 pos = int2(position.xy);
+    target0 = SMAALoad(colorTexMS, pos, 0);
+    target1 = SMAALoad(colorTexMS, pos, 1);
+}
+#endif
+
+//-----------------------------------------------------------------------------
+#endif // SMAA_INCLUDE_PS
diff --git a/src/Ryujinx.Graphics.OpenGL/Effects/Shaders/smaa_blend.glsl b/src/Ryujinx.Graphics.OpenGL/Effects/Shaders/smaa_blend.glsl
new file mode 100644
index 00000000..c875ce12
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Effects/Shaders/smaa_blend.glsl
@@ -0,0 +1,26 @@
+layout(rgba8, binding = 0) uniform image2D imgOutput;
+
+uniform sampler2D inputTexture;
+layout( location=0 ) uniform vec2 invResolution;
+uniform sampler2D samplerArea;
+uniform sampler2D samplerSearch;
+
+void main() {
+    ivec2 loc = ivec2(gl_GlobalInvocationID.x * 4, gl_GlobalInvocationID.y * 4);
+    for(int i = 0; i < 4; i++)
+    {
+        for(int j = 0; j < 4; j++)
+        {
+            ivec2 texelCoord = ivec2(loc.x + i, loc.y + j);
+            vec2 coord = (texelCoord + vec2(0.5)) / invResolution;
+            vec2 pixCoord;
+            vec4 offset[3];
+
+            SMAABlendingWeightCalculationVS(coord, pixCoord, offset);
+
+            vec4 oColor = SMAABlendingWeightCalculationPS(coord, pixCoord, offset, inputTexture, samplerArea, samplerSearch, ivec4(0));    
+
+            imageStore(imgOutput, texelCoord, oColor);
+        }
+    }
+}
diff --git a/src/Ryujinx.Graphics.OpenGL/Effects/Shaders/smaa_edge.glsl b/src/Ryujinx.Graphics.OpenGL/Effects/Shaders/smaa_edge.glsl
new file mode 100644
index 00000000..fd5d9715
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Effects/Shaders/smaa_edge.glsl
@@ -0,0 +1,24 @@
+layout(rgba8, binding = 0) uniform image2D imgOutput;
+
+uniform sampler2D inputTexture;
+layout( location=0 ) uniform vec2 invResolution;
+
+void main() 
+{
+    vec2 loc = ivec2(gl_GlobalInvocationID.x * 4, gl_GlobalInvocationID.y * 4);
+    for(int i = 0; i < 4; i++)
+    {
+        for(int j = 0; j < 4; j++)
+        {
+            ivec2 texelCoord = ivec2(loc.x + i, loc.y + j);
+            vec2 coord = (texelCoord + vec2(0.5)) / invResolution;
+            vec4 offset[3];
+            SMAAEdgeDetectionVS(coord, offset);
+            vec2 oColor = SMAAColorEdgeDetectionPS(coord, offset, inputTexture);
+            if (oColor != float2(-2.0, -2.0))
+            {
+                imageStore(imgOutput, texelCoord, vec4(oColor, 0.0, 1.0));
+            }
+        }
+    }
+}
\ No newline at end of file
diff --git a/src/Ryujinx.Graphics.OpenGL/Effects/Shaders/smaa_neighbour.glsl b/src/Ryujinx.Graphics.OpenGL/Effects/Shaders/smaa_neighbour.glsl
new file mode 100644
index 00000000..2e9432ae
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Effects/Shaders/smaa_neighbour.glsl
@@ -0,0 +1,26 @@
+layout(rgba8, binding = 0) uniform image2D imgOutput;
+
+uniform sampler2D inputTexture;
+layout( location=0 ) uniform vec2 invResolution;
+uniform sampler2D samplerBlend;
+
+void main() {
+    vec2 loc = ivec2(gl_GlobalInvocationID.x * 4, gl_GlobalInvocationID.y * 4);
+    for(int i = 0; i < 4; i++)
+    {
+        for(int j = 0; j < 4; j++)
+        {
+            ivec2 texelCoord = ivec2(loc.x + i, loc.y + j);
+            vec2 coord = (texelCoord + vec2(0.5)) / invResolution;
+            vec2 pixCoord;
+            vec4 offset;
+            
+            SMAANeighborhoodBlendingVS(coord, offset);
+
+            vec4 oColor = SMAANeighborhoodBlendingPS(coord, offset, inputTexture, samplerBlend);
+
+            imageStore(imgOutput,  texelCoord, oColor);
+        }
+    }
+
+}
diff --git a/src/Ryujinx.Graphics.OpenGL/Effects/SmaaPostProcessingEffect.cs b/src/Ryujinx.Graphics.OpenGL/Effects/SmaaPostProcessingEffect.cs
new file mode 100644
index 00000000..1ad300c8
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Effects/SmaaPostProcessingEffect.cs
@@ -0,0 +1,261 @@
+using OpenTK.Graphics.OpenGL;
+using Ryujinx.Common;
+using Ryujinx.Graphics.GAL;
+using Ryujinx.Graphics.OpenGL.Image;
+using System;
+
+namespace Ryujinx.Graphics.OpenGL.Effects.Smaa
+{
+    internal partial class SmaaPostProcessingEffect : IPostProcessingEffect
+    {
+        public const int AreaWidth = 160;
+        public const int AreaHeight = 560;
+        public const int SearchWidth = 64;
+        public const int SearchHeight = 16;
+
+        private readonly OpenGLRenderer _renderer;
+        private TextureStorage _outputTexture;
+        private TextureStorage _searchTexture;
+        private TextureStorage _areaTexture;
+        private int[] _edgeShaderPrograms;
+        private int[] _blendShaderPrograms;
+        private int[] _neighbourShaderPrograms;
+        private TextureStorage _edgeOutputTexture;
+        private TextureStorage _blendOutputTexture;
+        private string[] _qualities;
+        private int _inputUniform;
+        private int _outputUniform;
+        private int _samplerAreaUniform;
+        private int _samplerSearchUniform;
+        private int _samplerBlendUniform;
+        private int _resolutionUniform;
+        private int _quality = 1;
+
+        public int Quality
+        {
+            get => _quality; set
+            {
+                _quality = Math.Clamp(value, 0, _qualities.Length - 1);
+            }
+        }
+        public SmaaPostProcessingEffect(OpenGLRenderer renderer, int quality)
+        {
+            _renderer = renderer;
+
+            _edgeShaderPrograms = Array.Empty<int>();
+            _blendShaderPrograms = Array.Empty<int>();
+            _neighbourShaderPrograms = Array.Empty<int>();
+
+            _qualities = new string[] { "SMAA_PRESET_LOW", "SMAA_PRESET_MEDIUM", "SMAA_PRESET_HIGH", "SMAA_PRESET_ULTRA" };
+
+            Quality = quality;
+
+            Initialize();
+        }
+
+        public void Dispose()
+        {
+            _searchTexture?.Dispose();
+            _areaTexture?.Dispose();
+            _outputTexture?.Dispose();
+            _edgeOutputTexture?.Dispose();
+            _blendOutputTexture?.Dispose();
+
+            DeleteShaders();
+        }
+
+        private void DeleteShaders()
+        {
+            for (int i = 0; i < _edgeShaderPrograms.Length; i++)
+            {
+                GL.DeleteProgram(_edgeShaderPrograms[i]);
+                GL.DeleteProgram(_blendShaderPrograms[i]);
+                GL.DeleteProgram(_neighbourShaderPrograms[i]);
+            }
+        }
+
+        private unsafe void RecreateShaders(int width, int height)
+        {
+            string baseShader = EmbeddedResources.ReadAllText("Ryujinx.Graphics.OpenGL/Effects/Shaders/smaa.hlsl");
+            var pixelSizeDefine = $"#define SMAA_RT_METRICS float4(1.0 / {width}.0, 1.0 / {height}.0, {width}, {height}) \n";
+
+            _edgeShaderPrograms = new int[_qualities.Length];
+            _blendShaderPrograms = new int[_qualities.Length];
+            _neighbourShaderPrograms = new int[_qualities.Length];
+
+            for (int i = 0; i < +_edgeShaderPrograms.Length; i++)
+            {
+                var presets = $"#version 430 core \n#define {_qualities[i]} 1 \n{pixelSizeDefine}#define SMAA_GLSL_4 1 \nlayout (local_size_x = 16, local_size_y = 16) in;\n{baseShader}";
+
+                var edgeShaderData = EmbeddedResources.ReadAllText("Ryujinx.Graphics.OpenGL/Effects/Shaders/smaa_edge.glsl");
+                var blendShaderData = EmbeddedResources.ReadAllText("Ryujinx.Graphics.OpenGL/Effects/Shaders/smaa_blend.glsl");
+                var neighbourShaderData = EmbeddedResources.ReadAllText("Ryujinx.Graphics.OpenGL/Effects/Shaders/smaa_neighbour.glsl");
+
+                var shaders = new string[] { presets, edgeShaderData };
+                var edgeProgram = ShaderHelper.CompileProgram(shaders, ShaderType.ComputeShader);
+
+                shaders[1] = blendShaderData;
+                var blendProgram = ShaderHelper.CompileProgram(shaders, ShaderType.ComputeShader);
+
+                shaders[1] = neighbourShaderData;
+                var neighbourProgram = ShaderHelper.CompileProgram(shaders, ShaderType.ComputeShader);
+
+                _edgeShaderPrograms[i] = edgeProgram;
+                _blendShaderPrograms[i] = blendProgram;
+                _neighbourShaderPrograms[i] = neighbourProgram;
+            }
+
+            _inputUniform = GL.GetUniformLocation(_edgeShaderPrograms[0], "inputTexture");
+            _outputUniform = GL.GetUniformLocation(_edgeShaderPrograms[0], "imgOutput");
+            _samplerAreaUniform = GL.GetUniformLocation(_blendShaderPrograms[0], "samplerArea");
+            _samplerSearchUniform = GL.GetUniformLocation(_blendShaderPrograms[0], "samplerSearch");
+            _samplerBlendUniform = GL.GetUniformLocation(_neighbourShaderPrograms[0], "samplerBlend");
+            _resolutionUniform = GL.GetUniformLocation(_edgeShaderPrograms[0], "invResolution");
+        }
+
+        private void Initialize()
+        {
+            var areaInfo = new TextureCreateInfo(AreaWidth,
+                AreaHeight,
+                1,
+                1,
+                1,
+                1,
+                1,
+                1,
+                Format.R8G8Unorm,
+                DepthStencilMode.Depth,
+                Target.Texture2D,
+                SwizzleComponent.Red,
+                SwizzleComponent.Green,
+                SwizzleComponent.Blue,
+                SwizzleComponent.Alpha);
+
+            var searchInfo = new TextureCreateInfo(SearchWidth,
+                SearchHeight,
+                1,
+                1,
+                1,
+                1,
+                1,
+                1,
+                Format.R8Unorm,
+                DepthStencilMode.Depth,
+                Target.Texture2D,
+                SwizzleComponent.Red,
+                SwizzleComponent.Green,
+                SwizzleComponent.Blue,
+                SwizzleComponent.Alpha);
+
+            _areaTexture = new TextureStorage(_renderer, areaInfo, 1);
+            _searchTexture = new TextureStorage(_renderer, searchInfo, 1);
+
+            var areaTexture = EmbeddedResources.Read("Ryujinx.Graphics.OpenGL/Effects/Textures/SmaaAreaTexture.bin");
+            var searchTexture = EmbeddedResources.Read("Ryujinx.Graphics.OpenGL/Effects/Textures/SmaaSearchTexture.bin");
+
+            var areaView = _areaTexture.CreateDefaultView();
+            var searchView = _searchTexture.CreateDefaultView();
+
+            areaView.SetData(areaTexture);
+            searchView.SetData(searchTexture);
+        }
+
+        public TextureView Run(TextureView view, int width, int height)
+        {
+            if (_outputTexture == null || _outputTexture.Info.Width != view.Width || _outputTexture.Info.Height != view.Height)
+            {
+                _outputTexture?.Dispose();
+                _outputTexture = new TextureStorage(_renderer, view.Info, view.ScaleFactor);
+                _outputTexture.CreateDefaultView();
+                _edgeOutputTexture = new TextureStorage(_renderer, view.Info, view.ScaleFactor);
+                _edgeOutputTexture.CreateDefaultView();
+                _blendOutputTexture = new TextureStorage(_renderer, view.Info, view.ScaleFactor);
+                _blendOutputTexture.CreateDefaultView();
+
+                DeleteShaders();
+
+                RecreateShaders(view.Width, view.Height);
+            }
+
+            var textureView = _outputTexture.CreateView(view.Info, 0, 0) as TextureView;
+            var edgeOutput = _edgeOutputTexture.DefaultView as TextureView;
+            var blendOutput = _blendOutputTexture.DefaultView as TextureView;
+            var areaTexture = _areaTexture.DefaultView as TextureView;
+            var searchTexture = _searchTexture.DefaultView as TextureView;
+
+            var previousFramebuffer = GL.GetInteger(GetPName.FramebufferBinding);
+            int previousUnit = GL.GetInteger(GetPName.ActiveTexture);
+            GL.ActiveTexture(TextureUnit.Texture0);
+            int previousTextureBinding0 = GL.GetInteger(GetPName.TextureBinding2D);
+            GL.ActiveTexture(TextureUnit.Texture1);
+            int previousTextureBinding1 = GL.GetInteger(GetPName.TextureBinding2D);
+            GL.ActiveTexture(TextureUnit.Texture2);
+            int previousTextureBinding2 = GL.GetInteger(GetPName.TextureBinding2D);
+
+            var framebuffer = new Framebuffer();
+            framebuffer.Bind();
+            framebuffer.AttachColor(0, edgeOutput);
+            GL.Clear(ClearBufferMask.ColorBufferBit);
+            GL.ClearColor(0, 0, 0, 0);
+            framebuffer.AttachColor(0, blendOutput);
+            GL.Clear(ClearBufferMask.ColorBufferBit);
+            GL.ClearColor(0, 0, 0, 0);
+
+            GL.BindFramebuffer(FramebufferTarget.Framebuffer, previousFramebuffer);
+
+            framebuffer.Dispose();
+
+            var dispatchX = BitUtils.DivRoundUp(view.Width, IPostProcessingEffect.LocalGroupSize);
+            var dispatchY = BitUtils.DivRoundUp(view.Height, IPostProcessingEffect.LocalGroupSize);
+
+            int previousProgram = GL.GetInteger(GetPName.CurrentProgram);
+            GL.BindImageTexture(0, edgeOutput.Handle, 0, false, 0, TextureAccess.ReadWrite, SizedInternalFormat.Rgba8);
+            GL.UseProgram(_edgeShaderPrograms[Quality]);
+            view.Bind(0);
+            GL.Uniform1(_inputUniform, 0);
+            GL.Uniform1(_outputUniform, 0);
+            GL.Uniform2(_resolutionUniform, (float)view.Width, (float)view.Height);
+            GL.DispatchCompute(dispatchX, dispatchY, 1);
+            GL.MemoryBarrier(MemoryBarrierFlags.ShaderImageAccessBarrierBit);
+
+            GL.BindImageTexture(0, blendOutput.Handle, 0, false, 0, TextureAccess.ReadWrite, SizedInternalFormat.Rgba8);
+            GL.UseProgram(_blendShaderPrograms[Quality]);
+            edgeOutput.Bind(0);
+            areaTexture.Bind(1);
+            searchTexture.Bind(2);
+            GL.Uniform1(_inputUniform, 0);
+            GL.Uniform1(_outputUniform, 0);
+            GL.Uniform1(_samplerAreaUniform, 1);
+            GL.Uniform1(_samplerSearchUniform, 2);
+            GL.Uniform2(_resolutionUniform, (float)view.Width, (float)view.Height);
+            GL.DispatchCompute(dispatchX, dispatchY, 1);
+            GL.MemoryBarrier(MemoryBarrierFlags.ShaderImageAccessBarrierBit);
+
+            GL.BindImageTexture(0, textureView.Handle, 0, false, 0, TextureAccess.ReadWrite, SizedInternalFormat.Rgba8);
+            GL.UseProgram(_neighbourShaderPrograms[Quality]);
+            view.Bind(0);
+            blendOutput.Bind(1);
+            GL.Uniform1(_inputUniform, 0);
+            GL.Uniform1(_outputUniform, 0);
+            GL.Uniform1(_samplerBlendUniform, 1);
+            GL.Uniform2(_resolutionUniform, (float)view.Width, (float)view.Height);
+            GL.DispatchCompute(dispatchX, dispatchY, 1);
+            GL.MemoryBarrier(MemoryBarrierFlags.ShaderImageAccessBarrierBit);
+
+            (_renderer.Pipeline as Pipeline).RestoreImages1And2();
+
+            GL.UseProgram(previousProgram);
+
+            GL.ActiveTexture(TextureUnit.Texture0);
+            GL.BindTexture(TextureTarget.Texture2D, previousTextureBinding0);
+            GL.ActiveTexture(TextureUnit.Texture1);
+            GL.BindTexture(TextureTarget.Texture2D, previousTextureBinding1);
+            GL.ActiveTexture(TextureUnit.Texture2);
+            GL.BindTexture(TextureTarget.Texture2D, previousTextureBinding2);
+
+            GL.ActiveTexture((TextureUnit)previousUnit);
+
+            return textureView;
+        }
+    }
+}
diff --git a/src/Ryujinx.Graphics.OpenGL/Effects/Textures/SmaaAreaTexture.bin b/src/Ryujinx.Graphics.OpenGL/Effects/Textures/SmaaAreaTexture.bin
new file mode 100644
index 00000000..f4a7a1b4
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Effects/Textures/SmaaAreaTexture.bin
diff --git a/src/Ryujinx.Graphics.OpenGL/Effects/Textures/SmaaSearchTexture.bin b/src/Ryujinx.Graphics.OpenGL/Effects/Textures/SmaaSearchTexture.bin
new file mode 100644
index 00000000..db5bf73f
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Effects/Textures/SmaaSearchTexture.bin
diff --git a/src/Ryujinx.Graphics.OpenGL/EnumConversion.cs b/src/Ryujinx.Graphics.OpenGL/EnumConversion.cs
new file mode 100644
index 00000000..c9a1eaa9
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/EnumConversion.cs
@@ -0,0 +1,675 @@
+using OpenTK.Graphics.OpenGL;
+using Ryujinx.Common.Logging;
+using Ryujinx.Graphics.GAL;
+using Ryujinx.Graphics.Shader;
+
+namespace Ryujinx.Graphics.OpenGL
+{
+    static class EnumConversion
+    {
+        public static TextureWrapMode Convert(this AddressMode mode)
+        {
+            switch (mode)
+            {
+                case AddressMode.Clamp:
+                    return TextureWrapMode.Clamp;
+                case AddressMode.Repeat:
+                    return TextureWrapMode.Repeat;
+                case AddressMode.MirrorClamp:
+                    return (TextureWrapMode)ExtTextureMirrorClamp.MirrorClampExt;
+                case AddressMode.MirrorClampToEdge:
+                    return (TextureWrapMode)ExtTextureMirrorClamp.MirrorClampToEdgeExt;
+                case AddressMode.MirrorClampToBorder:
+                    return (TextureWrapMode)ExtTextureMirrorClamp.MirrorClampToBorderExt;
+                case AddressMode.ClampToBorder:
+                    return TextureWrapMode.ClampToBorder;
+                case AddressMode.MirroredRepeat:
+                    return TextureWrapMode.MirroredRepeat;
+                case AddressMode.ClampToEdge:
+                    return TextureWrapMode.ClampToEdge;
+            }
+
+            Logger.Debug?.Print(LogClass.Gpu, $"Invalid {nameof(AddressMode)} enum value: {mode}.");
+
+            return TextureWrapMode.Clamp;
+        }
+
+        public static NvBlendEquationAdvanced Convert(this AdvancedBlendOp op)
+        {
+            switch (op)
+            {
+                case AdvancedBlendOp.Zero:
+                    return NvBlendEquationAdvanced.Zero;
+                case AdvancedBlendOp.Src:
+                    return NvBlendEquationAdvanced.SrcNv;
+                case AdvancedBlendOp.Dst:
+                    return NvBlendEquationAdvanced.DstNv;
+                case AdvancedBlendOp.SrcOver:
+                    return NvBlendEquationAdvanced.SrcOverNv;
+                case AdvancedBlendOp.DstOver:
+                    return NvBlendEquationAdvanced.DstOverNv;
+                case AdvancedBlendOp.SrcIn:
+                    return NvBlendEquationAdvanced.SrcInNv;
+                case AdvancedBlendOp.DstIn:
+                    return NvBlendEquationAdvanced.DstInNv;
+                case AdvancedBlendOp.SrcOut:
+                    return NvBlendEquationAdvanced.SrcOutNv;
+                case AdvancedBlendOp.DstOut:
+                    return NvBlendEquationAdvanced.DstOutNv;
+                case AdvancedBlendOp.SrcAtop:
+                    return NvBlendEquationAdvanced.SrcAtopNv;
+                case AdvancedBlendOp.DstAtop:
+                    return NvBlendEquationAdvanced.DstAtopNv;
+                case AdvancedBlendOp.Xor:
+                    return NvBlendEquationAdvanced.XorNv;
+                case AdvancedBlendOp.Plus:
+                    return NvBlendEquationAdvanced.PlusNv;
+                case AdvancedBlendOp.PlusClamped:
+                    return NvBlendEquationAdvanced.PlusClampedNv;
+                case AdvancedBlendOp.PlusClampedAlpha:
+                    return NvBlendEquationAdvanced.PlusClampedAlphaNv;
+                case AdvancedBlendOp.PlusDarker:
+                    return NvBlendEquationAdvanced.PlusDarkerNv;
+                case AdvancedBlendOp.Multiply:
+                    return NvBlendEquationAdvanced.MultiplyNv;
+                case AdvancedBlendOp.Screen:
+                    return NvBlendEquationAdvanced.ScreenNv;
+                case AdvancedBlendOp.Overlay:
+                    return NvBlendEquationAdvanced.OverlayNv;
+                case AdvancedBlendOp.Darken:
+                    return NvBlendEquationAdvanced.DarkenNv;
+                case AdvancedBlendOp.Lighten:
+                    return NvBlendEquationAdvanced.LightenNv;
+                case AdvancedBlendOp.ColorDodge:
+                    return NvBlendEquationAdvanced.ColordodgeNv;
+                case AdvancedBlendOp.ColorBurn:
+                    return NvBlendEquationAdvanced.ColorburnNv;
+                case AdvancedBlendOp.HardLight:
+                    return NvBlendEquationAdvanced.HardlightNv;
+                case AdvancedBlendOp.SoftLight:
+                    return NvBlendEquationAdvanced.SoftlightNv;
+                case AdvancedBlendOp.Difference:
+                    return NvBlendEquationAdvanced.DifferenceNv;
+                case AdvancedBlendOp.Minus:
+                    return NvBlendEquationAdvanced.MinusNv;
+                case AdvancedBlendOp.MinusClamped:
+                    return NvBlendEquationAdvanced.MinusClampedNv;
+                case AdvancedBlendOp.Exclusion:
+                    return NvBlendEquationAdvanced.ExclusionNv;
+                case AdvancedBlendOp.Contrast:
+                    return NvBlendEquationAdvanced.ContrastNv;
+                case AdvancedBlendOp.Invert:
+                    return NvBlendEquationAdvanced.Invert;
+                case AdvancedBlendOp.InvertRGB:
+                    return NvBlendEquationAdvanced.InvertRgbNv;
+                case AdvancedBlendOp.InvertOvg:
+                    return NvBlendEquationAdvanced.InvertOvgNv;
+                case AdvancedBlendOp.LinearDodge:
+                    return NvBlendEquationAdvanced.LineardodgeNv;
+                case AdvancedBlendOp.LinearBurn:
+                    return NvBlendEquationAdvanced.LinearburnNv;
+                case AdvancedBlendOp.VividLight:
+                    return NvBlendEquationAdvanced.VividlightNv;
+                case AdvancedBlendOp.LinearLight:
+                    return NvBlendEquationAdvanced.LinearlightNv;
+                case AdvancedBlendOp.PinLight:
+                    return NvBlendEquationAdvanced.PinlightNv;
+                case AdvancedBlendOp.HardMix:
+                    return NvBlendEquationAdvanced.HardmixNv;
+                case AdvancedBlendOp.Red:
+                    return NvBlendEquationAdvanced.RedNv;
+                case AdvancedBlendOp.Green:
+                    return NvBlendEquationAdvanced.GreenNv;
+                case AdvancedBlendOp.Blue:
+                    return NvBlendEquationAdvanced.BlueNv;
+                case AdvancedBlendOp.HslHue:
+                    return NvBlendEquationAdvanced.HslHueNv;
+                case AdvancedBlendOp.HslSaturation:
+                    return NvBlendEquationAdvanced.HslSaturationNv;
+                case AdvancedBlendOp.HslColor:
+                    return NvBlendEquationAdvanced.HslColorNv;
+                case AdvancedBlendOp.HslLuminosity:
+                    return NvBlendEquationAdvanced.HslLuminosityNv;
+            }
+
+            Logger.Debug?.Print(LogClass.Gpu, $"Invalid {nameof(AdvancedBlendOp)} enum value: {op}.");
+
+            return NvBlendEquationAdvanced.Zero;
+        }
+
+        public static All Convert(this AdvancedBlendOverlap overlap)
+        {
+            switch (overlap)
+            {
+                case AdvancedBlendOverlap.Uncorrelated:
+                    return All.UncorrelatedNv;
+                case AdvancedBlendOverlap.Disjoint:
+                    return All.DisjointNv;
+                case AdvancedBlendOverlap.Conjoint:
+                    return All.ConjointNv;
+            }
+
+            Logger.Debug?.Print(LogClass.Gpu, $"Invalid {nameof(AdvancedBlendOverlap)} enum value: {overlap}.");
+
+            return All.UncorrelatedNv;
+        }
+
+        public static All Convert(this BlendFactor factor)
+        {
+            switch (factor)
+            {
+                case BlendFactor.Zero:
+                case BlendFactor.ZeroGl:
+                    return All.Zero;
+                case BlendFactor.One:
+                case BlendFactor.OneGl:
+                    return All.One;
+                case BlendFactor.SrcColor:
+                case BlendFactor.SrcColorGl:
+                    return All.SrcColor;
+                case BlendFactor.OneMinusSrcColor:
+                case BlendFactor.OneMinusSrcColorGl:
+                    return All.OneMinusSrcColor;
+                case BlendFactor.SrcAlpha:
+                case BlendFactor.SrcAlphaGl:
+                    return All.SrcAlpha;
+                case BlendFactor.OneMinusSrcAlpha:
+                case BlendFactor.OneMinusSrcAlphaGl:
+                    return All.OneMinusSrcAlpha;
+                case BlendFactor.DstAlpha:
+                case BlendFactor.DstAlphaGl:
+                    return All.DstAlpha;
+                case BlendFactor.OneMinusDstAlpha:
+                case BlendFactor.OneMinusDstAlphaGl:
+                    return All.OneMinusDstAlpha;
+                case BlendFactor.DstColor:
+                case BlendFactor.DstColorGl:
+                    return All.DstColor;
+                case BlendFactor.OneMinusDstColor:
+                case BlendFactor.OneMinusDstColorGl:
+                    return All.OneMinusDstColor;
+                case BlendFactor.SrcAlphaSaturate:
+                case BlendFactor.SrcAlphaSaturateGl:
+                    return All.SrcAlphaSaturate;
+                case BlendFactor.Src1Color:
+                case BlendFactor.Src1ColorGl:
+                    return All.Src1Color;
+                case BlendFactor.OneMinusSrc1Color:
+                case BlendFactor.OneMinusSrc1ColorGl:
+                    return All.OneMinusSrc1Color;
+                case BlendFactor.Src1Alpha:
+                case BlendFactor.Src1AlphaGl:
+                    return All.Src1Alpha;
+                case BlendFactor.OneMinusSrc1Alpha:
+                case BlendFactor.OneMinusSrc1AlphaGl:
+                    return All.OneMinusSrc1Alpha;
+                case BlendFactor.ConstantColor:
+                    return All.ConstantColor;
+                case BlendFactor.OneMinusConstantColor:
+                    return All.OneMinusConstantColor;
+                case BlendFactor.ConstantAlpha:
+                    return All.ConstantAlpha;
+                case BlendFactor.OneMinusConstantAlpha:
+                    return All.OneMinusConstantAlpha;
+            }
+
+            Logger.Debug?.Print(LogClass.Gpu, $"Invalid {nameof(BlendFactor)} enum value: {factor}.");
+
+            return All.Zero;
+        }
+
+        public static BlendEquationMode Convert(this BlendOp op)
+        {
+            switch (op)
+            {
+                case BlendOp.Add:
+                case BlendOp.AddGl:
+                    return BlendEquationMode.FuncAdd;
+                case BlendOp.Minimum:
+                case BlendOp.MinimumGl:
+                    return BlendEquationMode.Min;
+                case BlendOp.Maximum:
+                case BlendOp.MaximumGl:
+                    return BlendEquationMode.Max;
+                case BlendOp.Subtract:
+                case BlendOp.SubtractGl:
+                    return BlendEquationMode.FuncSubtract;
+                case BlendOp.ReverseSubtract:
+                case BlendOp.ReverseSubtractGl:
+                    return BlendEquationMode.FuncReverseSubtract;
+            }
+
+            Logger.Debug?.Print(LogClass.Gpu, $"Invalid {nameof(BlendOp)} enum value: {op}.");
+
+            return BlendEquationMode.FuncAdd;
+        }
+
+        public static TextureCompareMode Convert(this CompareMode mode)
+        {
+            switch (mode)
+            {
+                case CompareMode.None:
+                    return TextureCompareMode.None;
+                case CompareMode.CompareRToTexture:
+                    return TextureCompareMode.CompareRToTexture;
+            }
+
+            Logger.Debug?.Print(LogClass.Gpu, $"Invalid {nameof(CompareMode)} enum value: {mode}.");
+
+            return TextureCompareMode.None;
+        }
+
+        public static All Convert(this CompareOp op)
+        {
+            switch (op)
+            {
+                case CompareOp.Never:
+                case CompareOp.NeverGl:
+                    return All.Never;
+                case CompareOp.Less:
+                case CompareOp.LessGl:
+                    return All.Less;
+                case CompareOp.Equal:
+                case CompareOp.EqualGl:
+                    return All.Equal;
+                case CompareOp.LessOrEqual:
+                case CompareOp.LessOrEqualGl:
+                    return All.Lequal;
+                case CompareOp.Greater:
+                case CompareOp.GreaterGl:
+                    return All.Greater;
+                case CompareOp.NotEqual:
+                case CompareOp.NotEqualGl:
+                    return All.Notequal;
+                case CompareOp.GreaterOrEqual:
+                case CompareOp.GreaterOrEqualGl:
+                    return All.Gequal;
+                case CompareOp.Always:
+                case CompareOp.AlwaysGl:
+                    return All.Always;
+            }
+
+            Logger.Debug?.Print(LogClass.Gpu, $"Invalid {nameof(CompareOp)} enum value: {op}.");
+
+            return All.Never;
+        }
+
+        public static ClipDepthMode Convert(this DepthMode mode)
+        {
+            switch (mode)
+            {
+                case DepthMode.MinusOneToOne:
+                    return ClipDepthMode.NegativeOneToOne;
+                case DepthMode.ZeroToOne:
+                    return ClipDepthMode.ZeroToOne;
+            }
+
+            Logger.Debug?.Print(LogClass.Gpu, $"Invalid {nameof(DepthMode)} enum value: {mode}.");
+
+            return ClipDepthMode.NegativeOneToOne;
+        }
+
+        public static All Convert(this DepthStencilMode mode)
+        {
+            switch (mode)
+            {
+                case DepthStencilMode.Depth:
+                    return All.DepthComponent;
+                case DepthStencilMode.Stencil:
+                    return All.StencilIndex;
+            }
+
+            Logger.Debug?.Print(LogClass.Gpu, $"Invalid {nameof(DepthStencilMode)} enum value: {mode}.");
+
+            return All.Depth;
+        }
+
+        public static CullFaceMode Convert(this Face face)
+        {
+            switch (face)
+            {
+                case Face.Back:
+                    return CullFaceMode.Back;
+                case Face.Front:
+                    return CullFaceMode.Front;
+                case Face.FrontAndBack:
+                    return CullFaceMode.FrontAndBack;
+            }
+
+            Logger.Debug?.Print(LogClass.Gpu, $"Invalid {nameof(Face)} enum value: {face}.");
+
+            return CullFaceMode.Back;
+        }
+
+        public static FrontFaceDirection Convert(this FrontFace frontFace)
+        {
+            switch (frontFace)
+            {
+                case FrontFace.Clockwise:
+                    return FrontFaceDirection.Cw;
+                case FrontFace.CounterClockwise:
+                    return FrontFaceDirection.Ccw;
+            }
+
+            Logger.Debug?.Print(LogClass.Gpu, $"Invalid {nameof(FrontFace)} enum value: {frontFace}.");
+
+            return FrontFaceDirection.Cw;
+        }
+
+        public static DrawElementsType Convert(this IndexType type)
+        {
+            switch (type)
+            {
+                case IndexType.UByte:
+                    return DrawElementsType.UnsignedByte;
+                case IndexType.UShort:
+                    return DrawElementsType.UnsignedShort;
+                case IndexType.UInt:
+                    return DrawElementsType.UnsignedInt;
+            }
+
+            Logger.Debug?.Print(LogClass.Gpu, $"Invalid {nameof(IndexType)} enum value: {type}.");
+
+            return DrawElementsType.UnsignedByte;
+        }
+
+        public static TextureMagFilter Convert(this MagFilter filter)
+        {
+            switch (filter)
+            {
+                case MagFilter.Nearest:
+                    return TextureMagFilter.Nearest;
+                case MagFilter.Linear:
+                    return TextureMagFilter.Linear;
+            }
+
+            Logger.Debug?.Print(LogClass.Gpu, $"Invalid {nameof(MagFilter)} enum value: {filter}.");
+
+            return TextureMagFilter.Nearest;
+        }
+
+        public static TextureMinFilter Convert(this MinFilter filter)
+        {
+            switch (filter)
+            {
+                case MinFilter.Nearest:
+                    return TextureMinFilter.Nearest;
+                case MinFilter.Linear:
+                    return TextureMinFilter.Linear;
+                case MinFilter.NearestMipmapNearest:
+                    return TextureMinFilter.NearestMipmapNearest;
+                case MinFilter.LinearMipmapNearest:
+                    return TextureMinFilter.LinearMipmapNearest;
+                case MinFilter.NearestMipmapLinear:
+                    return TextureMinFilter.NearestMipmapLinear;
+                case MinFilter.LinearMipmapLinear:
+                    return TextureMinFilter.LinearMipmapLinear;
+            }
+
+            Logger.Debug?.Print(LogClass.Gpu, $"Invalid {nameof(MinFilter)} enum value: {filter}.");
+
+            return TextureMinFilter.Nearest;
+        }
+
+        public static OpenTK.Graphics.OpenGL.PolygonMode Convert(this GAL.PolygonMode mode)
+        {
+            switch (mode)
+            {
+                case GAL.PolygonMode.Point:
+                    return OpenTK.Graphics.OpenGL.PolygonMode.Point;
+                case GAL.PolygonMode.Line:
+                    return OpenTK.Graphics.OpenGL.PolygonMode.Line;
+                case GAL.PolygonMode.Fill:
+                    return OpenTK.Graphics.OpenGL.PolygonMode.Fill;
+            }
+
+            Logger.Debug?.Print(LogClass.Gpu, $"Invalid {nameof(GAL.PolygonMode)} enum value: {mode}.");
+
+            return OpenTK.Graphics.OpenGL.PolygonMode.Fill;
+        }
+
+        public static PrimitiveType Convert(this PrimitiveTopology topology)
+        {
+            switch (topology)
+            {
+                case PrimitiveTopology.Points:
+                    return PrimitiveType.Points;
+                case PrimitiveTopology.Lines:
+                    return PrimitiveType.Lines;
+                case PrimitiveTopology.LineLoop:
+                    return PrimitiveType.LineLoop;
+                case PrimitiveTopology.LineStrip:
+                    return PrimitiveType.LineStrip;
+                case PrimitiveTopology.Triangles:
+                    return PrimitiveType.Triangles;
+                case PrimitiveTopology.TriangleStrip:
+                    return PrimitiveType.TriangleStrip;
+                case PrimitiveTopology.TriangleFan:
+                    return PrimitiveType.TriangleFan;
+                case PrimitiveTopology.Quads:
+                    return PrimitiveType.Quads;
+                case PrimitiveTopology.QuadStrip:
+                    return PrimitiveType.QuadStrip;
+                case PrimitiveTopology.Polygon:
+                    return PrimitiveType.TriangleFan;
+                case PrimitiveTopology.LinesAdjacency:
+                    return PrimitiveType.LinesAdjacency;
+                case PrimitiveTopology.LineStripAdjacency:
+                    return PrimitiveType.LineStripAdjacency;
+                case PrimitiveTopology.TrianglesAdjacency:
+                    return PrimitiveType.TrianglesAdjacency;
+                case PrimitiveTopology.TriangleStripAdjacency:
+                    return PrimitiveType.TriangleStripAdjacency;
+                case PrimitiveTopology.Patches:
+                    return PrimitiveType.Patches;
+            }
+
+            Logger.Debug?.Print(LogClass.Gpu, $"Invalid {nameof(PrimitiveTopology)} enum value: {topology}.");
+
+            return PrimitiveType.Points;
+        }
+
+        public static TransformFeedbackPrimitiveType ConvertToTfType(this PrimitiveTopology topology)
+        {
+            switch (topology)
+            {
+                case PrimitiveTopology.Points:
+                    return TransformFeedbackPrimitiveType.Points;
+                case PrimitiveTopology.Lines:
+                case PrimitiveTopology.LineLoop:
+                case PrimitiveTopology.LineStrip:
+                case PrimitiveTopology.LinesAdjacency:
+                case PrimitiveTopology.LineStripAdjacency:
+                    return TransformFeedbackPrimitiveType.Lines;
+                case PrimitiveTopology.Triangles:
+                case PrimitiveTopology.TriangleStrip:
+                case PrimitiveTopology.TriangleFan:
+                case PrimitiveTopology.TrianglesAdjacency:
+                case PrimitiveTopology.TriangleStripAdjacency:
+                    return TransformFeedbackPrimitiveType.Triangles;
+            }
+
+            Logger.Debug?.Print(LogClass.Gpu, $"Invalid {nameof(PrimitiveTopology)} enum value: {topology}.");
+
+            return TransformFeedbackPrimitiveType.Points;
+        }
+
+        public static OpenTK.Graphics.OpenGL.StencilOp Convert(this GAL.StencilOp op)
+        {
+            switch (op)
+            {
+                case GAL.StencilOp.Keep:
+                case GAL.StencilOp.KeepGl:
+                    return OpenTK.Graphics.OpenGL.StencilOp.Keep;
+                case GAL.StencilOp.Zero:
+                case GAL.StencilOp.ZeroGl:
+                    return OpenTK.Graphics.OpenGL.StencilOp.Zero;
+                case GAL.StencilOp.Replace:
+                case GAL.StencilOp.ReplaceGl:
+                    return OpenTK.Graphics.OpenGL.StencilOp.Replace;
+                case GAL.StencilOp.IncrementAndClamp:
+                case GAL.StencilOp.IncrementAndClampGl:
+                    return OpenTK.Graphics.OpenGL.StencilOp.Incr;
+                case GAL.StencilOp.DecrementAndClamp:
+                case GAL.StencilOp.DecrementAndClampGl:
+                    return OpenTK.Graphics.OpenGL.StencilOp.Decr;
+                case GAL.StencilOp.Invert:
+                case GAL.StencilOp.InvertGl:
+                    return OpenTK.Graphics.OpenGL.StencilOp.Invert;
+                case GAL.StencilOp.IncrementAndWrap:
+                case GAL.StencilOp.IncrementAndWrapGl:
+                    return OpenTK.Graphics.OpenGL.StencilOp.IncrWrap;
+                case GAL.StencilOp.DecrementAndWrap:
+                case GAL.StencilOp.DecrementAndWrapGl:
+                    return OpenTK.Graphics.OpenGL.StencilOp.DecrWrap;
+            }
+
+            Logger.Debug?.Print(LogClass.Gpu, $"Invalid {nameof(GAL.StencilOp)} enum value: {op}.");
+
+            return OpenTK.Graphics.OpenGL.StencilOp.Keep;
+        }
+
+        public static All Convert(this SwizzleComponent swizzleComponent)
+        {
+            switch (swizzleComponent)
+            {
+                case SwizzleComponent.Zero:
+                    return All.Zero;
+                case SwizzleComponent.One:
+                    return All.One;
+                case SwizzleComponent.Red:
+                    return All.Red;
+                case SwizzleComponent.Green:
+                    return All.Green;
+                case SwizzleComponent.Blue:
+                    return All.Blue;
+                case SwizzleComponent.Alpha:
+                    return All.Alpha;
+            }
+
+            Logger.Debug?.Print(LogClass.Gpu, $"Invalid {nameof(SwizzleComponent)} enum value: {swizzleComponent}.");
+
+            return All.Zero;
+        }
+
+        public static ImageTarget ConvertToImageTarget(this Target target)
+        {
+            return (ImageTarget)target.Convert();
+        }
+
+        public static TextureTarget Convert(this Target target)
+        {
+            switch (target)
+            {
+                case Target.Texture1D:
+                    return TextureTarget.Texture1D;
+                case Target.Texture2D:
+                    return TextureTarget.Texture2D;
+                case Target.Texture3D:
+                    return TextureTarget.Texture3D;
+                case Target.Texture1DArray:
+                    return TextureTarget.Texture1DArray;
+                case Target.Texture2DArray:
+                    return TextureTarget.Texture2DArray;
+                case Target.Texture2DMultisample:
+                    return TextureTarget.Texture2DMultisample;
+                case Target.Texture2DMultisampleArray:
+                    return TextureTarget.Texture2DMultisampleArray;
+                case Target.Cubemap:
+                    return TextureTarget.TextureCubeMap;
+                case Target.CubemapArray:
+                    return TextureTarget.TextureCubeMapArray;
+                case Target.TextureBuffer:
+                    return TextureTarget.TextureBuffer;
+            }
+
+            Logger.Debug?.Print(LogClass.Gpu, $"Invalid {nameof(Target)} enum value: {target}.");
+
+            return TextureTarget.Texture2D;
+        }
+
+        public static NvViewportSwizzle Convert(this ViewportSwizzle swizzle)
+        {
+            switch (swizzle)
+            {
+                case ViewportSwizzle.PositiveX:
+                    return NvViewportSwizzle.ViewportSwizzlePositiveXNv;
+                case ViewportSwizzle.PositiveY:
+                    return NvViewportSwizzle.ViewportSwizzlePositiveYNv;
+                case ViewportSwizzle.PositiveZ:
+                    return NvViewportSwizzle.ViewportSwizzlePositiveZNv;
+                case ViewportSwizzle.PositiveW:
+                    return NvViewportSwizzle.ViewportSwizzlePositiveWNv;
+                case ViewportSwizzle.NegativeX:
+                    return NvViewportSwizzle.ViewportSwizzleNegativeXNv;
+                case ViewportSwizzle.NegativeY:
+                    return NvViewportSwizzle.ViewportSwizzleNegativeYNv;
+                case ViewportSwizzle.NegativeZ:
+                    return NvViewportSwizzle.ViewportSwizzleNegativeZNv;
+                case ViewportSwizzle.NegativeW:
+                    return NvViewportSwizzle.ViewportSwizzleNegativeWNv;
+            }
+
+            Logger.Debug?.Print(LogClass.Gpu, $"Invalid {nameof(ViewportSwizzle)} enum value: {swizzle}.");
+
+            return NvViewportSwizzle.ViewportSwizzlePositiveXNv;
+        }
+
+        public static All Convert(this LogicalOp op)
+        {
+            switch (op)
+            {
+                case LogicalOp.Clear:
+                    return All.Clear;
+                case LogicalOp.And:
+                    return All.And;
+                case LogicalOp.AndReverse:
+                    return All.AndReverse;
+                case LogicalOp.Copy:
+                    return All.Copy;
+                case LogicalOp.AndInverted:
+                    return All.AndInverted;
+                case LogicalOp.Noop:
+                    return All.Noop;
+                case LogicalOp.Xor:
+                    return All.Xor;
+                case LogicalOp.Or:
+                    return All.Or;
+                case LogicalOp.Nor:
+                    return All.Nor;
+                case LogicalOp.Equiv:
+                    return All.Equiv;
+                case LogicalOp.Invert:
+                    return All.Invert;
+                case LogicalOp.OrReverse:
+                    return All.OrReverse;
+                case LogicalOp.CopyInverted:
+                    return All.CopyInverted;
+                case LogicalOp.OrInverted:
+                    return All.OrInverted;
+                case LogicalOp.Nand:
+                    return All.Nand;
+                case LogicalOp.Set:
+                    return All.Set;
+            }
+
+            Logger.Debug?.Print(LogClass.Gpu, $"Invalid {nameof(LogicalOp)} enum value: {op}.");
+
+            return All.Never;
+        }
+
+        public static ShaderType Convert(this ShaderStage stage)
+        {
+            return stage switch
+            {
+                ShaderStage.Compute => ShaderType.ComputeShader,
+                ShaderStage.Vertex => ShaderType.VertexShader,
+                ShaderStage.TessellationControl => ShaderType.TessControlShader,
+                ShaderStage.TessellationEvaluation => ShaderType.TessEvaluationShader,
+                ShaderStage.Geometry => ShaderType.GeometryShader,
+                ShaderStage.Fragment => ShaderType.FragmentShader,
+                _ => ShaderType.VertexShader
+            };
+        }
+    }
+}
diff --git a/src/Ryujinx.Graphics.OpenGL/FormatInfo.cs b/src/Ryujinx.Graphics.OpenGL/FormatInfo.cs
new file mode 100644
index 00000000..aef4dbad
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/FormatInfo.cs
@@ -0,0 +1,45 @@
+using OpenTK.Graphics.OpenGL;
+
+namespace Ryujinx.Graphics.OpenGL
+{
+    readonly struct FormatInfo
+    {
+        public int  Components { get; }
+        public bool Normalized { get; }
+        public bool Scaled     { get; }
+
+        public PixelInternalFormat PixelInternalFormat { get; }
+        public PixelFormat         PixelFormat         { get; }
+        public PixelType           PixelType           { get; }
+
+        public bool IsCompressed { get; }
+
+        public FormatInfo(
+            int         components,
+            bool        normalized,
+            bool        scaled,
+            All         pixelInternalFormat,
+            PixelFormat pixelFormat,
+            PixelType   pixelType)
+        {
+            Components          = components;
+            Normalized          = normalized;
+            Scaled              = scaled;
+            PixelInternalFormat = (PixelInternalFormat)pixelInternalFormat;
+            PixelFormat         = pixelFormat;
+            PixelType           = pixelType;
+            IsCompressed        = false;
+        }
+
+        public FormatInfo(int components, bool normalized, bool scaled, All pixelFormat)
+        {
+            Components          = components;
+            Normalized          = normalized;
+            Scaled              = scaled;
+            PixelInternalFormat = 0;
+            PixelFormat         = (PixelFormat)pixelFormat;
+            PixelType           = 0;
+            IsCompressed        = true;
+        }
+    }
+}
diff --git a/src/Ryujinx.Graphics.OpenGL/FormatTable.cs b/src/Ryujinx.Graphics.OpenGL/FormatTable.cs
new file mode 100644
index 00000000..281f0004
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/FormatTable.cs
@@ -0,0 +1,225 @@
+using OpenTK.Graphics.OpenGL;
+using Ryujinx.Graphics.GAL;
+using System;
+
+namespace Ryujinx.Graphics.OpenGL
+{
+    struct FormatTable
+    {
+        private static FormatInfo[] _table;
+        private static SizedInternalFormat[] _tableImage;
+
+        static FormatTable()
+        {
+            int tableSize = Enum.GetNames<Format>().Length;
+
+            _table = new FormatInfo[tableSize];
+            _tableImage = new SizedInternalFormat[tableSize];
+
+            Add(Format.R8Unorm,             new FormatInfo(1, true,  false, All.R8,                PixelFormat.Red,            PixelType.UnsignedByte));
+            Add(Format.R8Snorm,             new FormatInfo(1, true,  false, All.R8Snorm,           PixelFormat.Red,            PixelType.Byte));
+            Add(Format.R8Uint,              new FormatInfo(1, false, false, All.R8ui,              PixelFormat.RedInteger,     PixelType.UnsignedByte));
+            Add(Format.R8Sint,              new FormatInfo(1, false, false, All.R8i,               PixelFormat.RedInteger,     PixelType.Byte));
+            Add(Format.R16Float,            new FormatInfo(1, false, false, All.R16f,              PixelFormat.Red,            PixelType.HalfFloat));
+            Add(Format.R16Unorm,            new FormatInfo(1, true,  false, All.R16,               PixelFormat.Red,            PixelType.UnsignedShort));
+            Add(Format.R16Snorm,            new FormatInfo(1, true,  false, All.R16Snorm,          PixelFormat.Red,            PixelType.Short));
+            Add(Format.R16Uint,             new FormatInfo(1, false, false, All.R16ui,             PixelFormat.RedInteger,     PixelType.UnsignedShort));
+            Add(Format.R16Sint,             new FormatInfo(1, false, false, All.R16i,              PixelFormat.RedInteger,     PixelType.Short));
+            Add(Format.R32Float,            new FormatInfo(1, false, false, All.R32f,              PixelFormat.Red,            PixelType.Float));
+            Add(Format.R32Uint,             new FormatInfo(1, false, false, All.R32ui,             PixelFormat.RedInteger,     PixelType.UnsignedInt));
+            Add(Format.R32Sint,             new FormatInfo(1, false, false, All.R32i,              PixelFormat.RedInteger,     PixelType.Int));
+            Add(Format.R8G8Unorm,           new FormatInfo(2, true,  false, All.Rg8,               PixelFormat.Rg,             PixelType.UnsignedByte));
+            Add(Format.R8G8Snorm,           new FormatInfo(2, true,  false, All.Rg8Snorm,          PixelFormat.Rg,             PixelType.Byte));
+            Add(Format.R8G8Uint,            new FormatInfo(2, false, false, All.Rg8ui,             PixelFormat.RgInteger,      PixelType.UnsignedByte));
+            Add(Format.R8G8Sint,            new FormatInfo(2, false, false, All.Rg8i,              PixelFormat.RgInteger,      PixelType.Byte));
+            Add(Format.R16G16Float,         new FormatInfo(2, false, false, All.Rg16f,             PixelFormat.Rg,             PixelType.HalfFloat));
+            Add(Format.R16G16Unorm,         new FormatInfo(2, true,  false, All.Rg16,              PixelFormat.Rg,             PixelType.UnsignedShort));
+            Add(Format.R16G16Snorm,         new FormatInfo(2, true,  false, All.Rg16Snorm,         PixelFormat.Rg,             PixelType.Short));
+            Add(Format.R16G16Uint,          new FormatInfo(2, false, false, All.Rg16ui,            PixelFormat.RgInteger,      PixelType.UnsignedShort));
+            Add(Format.R16G16Sint,          new FormatInfo(2, false, false, All.Rg16i,             PixelFormat.RgInteger,      PixelType.Short));
+            Add(Format.R32G32Float,         new FormatInfo(2, false, false, All.Rg32f,             PixelFormat.Rg,             PixelType.Float));
+            Add(Format.R32G32Uint,          new FormatInfo(2, false, false, All.Rg32ui,            PixelFormat.RgInteger,      PixelType.UnsignedInt));
+            Add(Format.R32G32Sint,          new FormatInfo(2, false, false, All.Rg32i,             PixelFormat.RgInteger,      PixelType.Int));
+            Add(Format.R8G8B8Unorm,         new FormatInfo(3, true,  false, All.Rgb8,              PixelFormat.Rgb,            PixelType.UnsignedByte));
+            Add(Format.R8G8B8Snorm,         new FormatInfo(3, true,  false, All.Rgb8Snorm,         PixelFormat.Rgb,            PixelType.Byte));
+            Add(Format.R8G8B8Uint,          new FormatInfo(3, false, false, All.Rgb8ui,            PixelFormat.RgbInteger,     PixelType.UnsignedByte));
+            Add(Format.R8G8B8Sint,          new FormatInfo(3, false, false, All.Rgb8i,             PixelFormat.RgbInteger,     PixelType.Byte));
+            Add(Format.R16G16B16Float,      new FormatInfo(3, false, false, All.Rgb16f,            PixelFormat.Rgb,            PixelType.HalfFloat));
+            Add(Format.R16G16B16Unorm,      new FormatInfo(3, true,  false, All.Rgb16,             PixelFormat.Rgb,            PixelType.UnsignedShort));
+            Add(Format.R16G16B16Snorm,      new FormatInfo(3, true,  false, All.Rgb16Snorm,        PixelFormat.Rgb,            PixelType.Short));
+            Add(Format.R16G16B16Uint,       new FormatInfo(3, false, false, All.Rgb16ui,           PixelFormat.RgbInteger,     PixelType.UnsignedShort));
+            Add(Format.R16G16B16Sint,       new FormatInfo(3, false, false, All.Rgb16i,            PixelFormat.RgbInteger,     PixelType.Short));
+            Add(Format.R32G32B32Float,      new FormatInfo(3, false, false, All.Rgb32f,            PixelFormat.Rgb,            PixelType.Float));
+            Add(Format.R32G32B32Uint,       new FormatInfo(3, false, false, All.Rgb32ui,           PixelFormat.RgbInteger,     PixelType.UnsignedInt));
+            Add(Format.R32G32B32Sint,       new FormatInfo(3, false, false, All.Rgb32i,            PixelFormat.RgbInteger,     PixelType.Int));
+            Add(Format.R8G8B8A8Unorm,       new FormatInfo(4, true,  false, All.Rgba8,             PixelFormat.Rgba,           PixelType.UnsignedByte));
+            Add(Format.R8G8B8A8Snorm,       new FormatInfo(4, true,  false, All.Rgba8Snorm,        PixelFormat.Rgba,           PixelType.Byte));
+            Add(Format.R8G8B8A8Uint,        new FormatInfo(4, false, false, All.Rgba8ui,           PixelFormat.RgbaInteger,    PixelType.UnsignedByte));
+            Add(Format.R8G8B8A8Sint,        new FormatInfo(4, false, false, All.Rgba8i,            PixelFormat.RgbaInteger,    PixelType.Byte));
+            Add(Format.R16G16B16A16Float,   new FormatInfo(4, false, false, All.Rgba16f,           PixelFormat.Rgba,           PixelType.HalfFloat));
+            Add(Format.R16G16B16A16Unorm,   new FormatInfo(4, true,  false, All.Rgba16,            PixelFormat.Rgba,           PixelType.UnsignedShort));
+            Add(Format.R16G16B16A16Snorm,   new FormatInfo(4, true,  false, All.Rgba16Snorm,       PixelFormat.Rgba,           PixelType.Short));
+            Add(Format.R16G16B16A16Uint,    new FormatInfo(4, false, false, All.Rgba16ui,          PixelFormat.RgbaInteger,    PixelType.UnsignedShort));
+            Add(Format.R16G16B16A16Sint,    new FormatInfo(4, false, false, All.Rgba16i,           PixelFormat.RgbaInteger,    PixelType.Short));
+            Add(Format.R32G32B32A32Float,   new FormatInfo(4, false, false, All.Rgba32f,           PixelFormat.Rgba,           PixelType.Float));
+            Add(Format.R32G32B32A32Uint,    new FormatInfo(4, false, false, All.Rgba32ui,          PixelFormat.RgbaInteger,    PixelType.UnsignedInt));
+            Add(Format.R32G32B32A32Sint,    new FormatInfo(4, false, false, All.Rgba32i,           PixelFormat.RgbaInteger,    PixelType.Int));
+            Add(Format.S8Uint,              new FormatInfo(1, false, false, All.StencilIndex8,     PixelFormat.StencilIndex,   PixelType.UnsignedByte));
+            Add(Format.D16Unorm,            new FormatInfo(1, false, false, All.DepthComponent16,  PixelFormat.DepthComponent, PixelType.UnsignedShort));
+            Add(Format.S8UintD24Unorm,      new FormatInfo(1, false, false, All.Depth24Stencil8,   PixelFormat.DepthStencil,   PixelType.UnsignedInt248));
+            Add(Format.D32Float,            new FormatInfo(1, false, false, All.DepthComponent32f, PixelFormat.DepthComponent, PixelType.Float));
+            Add(Format.D24UnormS8Uint,      new FormatInfo(1, false, false, All.Depth24Stencil8,   PixelFormat.DepthStencil,   PixelType.UnsignedInt248));
+            Add(Format.D32FloatS8Uint,      new FormatInfo(1, false, false, All.Depth32fStencil8,  PixelFormat.DepthStencil,   PixelType.Float32UnsignedInt248Rev));
+            Add(Format.R8G8B8A8Srgb,        new FormatInfo(4, false, false, All.Srgb8Alpha8,       PixelFormat.Rgba,           PixelType.UnsignedByte));
+            Add(Format.R4G4B4A4Unorm,       new FormatInfo(4, true,  false, All.Rgba4,             PixelFormat.Rgba,           PixelType.UnsignedShort4444Reversed));
+            Add(Format.R5G5B5X1Unorm,       new FormatInfo(4, true,  false, All.Rgb5,              PixelFormat.Rgb,            PixelType.UnsignedShort1555Reversed));
+            Add(Format.R5G5B5A1Unorm,       new FormatInfo(4, true,  false, All.Rgb5A1,            PixelFormat.Rgba,           PixelType.UnsignedShort1555Reversed));
+            Add(Format.R5G6B5Unorm,         new FormatInfo(3, true,  false, All.Rgb565,            PixelFormat.Rgb,            PixelType.UnsignedShort565Reversed));
+            Add(Format.R10G10B10A2Unorm,    new FormatInfo(4, true,  false, All.Rgb10A2,           PixelFormat.Rgba,           PixelType.UnsignedInt2101010Reversed));
+            Add(Format.R10G10B10A2Uint,     new FormatInfo(4, false, false, All.Rgb10A2ui,         PixelFormat.RgbaInteger,    PixelType.UnsignedInt2101010Reversed));
+            Add(Format.R11G11B10Float,      new FormatInfo(3, false, false, All.R11fG11fB10f,      PixelFormat.Rgb,            PixelType.UnsignedInt10F11F11FRev));
+            Add(Format.R9G9B9E5Float,       new FormatInfo(3, false, false, All.Rgb9E5,            PixelFormat.Rgb,            PixelType.UnsignedInt5999Rev));
+            Add(Format.Bc1RgbaUnorm,        new FormatInfo(4, true,  false, All.CompressedRgbaS3tcDxt1Ext));
+            Add(Format.Bc2Unorm,            new FormatInfo(4, true,  false, All.CompressedRgbaS3tcDxt3Ext));
+            Add(Format.Bc3Unorm,            new FormatInfo(4, true,  false, All.CompressedRgbaS3tcDxt5Ext));
+            Add(Format.Bc1RgbaSrgb,         new FormatInfo(4, true,  false, All.CompressedSrgbAlphaS3tcDxt1Ext));
+            Add(Format.Bc2Srgb,             new FormatInfo(4, false, false, All.CompressedSrgbAlphaS3tcDxt3Ext));
+            Add(Format.Bc3Srgb,             new FormatInfo(4, false, false, All.CompressedSrgbAlphaS3tcDxt5Ext));
+            Add(Format.Bc4Unorm,            new FormatInfo(1, true,  false, All.CompressedRedRgtc1));
+            Add(Format.Bc4Snorm,            new FormatInfo(1, true,  false, All.CompressedSignedRedRgtc1));
+            Add(Format.Bc5Unorm,            new FormatInfo(2, true,  false, All.CompressedRgRgtc2));
+            Add(Format.Bc5Snorm,            new FormatInfo(2, true,  false, All.CompressedSignedRgRgtc2));
+            Add(Format.Bc7Unorm,            new FormatInfo(4, true,  false, All.CompressedRgbaBptcUnorm));
+            Add(Format.Bc7Srgb,             new FormatInfo(4, false, false, All.CompressedSrgbAlphaBptcUnorm));
+            Add(Format.Bc6HSfloat,          new FormatInfo(4, false, false, All.CompressedRgbBptcSignedFloat));
+            Add(Format.Bc6HUfloat,          new FormatInfo(4, false, false, All.CompressedRgbBptcUnsignedFloat));
+            Add(Format.Etc2RgbUnorm,        new FormatInfo(4, false, false, All.CompressedRgb8Etc2));
+            Add(Format.Etc2RgbaUnorm,       new FormatInfo(4, false, false, All.CompressedRgba8Etc2Eac));
+            Add(Format.Etc2RgbPtaUnorm,     new FormatInfo(4, false, false, All.CompressedRgb8PunchthroughAlpha1Etc2));
+            Add(Format.Etc2RgbSrgb,         new FormatInfo(4, false, false, All.CompressedSrgb8Etc2));
+            Add(Format.Etc2RgbaSrgb,        new FormatInfo(4, false, false, All.CompressedSrgb8Alpha8Etc2Eac));
+            Add(Format.Etc2RgbPtaSrgb,      new FormatInfo(4, false, false, All.CompressedSrgb8PunchthroughAlpha1Etc2));
+            Add(Format.R8Uscaled,           new FormatInfo(1, false, true,  All.R8ui,              PixelFormat.RedInteger,     PixelType.UnsignedByte));
+            Add(Format.R8Sscaled,           new FormatInfo(1, false, true,  All.R8i,               PixelFormat.RedInteger,     PixelType.Byte));
+            Add(Format.R16Uscaled,          new FormatInfo(1, false, true,  All.R16ui,             PixelFormat.RedInteger,     PixelType.UnsignedShort));
+            Add(Format.R16Sscaled,          new FormatInfo(1, false, true,  All.R16i,              PixelFormat.RedInteger,     PixelType.Short));
+            Add(Format.R32Uscaled,          new FormatInfo(1, false, true,  All.R32ui,             PixelFormat.RedInteger,     PixelType.UnsignedInt));
+            Add(Format.R32Sscaled,          new FormatInfo(1, false, true,  All.R32i,              PixelFormat.RedInteger,     PixelType.Int));
+            Add(Format.R8G8Uscaled,         new FormatInfo(2, false, true,  All.Rg8ui,             PixelFormat.RgInteger,      PixelType.UnsignedByte));
+            Add(Format.R8G8Sscaled,         new FormatInfo(2, false, true,  All.Rg8i,              PixelFormat.RgInteger,      PixelType.Byte));
+            Add(Format.R16G16Uscaled,       new FormatInfo(2, false, true,  All.Rg16ui,            PixelFormat.RgInteger,      PixelType.UnsignedShort));
+            Add(Format.R16G16Sscaled,       new FormatInfo(2, false, true,  All.Rg16i,             PixelFormat.RgInteger,      PixelType.Short));
+            Add(Format.R32G32Uscaled,       new FormatInfo(2, false, true,  All.Rg32ui,            PixelFormat.RgInteger,      PixelType.UnsignedInt));
+            Add(Format.R32G32Sscaled,       new FormatInfo(2, false, true,  All.Rg32i,             PixelFormat.RgInteger,      PixelType.Int));
+            Add(Format.R8G8B8Uscaled,       new FormatInfo(3, false, true,  All.Rgb8ui,            PixelFormat.RgbInteger,     PixelType.UnsignedByte));
+            Add(Format.R8G8B8Sscaled,       new FormatInfo(3, false, true,  All.Rgb8i,             PixelFormat.RgbInteger,     PixelType.Byte));
+            Add(Format.R16G16B16Uscaled,    new FormatInfo(3, false, true,  All.Rgb16ui,           PixelFormat.RgbInteger,     PixelType.UnsignedShort));
+            Add(Format.R16G16B16Sscaled,    new FormatInfo(3, false, true,  All.Rgb16i,            PixelFormat.RgbInteger,     PixelType.Short));
+            Add(Format.R32G32B32Uscaled,    new FormatInfo(3, false, true,  All.Rgb32ui,           PixelFormat.RgbInteger,     PixelType.UnsignedInt));
+            Add(Format.R32G32B32Sscaled,    new FormatInfo(3, false, true,  All.Rgb32i,            PixelFormat.RgbInteger,     PixelType.Int));
+            Add(Format.R8G8B8A8Uscaled,     new FormatInfo(4, false, true,  All.Rgba8ui,           PixelFormat.RgbaInteger,    PixelType.UnsignedByte));
+            Add(Format.R8G8B8A8Sscaled,     new FormatInfo(4, false, true,  All.Rgba8i,            PixelFormat.RgbaInteger,    PixelType.Byte));
+            Add(Format.R16G16B16A16Uscaled, new FormatInfo(4, false, true,  All.Rgba16ui,          PixelFormat.RgbaInteger,    PixelType.UnsignedShort));
+            Add(Format.R16G16B16A16Sscaled, new FormatInfo(4, false, true,  All.Rgba16i,           PixelFormat.RgbaInteger,    PixelType.Short));
+            Add(Format.R32G32B32A32Uscaled, new FormatInfo(4, false, true,  All.Rgba32ui,          PixelFormat.RgbaInteger,    PixelType.UnsignedInt));
+            Add(Format.R32G32B32A32Sscaled, new FormatInfo(4, false, true,  All.Rgba32i,           PixelFormat.RgbaInteger,    PixelType.Int));
+            Add(Format.R10G10B10A2Snorm,    new FormatInfo(4, true,  false, All.Rgb10A2,           PixelFormat.Rgba,           (PixelType)All.Int2101010Rev));
+            Add(Format.R10G10B10A2Sint,     new FormatInfo(4, false, false, All.Rgb10A2,           PixelFormat.RgbaInteger,    (PixelType)All.Int2101010Rev));
+            Add(Format.R10G10B10A2Uscaled,  new FormatInfo(4, false, true,  All.Rgb10A2ui,         PixelFormat.RgbaInteger,    PixelType.UnsignedInt2101010Reversed));
+            Add(Format.R10G10B10A2Sscaled,  new FormatInfo(4, false, true,  All.Rgb10A2,           PixelFormat.RgbaInteger,    PixelType.UnsignedInt2101010Reversed));
+            Add(Format.Astc4x4Unorm,        new FormatInfo(4, true,  false, All.CompressedRgbaAstc4X4Khr));
+            Add(Format.Astc5x4Unorm,        new FormatInfo(4, true,  false, All.CompressedRgbaAstc5X4Khr));
+            Add(Format.Astc5x5Unorm,        new FormatInfo(4, true,  false, All.CompressedRgbaAstc5X5Khr));
+            Add(Format.Astc6x5Unorm,        new FormatInfo(4, true,  false, All.CompressedRgbaAstc6X5Khr));
+            Add(Format.Astc6x6Unorm,        new FormatInfo(4, true,  false, All.CompressedRgbaAstc6X6Khr));
+            Add(Format.Astc8x5Unorm,        new FormatInfo(4, true,  false, All.CompressedRgbaAstc8X5Khr));
+            Add(Format.Astc8x6Unorm,        new FormatInfo(4, true,  false, All.CompressedRgbaAstc8X6Khr));
+            Add(Format.Astc8x8Unorm,        new FormatInfo(4, true,  false, All.CompressedRgbaAstc8X8Khr));
+            Add(Format.Astc10x5Unorm,       new FormatInfo(4, true,  false, All.CompressedRgbaAstc10X5Khr));
+            Add(Format.Astc10x6Unorm,       new FormatInfo(4, true,  false, All.CompressedRgbaAstc10X6Khr));
+            Add(Format.Astc10x8Unorm,       new FormatInfo(4, true,  false, All.CompressedRgbaAstc10X8Khr));
+            Add(Format.Astc10x10Unorm,      new FormatInfo(4, true,  false, All.CompressedRgbaAstc10X10Khr));
+            Add(Format.Astc12x10Unorm,      new FormatInfo(4, true,  false, All.CompressedRgbaAstc12X10Khr));
+            Add(Format.Astc12x12Unorm,      new FormatInfo(4, true,  false, All.CompressedRgbaAstc12X12Khr));
+            Add(Format.Astc4x4Srgb,         new FormatInfo(4, false, false, All.CompressedSrgb8Alpha8Astc4X4Khr));
+            Add(Format.Astc5x4Srgb,         new FormatInfo(4, false, false, All.CompressedSrgb8Alpha8Astc5X4Khr));
+            Add(Format.Astc5x5Srgb,         new FormatInfo(4, false, false, All.CompressedSrgb8Alpha8Astc5X5Khr));
+            Add(Format.Astc6x5Srgb,         new FormatInfo(4, false, false, All.CompressedSrgb8Alpha8Astc6X5Khr));
+            Add(Format.Astc6x6Srgb,         new FormatInfo(4, false, false, All.CompressedSrgb8Alpha8Astc6X6Khr));
+            Add(Format.Astc8x5Srgb,         new FormatInfo(4, false, false, All.CompressedSrgb8Alpha8Astc8X5Khr));
+            Add(Format.Astc8x6Srgb,         new FormatInfo(4, false, false, All.CompressedSrgb8Alpha8Astc8X6Khr));
+            Add(Format.Astc8x8Srgb,         new FormatInfo(4, false, false, All.CompressedSrgb8Alpha8Astc8X8Khr));
+            Add(Format.Astc10x5Srgb,        new FormatInfo(4, false, false, All.CompressedSrgb8Alpha8Astc10X5Khr));
+            Add(Format.Astc10x6Srgb,        new FormatInfo(4, false, false, All.CompressedSrgb8Alpha8Astc10X6Khr));
+            Add(Format.Astc10x8Srgb,        new FormatInfo(4, false, false, All.CompressedSrgb8Alpha8Astc10X8Khr));
+            Add(Format.Astc10x10Srgb,       new FormatInfo(4, false, false, All.CompressedSrgb8Alpha8Astc10X10Khr));
+            Add(Format.Astc12x10Srgb,       new FormatInfo(4, false, false, All.CompressedSrgb8Alpha8Astc12X10Khr));
+            Add(Format.Astc12x12Srgb,       new FormatInfo(4, false, false, All.CompressedSrgb8Alpha8Astc12X12Khr));
+            Add(Format.B5G6R5Unorm,         new FormatInfo(3, true,  false, All.Rgb565,            PixelFormat.Rgb,            PixelType.UnsignedShort565Reversed));
+            Add(Format.B5G5R5A1Unorm,       new FormatInfo(4, true,  false, All.Rgb5A1,            PixelFormat.Rgba,           PixelType.UnsignedShort1555Reversed));
+            Add(Format.A1B5G5R5Unorm,       new FormatInfo(4, true,  false, All.Rgb5A1,            PixelFormat.Rgba,           PixelType.UnsignedShort5551));
+            Add(Format.B8G8R8A8Unorm,       new FormatInfo(4, true,  false, All.Rgba8,             PixelFormat.Rgba,           PixelType.UnsignedByte));
+            Add(Format.B8G8R8A8Srgb,        new FormatInfo(4, false, false, All.Srgb8Alpha8,       PixelFormat.Rgba,           PixelType.UnsignedByte));
+
+            Add(Format.R8Unorm,           SizedInternalFormat.R8);
+            Add(Format.R8Uint,            SizedInternalFormat.R8ui);
+            Add(Format.R8Sint,            SizedInternalFormat.R8i);
+            Add(Format.R16Float,          SizedInternalFormat.R16f);
+            Add(Format.R16Unorm,          SizedInternalFormat.R16);
+            Add(Format.R16Snorm,          (SizedInternalFormat)All.R16Snorm);
+            Add(Format.R16Uint,           SizedInternalFormat.R16ui);
+            Add(Format.R16Sint,           SizedInternalFormat.R16i);
+            Add(Format.R32Float,          SizedInternalFormat.R32f);
+            Add(Format.R32Uint,           SizedInternalFormat.R32ui);
+            Add(Format.R32Sint,           SizedInternalFormat.R32i);
+            Add(Format.R8G8Unorm,         SizedInternalFormat.Rg8);
+            Add(Format.R8G8Snorm,         (SizedInternalFormat)All.Rg8Snorm);
+            Add(Format.R8G8Uint,          SizedInternalFormat.Rg8ui);
+            Add(Format.R8G8Sint,          SizedInternalFormat.Rg8i);
+            Add(Format.R16G16Float,       SizedInternalFormat.Rg16f);
+            Add(Format.R16G16Unorm,       SizedInternalFormat.Rg16);
+            Add(Format.R16G16Snorm,       (SizedInternalFormat)All.Rg16Snorm);
+            Add(Format.R16G16Uint,        SizedInternalFormat.Rg16ui);
+            Add(Format.R16G16Sint,        SizedInternalFormat.Rg16i);
+            Add(Format.R32G32Float,       SizedInternalFormat.Rg32f);
+            Add(Format.R32G32Uint,        SizedInternalFormat.Rg32ui);
+            Add(Format.R32G32Sint,        SizedInternalFormat.Rg32i);
+            Add(Format.R8G8B8A8Unorm,     SizedInternalFormat.Rgba8);
+            Add(Format.R8G8B8A8Snorm,     (SizedInternalFormat)All.Rgba8Snorm);
+            Add(Format.R8G8B8A8Uint,      SizedInternalFormat.Rgba8ui);
+            Add(Format.R8G8B8A8Sint,      SizedInternalFormat.Rgba8i);
+            Add(Format.R16G16B16A16Float, SizedInternalFormat.Rgba16f);
+            Add(Format.R16G16B16A16Unorm, SizedInternalFormat.Rgba16);
+            Add(Format.R16G16B16A16Snorm, (SizedInternalFormat)All.Rgba16Snorm);
+            Add(Format.R16G16B16A16Uint,  SizedInternalFormat.Rgba16ui);
+            Add(Format.R16G16B16A16Sint,  SizedInternalFormat.Rgba16i);
+            Add(Format.R32G32B32A32Float, SizedInternalFormat.Rgba32f);
+            Add(Format.R32G32B32A32Uint,  SizedInternalFormat.Rgba32ui);
+            Add(Format.R32G32B32A32Sint,  SizedInternalFormat.Rgba32i);
+            Add(Format.R8G8B8A8Srgb,      SizedInternalFormat.Rgba8);
+            Add(Format.R10G10B10A2Unorm,  (SizedInternalFormat)All.Rgb10A2);
+            Add(Format.R10G10B10A2Uint,   (SizedInternalFormat)All.Rgb10A2ui);
+            Add(Format.R11G11B10Float,    (SizedInternalFormat)All.R11fG11fB10f);
+        }
+
+        private static void Add(Format format, FormatInfo info)
+        {
+            _table[(int)format] = info;
+        }
+
+        private static void Add(Format format, SizedInternalFormat sif)
+        {
+            _tableImage[(int)format] = sif;
+        }
+
+        public static FormatInfo GetFormatInfo(Format format)
+        {
+            return _table[(int)format];
+        }
+
+        public static SizedInternalFormat GetImageFormat(Format format)
+        {
+            return _tableImage[(int)format];
+        }
+    }
+}
diff --git a/src/Ryujinx.Graphics.OpenGL/Framebuffer.cs b/src/Ryujinx.Graphics.OpenGL/Framebuffer.cs
new file mode 100644
index 00000000..b180b857
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Framebuffer.cs
@@ -0,0 +1,247 @@
+using OpenTK.Graphics.OpenGL;
+using Ryujinx.Graphics.GAL;
+using Ryujinx.Graphics.OpenGL.Image;
+using System;
+using System.Runtime.CompilerServices;
+
+namespace Ryujinx.Graphics.OpenGL
+{
+    class Framebuffer : IDisposable
+    {
+        public int Handle { get; private set; }
+        private int _clearFbHandle;
+        private bool _clearFbInitialized;
+
+        private FramebufferAttachment _lastDsAttachment;
+
+        private readonly TextureView[] _colors;
+        private TextureView _depthStencil;
+
+        private int _colorsCount;
+        private bool _dualSourceBlend;
+
+        public Framebuffer()
+        {
+            Handle = GL.GenFramebuffer();
+            _clearFbHandle = GL.GenFramebuffer();
+
+            _colors = new TextureView[8];
+        }
+
+        public int Bind()
+        {
+            GL.BindFramebuffer(FramebufferTarget.Framebuffer, Handle);
+            return Handle;
+        }
+
+        [MethodImpl(MethodImplOptions.AggressiveInlining)]
+        public void AttachColor(int index, TextureView color)
+        {
+            if (_colors[index] == color)
+            {
+                return;
+            }
+
+            FramebufferAttachment attachment = FramebufferAttachment.ColorAttachment0 + index;
+
+            GL.FramebufferTexture(FramebufferTarget.Framebuffer, attachment, color?.Handle ?? 0, 0);
+
+            _colors[index] = color;
+        }
+
+        public void AttachDepthStencil(TextureView depthStencil)
+        {
+            // Detach the last depth/stencil buffer if there is any.
+            if (_lastDsAttachment != 0)
+            {
+                GL.FramebufferTexture(FramebufferTarget.Framebuffer, _lastDsAttachment, 0, 0);
+            }
+
+            if (depthStencil != null)
+            {
+                FramebufferAttachment attachment = GetAttachment(depthStencil.Format);
+
+                GL.FramebufferTexture(
+                    FramebufferTarget.Framebuffer,
+                    attachment,
+                    depthStencil.Handle,
+                    0);
+
+                _lastDsAttachment = attachment;
+            }
+            else
+            {
+                _lastDsAttachment = 0;
+            }
+
+            _depthStencil = depthStencil;
+        }
+
+        public void SetDualSourceBlend(bool enable)
+        {
+            bool oldEnable = _dualSourceBlend;
+
+            _dualSourceBlend = enable;
+
+            // When dual source blend is used,
+            // we can only have one draw buffer.
+            if (enable)
+            {
+                GL.DrawBuffer(DrawBufferMode.ColorAttachment0);
+            }
+            else if (oldEnable)
+            {
+                SetDrawBuffersImpl(_colorsCount);
+            }
+        }
+
+        public void SetDrawBuffers(int colorsCount)
+        {
+            if (_colorsCount != colorsCount && !_dualSourceBlend)
+            {
+                SetDrawBuffersImpl(colorsCount);
+            }
+
+            _colorsCount = colorsCount;
+        }
+
+        private void SetDrawBuffersImpl(int colorsCount)
+        {
+            DrawBuffersEnum[] drawBuffers = new DrawBuffersEnum[colorsCount];
+
+            for (int index = 0; index < colorsCount; index++)
+            {
+                drawBuffers[index] = DrawBuffersEnum.ColorAttachment0 + index;
+            }
+
+            GL.DrawBuffers(colorsCount, drawBuffers);
+        }
+
+        private static FramebufferAttachment GetAttachment(Format format)
+        {
+            if (IsPackedDepthStencilFormat(format))
+            {
+                return FramebufferAttachment.DepthStencilAttachment;
+            }
+            else if (IsDepthOnlyFormat(format))
+            {
+                return FramebufferAttachment.DepthAttachment;
+            }
+            else
+            {
+                return FramebufferAttachment.StencilAttachment;
+            }
+        }
+
+        private static bool IsPackedDepthStencilFormat(Format format)
+        {
+            return format == Format.D24UnormS8Uint ||
+                   format == Format.D32FloatS8Uint ||
+                   format == Format.S8UintD24Unorm;
+        }
+
+        private static bool IsDepthOnlyFormat(Format format)
+        {
+            return format == Format.D16Unorm || format == Format.D32Float;
+        }
+
+        public int GetColorLayerCount(int index)
+        {
+            return _colors[index]?.Info.GetDepthOrLayers() ?? 0;
+        }
+
+        public int GetDepthStencilLayerCount()
+        {
+            return _depthStencil?.Info.GetDepthOrLayers() ?? 0;
+        }
+
+        public void AttachColorLayerForClear(int index, int layer)
+        {
+            TextureView color = _colors[index];
+
+            if (!IsLayered(color))
+            {
+                return;
+            }
+
+            BindClearFb();
+            GL.FramebufferTextureLayer(FramebufferTarget.Framebuffer, FramebufferAttachment.ColorAttachment0 + index, color.Handle, 0, layer);
+        }
+
+        public void DetachColorLayerForClear(int index)
+        {
+            TextureView color = _colors[index];
+
+            if (!IsLayered(color))
+            {
+                return;
+            }
+
+            GL.FramebufferTexture(FramebufferTarget.Framebuffer, FramebufferAttachment.ColorAttachment0 + index, 0, 0);
+            Bind();
+        }
+
+        public void AttachDepthStencilLayerForClear(int layer)
+        {
+            TextureView depthStencil = _depthStencil;
+
+            if (!IsLayered(depthStencil))
+            {
+                return;
+            }
+
+            BindClearFb();
+            GL.FramebufferTextureLayer(FramebufferTarget.Framebuffer, GetAttachment(depthStencil.Format), depthStencil.Handle, 0, layer);
+        }
+
+        public void DetachDepthStencilLayerForClear()
+        {
+            TextureView depthStencil = _depthStencil;
+
+            if (!IsLayered(depthStencil))
+            {
+                return;
+            }
+
+            GL.FramebufferTexture(FramebufferTarget.Framebuffer, GetAttachment(depthStencil.Format), 0, 0);
+            Bind();
+        }
+
+        private void BindClearFb()
+        {
+            GL.BindFramebuffer(FramebufferTarget.Framebuffer, _clearFbHandle);
+
+            if (!_clearFbInitialized)
+            {
+                SetDrawBuffersImpl(Constants.MaxRenderTargets);
+                _clearFbInitialized = true;
+            }
+        }
+
+        private static bool IsLayered(TextureView view)
+        {
+            return view != null &&
+                   view.Target != Target.Texture1D &&
+                   view.Target != Target.Texture2D &&
+                   view.Target != Target.Texture2DMultisample &&
+                   view.Target != Target.TextureBuffer;
+        }
+
+        public void Dispose()
+        {
+            if (Handle != 0)
+            {
+                GL.DeleteFramebuffer(Handle);
+
+                Handle = 0;
+            }
+
+            if (_clearFbHandle != 0)
+            {
+                GL.DeleteFramebuffer(_clearFbHandle);
+
+                _clearFbHandle = 0;
+            }
+        }
+    }
+}
diff --git a/src/Ryujinx.Graphics.OpenGL/Handle.cs b/src/Ryujinx.Graphics.OpenGL/Handle.cs
new file mode 100644
index 00000000..4b2f05e6
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Handle.cs
@@ -0,0 +1,23 @@
+using Ryujinx.Graphics.GAL;
+using System.Diagnostics;
+using System.Runtime.CompilerServices;
+
+namespace Ryujinx.Graphics.OpenGL
+{
+    static class Handle
+    {
+        public static T FromInt32<T>(int handle) where T : unmanaged
+        {
+            Debug.Assert(Unsafe.SizeOf<T>() == sizeof(ulong));
+
+            ulong handle64 = (uint)handle;
+
+            return Unsafe.As<ulong, T>(ref handle64);
+        }
+
+        public static int ToInt32(this BufferHandle handle)
+        {
+            return (int)Unsafe.As<BufferHandle, ulong>(ref handle);
+        }
+    }
+}
diff --git a/src/Ryujinx.Graphics.OpenGL/Helper/GLXHelper.cs b/src/Ryujinx.Graphics.OpenGL/Helper/GLXHelper.cs
new file mode 100644
index 00000000..96354946
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Helper/GLXHelper.cs
@@ -0,0 +1,36 @@
+using System;
+using System.Runtime.InteropServices;
+using System.Runtime.Versioning;
+
+namespace Ryujinx.Graphics.OpenGL.Helper
+{
+    [SupportedOSPlatform("linux")]
+    internal static partial class GLXHelper
+    {
+        private const string LibraryName = "glx.dll";
+
+        static GLXHelper()
+        {
+            NativeLibrary.SetDllImportResolver(typeof(GLXHelper).Assembly, (name, assembly, path) =>
+            {
+                if (name != LibraryName)
+                {
+                    return IntPtr.Zero;
+                }
+
+                if (!NativeLibrary.TryLoad("libGL.so.1", assembly, path, out IntPtr result))
+                {
+                    if (!NativeLibrary.TryLoad("libGL.so", assembly, path, out result))
+                    {
+                        return IntPtr.Zero;
+                    }
+                }
+
+                return result;
+            });
+        }
+
+        [LibraryImport(LibraryName, EntryPoint = "glXGetCurrentContext")]
+        public static partial IntPtr GetCurrentContext();
+    }
+}
diff --git a/src/Ryujinx.Graphics.OpenGL/Helper/WGLHelper.cs b/src/Ryujinx.Graphics.OpenGL/Helper/WGLHelper.cs
new file mode 100644
index 00000000..df497ae2
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Helper/WGLHelper.cs
@@ -0,0 +1,15 @@
+using System;
+using System.Runtime.InteropServices;
+using System.Runtime.Versioning;
+
+namespace Ryujinx.Graphics.OpenGL.Helper
+{
+    [SupportedOSPlatform("windows")]
+    internal static partial class WGLHelper
+    {
+        private const string LibraryName = "OPENGL32.DLL";
+
+        [LibraryImport(LibraryName, EntryPoint = "wglGetCurrentContext")]
+        public static partial IntPtr GetCurrentContext();
+    }
+}
diff --git a/src/Ryujinx.Graphics.OpenGL/HwCapabilities.cs b/src/Ryujinx.Graphics.OpenGL/HwCapabilities.cs
new file mode 100644
index 00000000..bf365b4d
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/HwCapabilities.cs
@@ -0,0 +1,142 @@
+using OpenTK.Graphics.OpenGL;
+using System;
+
+namespace Ryujinx.Graphics.OpenGL
+{
+    static class HwCapabilities
+    {
+        private static readonly Lazy<bool> _supportsAlphaToCoverageDitherControl = new Lazy<bool>(() => HasExtension("GL_NV_alpha_to_coverage_dither_control"));
+        private static readonly Lazy<bool> _supportsAstcCompression              = new Lazy<bool>(() => HasExtension("GL_KHR_texture_compression_astc_ldr"));
+        private static readonly Lazy<bool> _supportsBlendEquationAdvanced        = new Lazy<bool>(() => HasExtension("GL_NV_blend_equation_advanced"));
+        private static readonly Lazy<bool> _supportsDrawTexture                  = new Lazy<bool>(() => HasExtension("GL_NV_draw_texture"));
+        private static readonly Lazy<bool> _supportsFragmentShaderInterlock      = new Lazy<bool>(() => HasExtension("GL_ARB_fragment_shader_interlock"));
+        private static readonly Lazy<bool> _supportsFragmentShaderOrdering       = new Lazy<bool>(() => HasExtension("GL_INTEL_fragment_shader_ordering"));
+        private static readonly Lazy<bool> _supportsGeometryShaderPassthrough    = new Lazy<bool>(() => HasExtension("GL_NV_geometry_shader_passthrough"));
+        private static readonly Lazy<bool> _supportsImageLoadFormatted           = new Lazy<bool>(() => HasExtension("GL_EXT_shader_image_load_formatted"));
+        private static readonly Lazy<bool> _supportsIndirectParameters           = new Lazy<bool>(() => HasExtension("GL_ARB_indirect_parameters"));
+        private static readonly Lazy<bool> _supportsParallelShaderCompile        = new Lazy<bool>(() => HasExtension("GL_ARB_parallel_shader_compile"));
+        private static readonly Lazy<bool> _supportsPolygonOffsetClamp           = new Lazy<bool>(() => HasExtension("GL_EXT_polygon_offset_clamp"));
+        private static readonly Lazy<bool> _supportsQuads                        = new Lazy<bool>(SupportsQuadsCheck);
+        private static readonly Lazy<bool> _supportsSeamlessCubemapPerTexture    = new Lazy<bool>(() => HasExtension("GL_ARB_seamless_cubemap_per_texture"));
+        private static readonly Lazy<bool> _supportsShaderBallot                 = new Lazy<bool>(() => HasExtension("GL_ARB_shader_ballot"));
+        private static readonly Lazy<bool> _supportsShaderViewportLayerArray     = new Lazy<bool>(() => HasExtension("GL_ARB_shader_viewport_layer_array"));
+        private static readonly Lazy<bool> _supportsViewportArray2               = new Lazy<bool>(() => HasExtension("GL_NV_viewport_array2"));
+        private static readonly Lazy<bool> _supportsTextureCompressionBptc       = new Lazy<bool>(() => HasExtension("GL_EXT_texture_compression_bptc"));
+        private static readonly Lazy<bool> _supportsTextureCompressionRgtc       = new Lazy<bool>(() => HasExtension("GL_EXT_texture_compression_rgtc"));
+        private static readonly Lazy<bool> _supportsTextureCompressionS3tc       = new Lazy<bool>(() => HasExtension("GL_EXT_texture_compression_s3tc"));
+        private static readonly Lazy<bool> _supportsTextureShadowLod             = new Lazy<bool>(() => HasExtension("GL_EXT_texture_shadow_lod"));
+        private static readonly Lazy<bool> _supportsViewportSwizzle              = new Lazy<bool>(() => HasExtension("GL_NV_viewport_swizzle"));
+
+        private static readonly Lazy<int> _maximumComputeSharedMemorySize = new Lazy<int>(() => GetLimit(All.MaxComputeSharedMemorySize));
+        private static readonly Lazy<int> _storageBufferOffsetAlignment   = new Lazy<int>(() => GetLimit(All.ShaderStorageBufferOffsetAlignment));
+
+        public enum GpuVendor
+        {
+            Unknown,
+            AmdWindows,
+            AmdUnix,
+            IntelWindows,
+            IntelUnix,
+            Nvidia
+        }
+
+        private static readonly Lazy<GpuVendor> _gpuVendor = new Lazy<GpuVendor>(GetGpuVendor);
+
+        private static bool _isAMD   => _gpuVendor.Value == GpuVendor.AmdWindows || _gpuVendor.Value == GpuVendor.AmdUnix;
+        private static bool _isIntel => _gpuVendor.Value == GpuVendor.IntelWindows || _gpuVendor.Value == GpuVendor.IntelUnix;
+
+        public static GpuVendor Vendor => _gpuVendor.Value;
+
+        private static Lazy<float> _maxSupportedAnisotropy = new Lazy<float>(GL.GetFloat((GetPName)All.MaxTextureMaxAnisotropy));
+
+        public static bool UsePersistentBufferForFlush       => _gpuVendor.Value == GpuVendor.AmdWindows || _gpuVendor.Value == GpuVendor.Nvidia;
+
+        public static bool SupportsAlphaToCoverageDitherControl => _supportsAlphaToCoverageDitherControl.Value;
+        public static bool SupportsAstcCompression              => _supportsAstcCompression.Value;
+        public static bool SupportsBlendEquationAdvanced        => _supportsBlendEquationAdvanced.Value;
+        public static bool SupportsDrawTexture                  => _supportsDrawTexture.Value;
+        public static bool SupportsFragmentShaderInterlock      => _supportsFragmentShaderInterlock.Value;
+        public static bool SupportsFragmentShaderOrdering       => _supportsFragmentShaderOrdering.Value;
+        public static bool SupportsGeometryShaderPassthrough    => _supportsGeometryShaderPassthrough.Value;
+        public static bool SupportsImageLoadFormatted           => _supportsImageLoadFormatted.Value;
+        public static bool SupportsIndirectParameters           => _supportsIndirectParameters.Value;
+        public static bool SupportsParallelShaderCompile        => _supportsParallelShaderCompile.Value;
+        public static bool SupportsPolygonOffsetClamp           => _supportsPolygonOffsetClamp.Value;
+        public static bool SupportsQuads                        => _supportsQuads.Value;
+        public static bool SupportsSeamlessCubemapPerTexture    => _supportsSeamlessCubemapPerTexture.Value;
+        public static bool SupportsShaderBallot                 => _supportsShaderBallot.Value;
+        public static bool SupportsShaderViewportLayerArray     => _supportsShaderViewportLayerArray.Value;
+        public static bool SupportsViewportArray2               => _supportsViewportArray2.Value;
+        public static bool SupportsTextureCompressionBptc       => _supportsTextureCompressionBptc.Value;
+        public static bool SupportsTextureCompressionRgtc       => _supportsTextureCompressionRgtc.Value;
+        public static bool SupportsTextureCompressionS3tc       => _supportsTextureCompressionS3tc.Value;
+        public static bool SupportsTextureShadowLod             => _supportsTextureShadowLod.Value;
+        public static bool SupportsViewportSwizzle              => _supportsViewportSwizzle.Value;
+
+        public static bool SupportsMismatchingViewFormat    => _gpuVendor.Value != GpuVendor.AmdWindows && _gpuVendor.Value != GpuVendor.IntelWindows;
+        public static bool SupportsNonConstantTextureOffset => _gpuVendor.Value == GpuVendor.Nvidia;
+        public static bool RequiresSyncFlush                => _gpuVendor.Value == GpuVendor.AmdWindows || _isIntel;
+
+        public static int MaximumComputeSharedMemorySize => _maximumComputeSharedMemorySize.Value;
+        public static int StorageBufferOffsetAlignment   => _storageBufferOffsetAlignment.Value;
+
+        public static float MaximumSupportedAnisotropy => _maxSupportedAnisotropy.Value;
+
+        private static bool HasExtension(string name)
+        {
+            int numExtensions = GL.GetInteger(GetPName.NumExtensions);
+
+            for (int extension = 0; extension < numExtensions; extension++)
+            {
+                if (GL.GetString(StringNameIndexed.Extensions, extension) == name)
+                {
+                    return true;
+                }
+            }
+
+            return false;
+        }
+
+        private static int GetLimit(All name)
+        {
+            return GL.GetInteger((GetPName)name);
+        }
+
+        private static GpuVendor GetGpuVendor()
+        {
+            string vendor = GL.GetString(StringName.Vendor).ToLower();
+
+            if (vendor == "nvidia corporation")
+            {
+                return GpuVendor.Nvidia;
+            }
+            else if (vendor == "intel")
+            {
+                string renderer = GL.GetString(StringName.Renderer).ToLower();
+
+                return renderer.Contains("mesa") ? GpuVendor.IntelUnix : GpuVendor.IntelWindows;
+            }
+            else if (vendor == "ati technologies inc." || vendor == "advanced micro devices, inc.")
+            {
+                return GpuVendor.AmdWindows;
+            }
+            else if (vendor == "amd" || vendor == "x.org")
+            {
+                return GpuVendor.AmdUnix;
+            }
+            else
+            {
+                return GpuVendor.Unknown;
+            }
+        }
+
+        private static bool SupportsQuadsCheck()
+        {
+            GL.GetError(); // Clear any existing error.
+            GL.Begin(PrimitiveType.Quads);
+            GL.End();
+
+            return GL.GetError() == ErrorCode.NoError;
+        }
+    }
+}
\ No newline at end of file
diff --git a/src/Ryujinx.Graphics.OpenGL/IOpenGLContext.cs b/src/Ryujinx.Graphics.OpenGL/IOpenGLContext.cs
new file mode 100644
index 00000000..b1f6d72d
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/IOpenGLContext.cs
@@ -0,0 +1,27 @@
+using Ryujinx.Graphics.OpenGL.Helper;
+using System;
+
+namespace Ryujinx.Graphics.OpenGL
+{
+    public interface IOpenGLContext : IDisposable
+    {
+        void MakeCurrent();
+
+        // TODO: Support more APIs per platform.
+        static bool HasContext()
+        {
+            if (OperatingSystem.IsWindows())
+            {
+                return WGLHelper.GetCurrentContext() != IntPtr.Zero;
+            }
+            else if (OperatingSystem.IsLinux())
+            {
+                return GLXHelper.GetCurrentContext() != IntPtr.Zero;
+            }
+            else
+            {
+                return false;
+            }
+        }
+    }
+}
diff --git a/src/Ryujinx.Graphics.OpenGL/Image/FormatConverter.cs b/src/Ryujinx.Graphics.OpenGL/Image/FormatConverter.cs
new file mode 100644
index 00000000..c4bbf745
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Image/FormatConverter.cs
@@ -0,0 +1,149 @@
+using System;
+using System.Numerics;
+using System.Runtime.InteropServices;
+using System.Runtime.Intrinsics;
+using System.Runtime.Intrinsics.X86;
+
+namespace Ryujinx.Graphics.OpenGL.Image
+{
+    static class FormatConverter
+    {
+        public unsafe static byte[] ConvertS8D24ToD24S8(ReadOnlySpan<byte> data)
+        {
+            byte[] output = new byte[data.Length];
+
+            int start = 0;
+
+            if (Avx2.IsSupported)
+            {
+                var mask = Vector256.Create(
+                    (byte)3, (byte)0, (byte)1, (byte)2,
+                    (byte)7, (byte)4, (byte)5, (byte)6,
+                    (byte)11, (byte)8, (byte)9, (byte)10,
+                    (byte)15, (byte)12, (byte)13, (byte)14,
+                    (byte)19, (byte)16, (byte)17, (byte)18,
+                    (byte)23, (byte)20, (byte)21, (byte)22,
+                    (byte)27, (byte)24, (byte)25, (byte)26,
+                    (byte)31, (byte)28, (byte)29, (byte)30);
+
+                int sizeAligned = data.Length & ~31;
+
+                fixed (byte* pInput = data, pOutput = output)
+                {
+                    for (uint i = 0; i < sizeAligned; i += 32)
+                    {
+                        var dataVec = Avx.LoadVector256(pInput + i);
+
+                        dataVec = Avx2.Shuffle(dataVec, mask);
+
+                        Avx.Store(pOutput + i, dataVec);
+                    }
+                }
+
+                start = sizeAligned;
+            }
+            else if (Ssse3.IsSupported)
+            {
+                var mask = Vector128.Create(
+                    (byte)3, (byte)0, (byte)1, (byte)2,
+                    (byte)7, (byte)4, (byte)5, (byte)6,
+                    (byte)11, (byte)8, (byte)9, (byte)10,
+                    (byte)15, (byte)12, (byte)13, (byte)14);
+
+                int sizeAligned = data.Length & ~15;
+
+                fixed (byte* pInput = data, pOutput = output)
+                {
+                    for (uint i = 0; i < sizeAligned; i += 16)
+                    {
+                        var dataVec = Sse2.LoadVector128(pInput + i);
+
+                        dataVec = Ssse3.Shuffle(dataVec, mask);
+
+                        Sse2.Store(pOutput + i, dataVec);
+                    }
+                }
+
+                start = sizeAligned;
+            }
+
+            var outSpan = MemoryMarshal.Cast<byte, uint>(output);
+            var dataSpan = MemoryMarshal.Cast<byte, uint>(data);
+            for (int i = start / sizeof(uint); i < dataSpan.Length; i++)
+            {
+                outSpan[i] = BitOperations.RotateLeft(dataSpan[i], 8);
+            }
+
+            return output;
+        }
+
+        public unsafe static byte[] ConvertD24S8ToS8D24(ReadOnlySpan<byte> data)
+        {
+            byte[] output = new byte[data.Length];
+
+            int start = 0;
+
+            if (Avx2.IsSupported)
+            {
+                var mask = Vector256.Create(
+                    (byte)1, (byte)2, (byte)3, (byte)0,
+                    (byte)5, (byte)6, (byte)7, (byte)4,
+                    (byte)9, (byte)10, (byte)11, (byte)8,
+                    (byte)13, (byte)14, (byte)15, (byte)12,
+                    (byte)17, (byte)18, (byte)19, (byte)16,
+                    (byte)21, (byte)22, (byte)23, (byte)20,
+                    (byte)25, (byte)26, (byte)27, (byte)24,
+                    (byte)29, (byte)30, (byte)31, (byte)28);
+
+                int sizeAligned = data.Length & ~31;
+
+                fixed (byte* pInput = data, pOutput = output)
+                {
+                    for (uint i = 0; i < sizeAligned; i += 32)
+                    {
+                        var dataVec = Avx.LoadVector256(pInput + i);
+
+                        dataVec = Avx2.Shuffle(dataVec, mask);
+
+                        Avx.Store(pOutput + i, dataVec);
+                    }
+                }
+
+                start = sizeAligned;
+            }
+            else if (Ssse3.IsSupported)
+            {
+                var mask = Vector128.Create(
+                    (byte)1, (byte)2, (byte)3, (byte)0,
+                    (byte)5, (byte)6, (byte)7, (byte)4,
+                    (byte)9, (byte)10, (byte)11, (byte)8,
+                    (byte)13, (byte)14, (byte)15, (byte)12);
+
+                int sizeAligned = data.Length & ~15;
+
+                fixed (byte* pInput = data, pOutput = output)
+                {
+                    for (uint i = 0; i < sizeAligned; i += 16)
+                    {
+                        var dataVec = Sse2.LoadVector128(pInput + i);
+
+                        dataVec = Ssse3.Shuffle(dataVec, mask);
+
+                        Sse2.Store(pOutput + i, dataVec);
+                    }
+                }
+
+                start = sizeAligned;
+            }
+
+            var outSpan = MemoryMarshal.Cast<byte, uint>(output);
+            var dataSpan = MemoryMarshal.Cast<byte, uint>(data);
+            for (int i = start / sizeof(uint); i < dataSpan.Length; i++)
+            {
+                outSpan[i] = BitOperations.RotateRight(dataSpan[i], 8);
+            }
+
+            return output;
+        }
+    }
+}
diff --git a/src/Ryujinx.Graphics.OpenGL/Image/ITextureInfo.cs b/src/Ryujinx.Graphics.OpenGL/Image/ITextureInfo.cs
new file mode 100644
index 00000000..4c8d7fef
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Image/ITextureInfo.cs
@@ -0,0 +1,14 @@
+using Ryujinx.Graphics.GAL;
+
+namespace Ryujinx.Graphics.OpenGL.Image
+{
+    interface ITextureInfo
+    {
+        ITextureInfo Storage { get; }
+        int Handle { get; }
+        int FirstLayer => 0;
+        int FirstLevel => 0;
+
+        TextureCreateInfo Info { get; }
+    }
+}
diff --git a/src/Ryujinx.Graphics.OpenGL/Image/IntermmediatePool.cs b/src/Ryujinx.Graphics.OpenGL/Image/IntermmediatePool.cs
new file mode 100644
index 00000000..4f167e89
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Image/IntermmediatePool.cs
@@ -0,0 +1,103 @@
+using Ryujinx.Graphics.GAL;
+using System;
+using System.Collections.Generic;
+
+namespace Ryujinx.Graphics.OpenGL.Image
+{
+    class IntermediatePool : IDisposable
+    {
+        private readonly OpenGLRenderer _renderer;
+        private readonly List<TextureView> _entries;
+
+        public IntermediatePool(OpenGLRenderer renderer)
+        {
+            _renderer = renderer;
+            _entries = new List<TextureView>();
+        }
+
+        public TextureView GetOrCreateWithAtLeast(
+            Target target,
+            int blockWidth,
+            int blockHeight,
+            int bytesPerPixel,
+            Format format,
+            int width,
+            int height,
+            int depth,
+            int levels,
+            int samples)
+        {
+            TextureView entry;
+
+            for (int i = 0; i < _entries.Count; i++)
+            {
+                entry = _entries[i];
+
+                if (entry.Target == target && entry.Format == format && entry.Info.Samples == samples)
+                {
+                    if (entry.Width < width ||
+                        entry.Height < height ||
+                        entry.Info.Depth < depth ||
+                        entry.Info.Levels < levels)
+                    {
+                        width = Math.Max(width, entry.Width);
+                        height = Math.Max(height, entry.Height);
+                        depth = Math.Max(depth, entry.Info.Depth);
+                        levels = Math.Max(levels, entry.Info.Levels);
+
+                        entry.Dispose();
+                        entry = CreateNew(target, blockWidth, blockHeight, bytesPerPixel, format, width, height, depth, levels, samples);
+                        _entries[i] = entry;
+                    }
+
+                    return entry;
+                }
+            }
+
+            entry = CreateNew(target, blockWidth, blockHeight, bytesPerPixel, format, width, height, depth, levels, samples);
+            _entries.Add(entry);
+
+            return entry;
+        }
+
+        private TextureView CreateNew(
+            Target target,
+            int blockWidth,
+            int blockHeight,
+            int bytesPerPixel,
+            Format format,
+            int width,
+            int height,
+            int depth,
+            int levels,
+            int samples)
+        {
+            return (TextureView)_renderer.CreateTexture(new TextureCreateInfo(
+                width,
+                height,
+                depth,
+                levels,
+                samples,
+                blockWidth,
+                blockHeight,
+                bytesPerPixel,
+                format,
+                DepthStencilMode.Depth,
+                target,
+                SwizzleComponent.Red,
+                SwizzleComponent.Green,
+                SwizzleComponent.Blue,
+                SwizzleComponent.Alpha), 1f);
+        }
+
+        public void Dispose()
+        {
+            foreach (TextureView entry in _entries)
+            {
+                entry.Dispose();
+            }
+
+            _entries.Clear();
+        }
+    }
+}
\ No newline at end of file
diff --git a/src/Ryujinx.Graphics.OpenGL/Image/Sampler.cs b/src/Ryujinx.Graphics.OpenGL/Image/Sampler.cs
new file mode 100644
index 00000000..f705aa3e
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Image/Sampler.cs
@@ -0,0 +1,64 @@
+using OpenTK.Graphics.OpenGL;
+using Ryujinx.Graphics.GAL;
+
+namespace Ryujinx.Graphics.OpenGL.Image
+{
+    class Sampler : ISampler
+    {
+        public int Handle { get; private set; }
+
+        public Sampler(SamplerCreateInfo info)
+        {
+            Handle = GL.GenSampler();
+
+            GL.SamplerParameter(Handle, SamplerParameterName.TextureMinFilter, (int)info.MinFilter.Convert());
+            GL.SamplerParameter(Handle, SamplerParameterName.TextureMagFilter, (int)info.MagFilter.Convert());
+
+            if (HwCapabilities.SupportsSeamlessCubemapPerTexture)
+            {
+                GL.SamplerParameter(Handle, (SamplerParameterName)ArbSeamlessCubemapPerTexture.TextureCubeMapSeamless, info.SeamlessCubemap ? 1 : 0);
+            }
+
+            GL.SamplerParameter(Handle, SamplerParameterName.TextureWrapS, (int)info.AddressU.Convert());
+            GL.SamplerParameter(Handle, SamplerParameterName.TextureWrapT, (int)info.AddressV.Convert());
+            GL.SamplerParameter(Handle, SamplerParameterName.TextureWrapR, (int)info.AddressP.Convert());
+
+            GL.SamplerParameter(Handle, SamplerParameterName.TextureCompareMode, (int)info.CompareMode.Convert());
+            GL.SamplerParameter(Handle, SamplerParameterName.TextureCompareFunc, (int)info.CompareOp.Convert());
+
+            unsafe
+            {
+                float* borderColor = stackalloc float[4]
+                {
+                    info.BorderColor.Red,
+                    info.BorderColor.Green,
+                    info.BorderColor.Blue,
+                    info.BorderColor.Alpha
+                };
+
+                GL.SamplerParameter(Handle, SamplerParameterName.TextureBorderColor, borderColor);
+            }
+
+            GL.SamplerParameter(Handle, SamplerParameterName.TextureMinLod,  info.MinLod);
+            GL.SamplerParameter(Handle, SamplerParameterName.TextureMaxLod,  info.MaxLod);
+            GL.SamplerParameter(Handle, SamplerParameterName.TextureLodBias, info.MipLodBias);
+
+            GL.SamplerParameter(Handle, SamplerParameterName.TextureMaxAnisotropyExt, info.MaxAnisotropy);
+        }
+
+        public void Bind(int unit)
+        {
+            GL.BindSampler(unit, Handle);
+        }
+
+        public void Dispose()
+        {
+            if (Handle != 0)
+            {
+                GL.DeleteSampler(Handle);
+
+                Handle = 0;
+            }
+        }
+    }
+}
diff --git a/src/Ryujinx.Graphics.OpenGL/Image/TextureBase.cs b/src/Ryujinx.Graphics.OpenGL/Image/TextureBase.cs
new file mode 100644
index 00000000..2ab9dffb
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Image/TextureBase.cs
@@ -0,0 +1,44 @@
+using OpenTK.Graphics.OpenGL;
+using Ryujinx.Graphics.GAL;
+
+namespace Ryujinx.Graphics.OpenGL.Image
+{
+    class TextureBase
+    {
+        public int Handle { get; protected set; }
+
+        public TextureCreateInfo Info { get; }
+
+        public int Width => Info.Width;
+        public int Height => Info.Height;
+        public float ScaleFactor { get; }
+
+        public Target Target => Info.Target;
+        public Format Format => Info.Format;
+
+        public TextureBase(TextureCreateInfo info, float scaleFactor = 1f)
+        {
+            Info = info;
+            ScaleFactor = scaleFactor;
+
+            Handle = GL.GenTexture();
+        }
+
+        public void Bind(int unit)
+        {
+            Bind(Target.Convert(), unit);
+        }
+
+        protected void Bind(TextureTarget target, int unit)
+        {
+            GL.ActiveTexture(TextureUnit.Texture0 + unit);
+            GL.BindTexture(target, Handle);
+        }
+
+        public static void ClearBinding(int unit)
+        {
+            GL.ActiveTexture(TextureUnit.Texture0 + unit);
+            GL.BindTextureUnit(unit, 0);
+        }
+    }
+}
diff --git a/src/Ryujinx.Graphics.OpenGL/Image/TextureBuffer.cs b/src/Ryujinx.Graphics.OpenGL/Image/TextureBuffer.cs
new file mode 100644
index 00000000..1e9e4d6b
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Image/TextureBuffer.cs
@@ -0,0 +1,108 @@
+using OpenTK.Graphics.OpenGL;
+using Ryujinx.Common.Memory;
+using Ryujinx.Graphics.GAL;
+using System;
+
+namespace Ryujinx.Graphics.OpenGL.Image
+{
+    class TextureBuffer : TextureBase, ITexture
+    {
+        private OpenGLRenderer _renderer;
+        private int _bufferOffset;
+        private int _bufferSize;
+        private int _bufferCount;
+
+        private BufferHandle _buffer;
+
+        public TextureBuffer(OpenGLRenderer renderer, TextureCreateInfo info) : base(info)
+        {
+            _renderer = renderer;
+        }
+
+        public void CopyTo(ITexture destination, int firstLayer, int firstLevel)
+        {
+            throw new NotSupportedException();
+        }
+
+        public void CopyTo(ITexture destination, int srcLayer, int dstLayer, int srcLevel, int dstLevel)
+        {
+            throw new NotSupportedException();
+        }
+
+        public void CopyTo(ITexture destination, Extents2D srcRegion, Extents2D dstRegion, bool linearFilter)
+        {
+            throw new NotSupportedException();
+        }
+
+        public ITexture CreateView(TextureCreateInfo info, int firstLayer, int firstLevel)
+        {
+            throw new NotSupportedException();
+        }
+
+        public PinnedSpan<byte> GetData()
+        {
+            return Buffer.GetData(_renderer, _buffer, _bufferOffset, _bufferSize);
+        }
+
+        public PinnedSpan<byte> GetData(int layer, int level)
+        {
+            return GetData();
+        }
+
+        public void SetData(SpanOrArray<byte> data)
+        {
+            var dataSpan = data.AsSpan();
+
+            Buffer.SetData(_buffer, _bufferOffset, dataSpan.Slice(0, Math.Min(dataSpan.Length, _bufferSize)));
+        }
+
+        public void SetData(SpanOrArray<byte> data, int layer, int level)
+        {
+            throw new NotSupportedException();
+        }
+
+        public void SetData(SpanOrArray<byte> data, int layer, int level, Rectangle<int> region)
+        {
+            throw new NotSupportedException();
+        }
+
+        public void SetStorage(BufferRange buffer)
+        {
+            if (_buffer != BufferHandle.Null &&
+                _buffer == buffer.Handle &&
+                buffer.Offset == _bufferOffset &&
+                buffer.Size == _bufferSize &&
+                _renderer.BufferCount == _bufferCount)
+            {
+                // Only rebind the buffer when more have been created.
+                return;
+            }
+
+            _buffer = buffer.Handle;
+            _bufferOffset = buffer.Offset;
+            _bufferSize = buffer.Size;
+            _bufferCount = _renderer.BufferCount;
+
+            Bind(0);
+
+            SizedInternalFormat format = (SizedInternalFormat)FormatTable.GetFormatInfo(Info.Format).PixelInternalFormat;
+
+            GL.TexBufferRange(TextureBufferTarget.TextureBuffer, format, _buffer.ToInt32(), (IntPtr)buffer.Offset, buffer.Size);
+        }
+
+        public void Dispose()
+        {
+            if (Handle != 0)
+            {
+                GL.DeleteTexture(Handle);
+
+                Handle = 0;
+            }
+        }
+
+        public void Release()
+        {
+            Dispose();
+        }
+    }
+}
diff --git a/src/Ryujinx.Graphics.OpenGL/Image/TextureCopy.cs b/src/Ryujinx.Graphics.OpenGL/Image/TextureCopy.cs
new file mode 100644
index 00000000..a4b08787
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Image/TextureCopy.cs
@@ -0,0 +1,524 @@
+using OpenTK.Graphics.OpenGL;
+using Ryujinx.Common;
+using Ryujinx.Graphics.GAL;
+using System;
+
+namespace Ryujinx.Graphics.OpenGL.Image
+{
+    class TextureCopy : IDisposable
+    {
+        private readonly OpenGLRenderer _renderer;
+
+        private int _srcFramebuffer;
+        private int _dstFramebuffer;
+
+        private int _copyPboHandle;
+        private int _copyPboSize;
+
+        public IntermediatePool IntermediatePool { get; }
+
+        public TextureCopy(OpenGLRenderer renderer)
+        {
+            _renderer = renderer;
+            IntermediatePool = new IntermediatePool(renderer);
+        }
+
+        public void Copy(
+            TextureView src,
+            TextureView dst,
+            Extents2D   srcRegion,
+            Extents2D   dstRegion,
+            bool        linearFilter,
+            int         srcLayer = 0,
+            int         dstLayer = 0,
+            int         srcLevel = 0,
+            int         dstLevel = 0)
+        {
+            int levels = Math.Min(src.Info.Levels - srcLevel, dst.Info.Levels - dstLevel);
+            int layers = Math.Min(src.Info.GetLayers() - srcLayer, dst.Info.GetLayers() - dstLayer);
+
+            Copy(src, dst, srcRegion, dstRegion, linearFilter, srcLayer, dstLayer, srcLevel, dstLevel, layers, levels);
+        }
+
+        public void Copy(
+            TextureView src,
+            TextureView dst,
+            Extents2D   srcRegion,
+            Extents2D   dstRegion,
+            bool        linearFilter,
+            int         srcLayer,
+            int         dstLayer,
+            int         srcLevel,
+            int         dstLevel,
+            int         layers,
+            int         levels)
+        {
+            TextureView srcConverted = src.Format.IsBgr() != dst.Format.IsBgr() ? BgraSwap(src) : src;
+
+            (int oldDrawFramebufferHandle, int oldReadFramebufferHandle) = ((Pipeline)_renderer.Pipeline).GetBoundFramebuffers();
+
+            GL.BindFramebuffer(FramebufferTarget.ReadFramebuffer, GetSrcFramebufferLazy());
+            GL.BindFramebuffer(FramebufferTarget.DrawFramebuffer, GetDstFramebufferLazy());
+
+            if (srcLevel != 0)
+            {
+                srcRegion = srcRegion.Reduce(srcLevel);
+            }
+
+            if (dstLevel != 0)
+            {
+                dstRegion = dstRegion.Reduce(dstLevel);
+            }
+
+            for (int level = 0; level < levels; level++)
+            {
+                for (int layer = 0; layer < layers; layer++)
+                {
+                    if ((srcLayer | dstLayer) != 0 || layers > 1)
+                    {
+                        Attach(FramebufferTarget.ReadFramebuffer, src.Format, srcConverted.Handle, srcLevel + level, srcLayer + layer);
+                        Attach(FramebufferTarget.DrawFramebuffer, dst.Format, dst.Handle, dstLevel + level, dstLayer + layer);
+                    }
+                    else
+                    {
+                        Attach(FramebufferTarget.ReadFramebuffer, src.Format, srcConverted.Handle, srcLevel + level);
+                        Attach(FramebufferTarget.DrawFramebuffer, dst.Format, dst.Handle, dstLevel + level);
+                    }
+
+                    ClearBufferMask mask = GetMask(src.Format);
+
+                    if ((mask & (ClearBufferMask.DepthBufferBit | ClearBufferMask.StencilBufferBit)) != 0 || src.Format.IsInteger())
+                    {
+                        linearFilter = false;
+                    }
+
+                    BlitFramebufferFilter filter = linearFilter
+                        ? BlitFramebufferFilter.Linear
+                        : BlitFramebufferFilter.Nearest;
+
+                    GL.ReadBuffer(ReadBufferMode.ColorAttachment0);
+                    GL.DrawBuffer(DrawBufferMode.ColorAttachment0);
+
+                    GL.Disable(EnableCap.RasterizerDiscard);
+                    GL.Disable(IndexedEnableCap.ScissorTest, 0);
+
+                    GL.BlitFramebuffer(
+                        srcRegion.X1,
+                        srcRegion.Y1,
+                        srcRegion.X2,
+                        srcRegion.Y2,
+                        dstRegion.X1,
+                        dstRegion.Y1,
+                        dstRegion.X2,
+                        dstRegion.Y2,
+                        mask,
+                        filter);
+                }
+
+                if (level < levels - 1)
+                {
+                    srcRegion = srcRegion.Reduce(1);
+                    dstRegion = dstRegion.Reduce(1);
+                }
+            }
+
+            Attach(FramebufferTarget.ReadFramebuffer, src.Format, 0);
+            Attach(FramebufferTarget.DrawFramebuffer, dst.Format, 0);
+
+            GL.BindFramebuffer(FramebufferTarget.ReadFramebuffer, oldReadFramebufferHandle);
+            GL.BindFramebuffer(FramebufferTarget.DrawFramebuffer, oldDrawFramebufferHandle);
+
+            ((Pipeline)_renderer.Pipeline).RestoreScissor0Enable();
+            ((Pipeline)_renderer.Pipeline).RestoreRasterizerDiscard();
+
+            if (srcConverted != src)
+            {
+                srcConverted.Dispose();
+            }
+        }
+
+        public void CopyUnscaled(
+            ITextureInfo src,
+            ITextureInfo dst,
+            int srcLayer,
+            int dstLayer,
+            int srcLevel,
+            int dstLevel)
+        {
+            TextureCreateInfo srcInfo = src.Info;
+            TextureCreateInfo dstInfo = dst.Info;
+
+            int srcDepth = srcInfo.GetDepthOrLayers();
+            int srcLevels = srcInfo.Levels;
+
+            int dstDepth = dstInfo.GetDepthOrLayers();
+            int dstLevels = dstInfo.Levels;
+
+            if (dstInfo.Target == Target.Texture3D)
+            {
+                dstDepth = Math.Max(1, dstDepth >> dstLevel);
+            }
+
+            int depth = Math.Min(srcDepth, dstDepth);
+            int levels = Math.Min(srcLevels, dstLevels);
+
+            CopyUnscaled(src, dst, srcLayer, dstLayer, srcLevel, dstLevel, depth, levels);
+        }
+
+        public void CopyUnscaled(
+            ITextureInfo src,
+            ITextureInfo dst,
+            int srcLayer,
+            int dstLayer,
+            int srcLevel,
+            int dstLevel,
+            int depth,
+            int levels)
+        {
+            TextureCreateInfo srcInfo = src.Info;
+            TextureCreateInfo dstInfo = dst.Info;
+
+            int srcHandle = src.Handle;
+            int dstHandle = dst.Handle;
+
+            int srcWidth = srcInfo.Width;
+            int srcHeight = srcInfo.Height;
+
+            int dstWidth = dstInfo.Width;
+            int dstHeight = dstInfo.Height;
+
+            srcWidth = Math.Max(1, srcWidth >> srcLevel);
+            srcHeight = Math.Max(1, srcHeight >> srcLevel);
+
+            dstWidth = Math.Max(1, dstWidth >> dstLevel);
+            dstHeight = Math.Max(1, dstHeight >> dstLevel);
+
+            int blockWidth = 1;
+            int blockHeight = 1;
+            bool sizeInBlocks = false;
+
+            // When copying from a compressed to a non-compressed format,
+            // the non-compressed texture will have the size of the texture
+            // in blocks (not in texels), so we must adjust that size to
+            // match the size in texels of the compressed texture.
+            if (!srcInfo.IsCompressed && dstInfo.IsCompressed)
+            {
+                srcWidth *= dstInfo.BlockWidth;
+                srcHeight *= dstInfo.BlockHeight;
+                blockWidth = dstInfo.BlockWidth;
+                blockHeight = dstInfo.BlockHeight;
+
+                sizeInBlocks = true;
+            }
+            else if (srcInfo.IsCompressed && !dstInfo.IsCompressed)
+            {
+                dstWidth *= srcInfo.BlockWidth;
+                dstHeight *= srcInfo.BlockHeight;
+                blockWidth = srcInfo.BlockWidth;
+                blockHeight = srcInfo.BlockHeight;
+            }
+
+            int width = Math.Min(srcWidth, dstWidth);
+            int height = Math.Min(srcHeight, dstHeight);
+
+            for (int level = 0; level < levels; level++)
+            {
+                // Stop copy if we are already out of the levels range.
+                if (level >= srcInfo.Levels || dstLevel + level >= dstInfo.Levels)
+                {
+                    break;
+                }
+
+                if ((width % blockWidth != 0 || height % blockHeight != 0) && src is TextureView srcView && dst is TextureView dstView)
+                {
+                    PboCopy(srcView, dstView, srcLayer, dstLayer, srcLevel + level, dstLevel + level, width, height);
+                }
+                else
+                {
+                    int copyWidth = sizeInBlocks ? BitUtils.DivRoundUp(width, blockWidth) : width;
+                    int copyHeight = sizeInBlocks ? BitUtils.DivRoundUp(height, blockHeight) : height;
+
+                    if (HwCapabilities.Vendor == HwCapabilities.GpuVendor.IntelWindows)
+                    {
+                        GL.CopyImageSubData(
+                            src.Storage.Handle,
+                            src.Storage.Info.Target.ConvertToImageTarget(),
+                            src.FirstLevel + srcLevel + level,
+                            0,
+                            0,
+                            src.FirstLayer + srcLayer,
+                            dst.Storage.Handle,
+                            dst.Storage.Info.Target.ConvertToImageTarget(),
+                            dst.FirstLevel + dstLevel + level,
+                            0,
+                            0,
+                            dst.FirstLayer + dstLayer,
+                            copyWidth,
+                            copyHeight,
+                            depth);
+                    }
+                    else
+                    {
+                        GL.CopyImageSubData(
+                            srcHandle,
+                            srcInfo.Target.ConvertToImageTarget(),
+                            srcLevel + level,
+                            0,
+                            0,
+                            srcLayer,
+                            dstHandle,
+                            dstInfo.Target.ConvertToImageTarget(),
+                            dstLevel + level,
+                            0,
+                            0,
+                            dstLayer,
+                            copyWidth,
+                            copyHeight,
+                            depth);
+                    }
+                }
+
+                width = Math.Max(1, width >> 1);
+                height = Math.Max(1, height >> 1);
+
+                if (srcInfo.Target == Target.Texture3D)
+                {
+                    depth = Math.Max(1, depth >> 1);
+                }
+            }
+        }
+
+        private static FramebufferAttachment AttachmentForFormat(Format format)
+        {
+            if (format == Format.D24UnormS8Uint || format == Format.D32FloatS8Uint)
+            {
+                return FramebufferAttachment.DepthStencilAttachment;
+            }
+            else if (IsDepthOnly(format))
+            {
+                return FramebufferAttachment.DepthAttachment;
+            }
+            else if (format == Format.S8Uint)
+            {
+                return FramebufferAttachment.StencilAttachment;
+            }
+            else
+            {
+                return FramebufferAttachment.ColorAttachment0;
+            }
+        }
+
+        private static void Attach(FramebufferTarget target, Format format, int handle, int level = 0)
+        {
+            FramebufferAttachment attachment = AttachmentForFormat(format);
+
+            GL.FramebufferTexture(target, attachment, handle, level);
+        }
+
+        private static void Attach(FramebufferTarget target, Format format, int handle, int level, int layer)
+        {
+            FramebufferAttachment attachment = AttachmentForFormat(format);
+
+            GL.FramebufferTextureLayer(target, attachment, handle, level, layer);
+        }
+
+        private static ClearBufferMask GetMask(Format format)
+        {
+            if (format == Format.D24UnormS8Uint || format == Format.D32FloatS8Uint || format == Format.S8UintD24Unorm)
+            {
+                return ClearBufferMask.DepthBufferBit | ClearBufferMask.StencilBufferBit;
+            }
+            else if (IsDepthOnly(format))
+            {
+                return ClearBufferMask.DepthBufferBit;
+            }
+            else if (format == Format.S8Uint)
+            {
+                return ClearBufferMask.StencilBufferBit;
+            }
+            else
+            {
+                return ClearBufferMask.ColorBufferBit;
+            }
+        }
+
+        private static bool IsDepthOnly(Format format)
+        {
+            return format == Format.D16Unorm || format == Format.D32Float;
+        }
+
+        public TextureView BgraSwap(TextureView from)
+        {
+            TextureView to = (TextureView)_renderer.CreateTexture(from.Info, from.ScaleFactor);
+
+            EnsurePbo(from);
+
+            GL.BindBuffer(BufferTarget.PixelPackBuffer, _copyPboHandle);
+
+            from.WriteToPbo(0, forceBgra: true);
+
+            GL.BindBuffer(BufferTarget.PixelPackBuffer, 0);
+            GL.BindBuffer(BufferTarget.PixelUnpackBuffer, _copyPboHandle);
+
+            to.ReadFromPbo(0, _copyPboSize);
+
+            GL.BindBuffer(BufferTarget.PixelUnpackBuffer, 0);
+
+            return to;
+        }
+
+        private TextureView PboCopy(TextureView from, TextureView to, int srcLayer, int dstLayer, int srcLevel, int dstLevel, int width, int height)
+        {
+            int dstWidth = width;
+            int dstHeight = height;
+
+            // The size of the source texture.
+            int unpackWidth = from.Width;
+            int unpackHeight = from.Height;
+
+            if (from.Info.IsCompressed != to.Info.IsCompressed)
+            {
+                if (from.Info.IsCompressed)
+                {
+                    // Dest size is in pixels, but should be in blocks
+                    dstWidth = BitUtils.DivRoundUp(width, from.Info.BlockWidth);
+                    dstHeight = BitUtils.DivRoundUp(height, from.Info.BlockHeight);
+
+                    // When copying from a compressed texture, the source size must be taken in blocks for unpacking to the uncompressed block texture.
+                    unpackWidth = BitUtils.DivRoundUp(from.Info.Width, from.Info.BlockWidth);
+                    unpackHeight = BitUtils.DivRoundUp(from.Info.Height, from.Info.BlockHeight);
+                }
+                else
+                {
+                    // When copying to a compressed texture, the source size must be scaled by the block width for unpacking on the compressed target.
+                    unpackWidth = from.Info.Width * to.Info.BlockWidth;
+                    unpackHeight = from.Info.Height * to.Info.BlockHeight;
+                }
+            }
+
+            EnsurePbo(from);
+
+            GL.BindBuffer(BufferTarget.PixelPackBuffer, _copyPboHandle);
+
+            // The source texture is written out in full, then the destination is taken as a slice from the data using unpack params.
+            // The offset points to the base at which the requested layer is at.
+
+            int offset = from.WriteToPbo2D(0, srcLayer, srcLevel);
+
+            // If the destination size is not an exact match for the source unpack parameters, we need to set them to slice the data correctly.
+
+            bool slice = (unpackWidth != dstWidth || unpackHeight != dstHeight);
+
+            if (slice)
+            {
+                // Set unpack parameters to take a slice of width/height:
+                GL.PixelStore(PixelStoreParameter.UnpackRowLength, unpackWidth);
+                GL.PixelStore(PixelStoreParameter.UnpackImageHeight, unpackHeight);
+
+                if (to.Info.IsCompressed)
+                {
+                    GL.PixelStore(PixelStoreParameter.UnpackCompressedBlockWidth, to.Info.BlockWidth);
+                    GL.PixelStore(PixelStoreParameter.UnpackCompressedBlockHeight, to.Info.BlockHeight);
+                    GL.PixelStore(PixelStoreParameter.UnpackCompressedBlockDepth, 1);
+                    GL.PixelStore(PixelStoreParameter.UnpackCompressedBlockSize, to.Info.BytesPerPixel);
+                }
+            }
+
+            GL.BindBuffer(BufferTarget.PixelPackBuffer, 0);
+            GL.BindBuffer(BufferTarget.PixelUnpackBuffer, _copyPboHandle);
+
+            to.ReadFromPbo2D(offset, dstLayer, dstLevel, dstWidth, dstHeight);
+
+            if (slice)
+            {
+                // Reset unpack parameters
+                GL.PixelStore(PixelStoreParameter.UnpackRowLength, 0);
+                GL.PixelStore(PixelStoreParameter.UnpackImageHeight, 0);
+
+                if (to.Info.IsCompressed)
+                {
+                    GL.PixelStore(PixelStoreParameter.UnpackCompressedBlockWidth, 0);
+                    GL.PixelStore(PixelStoreParameter.UnpackCompressedBlockHeight, 0);
+                    GL.PixelStore(PixelStoreParameter.UnpackCompressedBlockDepth, 0);
+                    GL.PixelStore(PixelStoreParameter.UnpackCompressedBlockSize, 0);
+                }
+            }
+
+            GL.BindBuffer(BufferTarget.PixelUnpackBuffer, 0);
+
+            return to;
+        }
+
+        private void EnsurePbo(TextureView view)
+        {
+            int requiredSize = 0;
+
+            for (int level = 0; level < view.Info.Levels; level++)
+            {
+                requiredSize += view.Info.GetMipSize(level);
+            }
+
+            if (_copyPboSize < requiredSize && _copyPboHandle != 0)
+            {
+                GL.DeleteBuffer(_copyPboHandle);
+
+                _copyPboHandle = 0;
+            }
+
+            if (_copyPboHandle == 0)
+            {
+                _copyPboHandle = GL.GenBuffer();
+                _copyPboSize = requiredSize;
+
+                GL.BindBuffer(BufferTarget.PixelPackBuffer, _copyPboHandle);
+                GL.BufferData(BufferTarget.PixelPackBuffer, requiredSize, IntPtr.Zero, BufferUsageHint.DynamicCopy);
+            }
+        }
+
+        private int GetSrcFramebufferLazy()
+        {
+            if (_srcFramebuffer == 0)
+            {
+                _srcFramebuffer = GL.GenFramebuffer();
+            }
+
+            return _srcFramebuffer;
+        }
+
+        private int GetDstFramebufferLazy()
+        {
+            if (_dstFramebuffer == 0)
+            {
+                _dstFramebuffer = GL.GenFramebuffer();
+            }
+
+            return _dstFramebuffer;
+        }
+
+        public void Dispose()
+        {
+            if (_srcFramebuffer != 0)
+            {
+                GL.DeleteFramebuffer(_srcFramebuffer);
+
+                _srcFramebuffer = 0;
+            }
+
+            if (_dstFramebuffer != 0)
+            {
+                GL.DeleteFramebuffer(_dstFramebuffer);
+
+                _dstFramebuffer = 0;
+            }
+
+            if (_copyPboHandle != 0)
+            {
+                GL.DeleteBuffer(_copyPboHandle);
+
+                _copyPboHandle = 0;
+            }
+
+            IntermediatePool.Dispose();
+        }
+    }
+}
diff --git a/src/Ryujinx.Graphics.OpenGL/Image/TextureCopyIncompatible.cs b/src/Ryujinx.Graphics.OpenGL/Image/TextureCopyIncompatible.cs
new file mode 100644
index 00000000..c8fbfbc6
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Image/TextureCopyIncompatible.cs
@@ -0,0 +1,252 @@
+using OpenTK.Graphics.OpenGL;
+using Ryujinx.Graphics.GAL;
+using System;
+using System.Collections.Generic;
+using System.Globalization;
+using System.Numerics;
+
+namespace Ryujinx.Graphics.OpenGL.Image
+{
+    class TextureCopyIncompatible
+    {
+        private const string ComputeShaderShortening = @"#version 450 core
+
+layout (binding = 0, $SRC_FORMAT$) uniform uimage2D src;
+layout (binding = 1, $DST_FORMAT$) uniform uimage2D dst;
+
+layout (local_size_x = 32, local_size_y = 32, local_size_z = 1) in;
+
+void main()
+{
+    uvec2 coords = gl_GlobalInvocationID.xy;
+    ivec2 imageSz = imageSize(src);
+
+    if (int(coords.x) >= imageSz.x || int(coords.y) >= imageSz.y)
+    {
+        return;
+    }
+
+    uint coordsShifted = coords.x << $RATIO_LOG2$;
+
+    uvec2 dstCoords0 = uvec2(coordsShifted, coords.y);
+    uvec2 dstCoords1 = uvec2(coordsShifted + 1, coords.y);
+    uvec2 dstCoords2 = uvec2(coordsShifted + 2, coords.y);
+    uvec2 dstCoords3 = uvec2(coordsShifted + 3, coords.y);
+
+    uvec4 rgba = imageLoad(src, ivec2(coords));
+
+    imageStore(dst, ivec2(dstCoords0), rgba.rrrr);
+    imageStore(dst, ivec2(dstCoords1), rgba.gggg);
+    imageStore(dst, ivec2(dstCoords2), rgba.bbbb);
+    imageStore(dst, ivec2(dstCoords3), rgba.aaaa);
+}";
+
+        private const string ComputeShaderWidening = @"#version 450 core
+
+layout (binding = 0, $SRC_FORMAT$) uniform uimage2D src;
+layout (binding = 1, $DST_FORMAT$) uniform uimage2D dst;
+
+layout (local_size_x = 32, local_size_y = 32, local_size_z = 1) in;
+
+void main()
+{
+    uvec2 coords = gl_GlobalInvocationID.xy;
+    ivec2 imageSz = imageSize(dst);
+
+    if (int(coords.x) >= imageSz.x || int(coords.y) >= imageSz.y)
+    {
+        return;
+    }
+
+    uvec2 srcCoords = uvec2(coords.x << $RATIO_LOG2$, coords.y);
+
+    uint r = imageLoad(src, ivec2(srcCoords) + ivec2(0, 0)).r;
+    uint g = imageLoad(src, ivec2(srcCoords) + ivec2(1, 0)).r;
+    uint b = imageLoad(src, ivec2(srcCoords) + ivec2(2, 0)).r;
+    uint a = imageLoad(src, ivec2(srcCoords) + ivec2(3, 0)).r;
+
+    imageStore(dst, ivec2(coords), uvec4(r, g, b, a));
+}";
+
+        private readonly OpenGLRenderer _renderer;
+        private readonly Dictionary<int, int> _shorteningProgramHandles;
+        private readonly Dictionary<int, int> _wideningProgramHandles;
+
+        public TextureCopyIncompatible(OpenGLRenderer renderer)
+        {
+            _renderer = renderer;
+            _shorteningProgramHandles = new Dictionary<int, int>();
+            _wideningProgramHandles = new Dictionary<int, int>();
+        }
+
+        public void CopyIncompatibleFormats(ITextureInfo src, ITextureInfo dst, int srcLayer, int dstLayer, int srcLevel, int dstLevel, int depth, int levels)
+        {
+            TextureCreateInfo srcInfo = src.Info;
+            TextureCreateInfo dstInfo = dst.Info;
+
+            int srcBpp = src.Info.BytesPerPixel;
+            int dstBpp = dst.Info.BytesPerPixel;
+
+            // Calculate ideal component size, given our constraints:
+            // - Component size must not exceed bytes per pixel of source and destination image formats.
+            // - Maximum component size is 4 (R32).
+            int componentSize = Math.Min(Math.Min(srcBpp, dstBpp), 4);
+
+            int srcComponentsCount = srcBpp / componentSize;
+            int dstComponentsCount = dstBpp / componentSize;
+
+            var srcFormat = GetFormat(componentSize, srcComponentsCount);
+            var dstFormat = GetFormat(componentSize, dstComponentsCount);
+
+            GL.UseProgram(srcBpp < dstBpp
+                ? GetWideningShader(componentSize, srcComponentsCount, dstComponentsCount)
+                : GetShorteningShader(componentSize, srcComponentsCount, dstComponentsCount));
+
+            for (int l = 0; l < levels; l++)
+            {
+                int srcWidth = Math.Max(1, src.Info.Width >> l);
+                int srcHeight = Math.Max(1, src.Info.Height >> l);
+
+                int dstWidth = Math.Max(1, dst.Info.Width >> l);
+                int dstHeight = Math.Max(1, dst.Info.Height >> l);
+
+                int width = Math.Min(srcWidth, dstWidth);
+                int height = Math.Min(srcHeight, dstHeight);
+
+                for (int z = 0; z < depth; z++)
+                {
+                    GL.BindImageTexture(0, src.Handle, srcLevel + l, false, srcLayer + z, TextureAccess.ReadOnly, srcFormat);
+                    GL.BindImageTexture(1, dst.Handle, dstLevel + l, false, dstLayer + z, TextureAccess.WriteOnly, dstFormat);
+
+                    GL.DispatchCompute((width + 31) / 32, (height + 31) / 32, 1);
+                }
+            }
+
+            Pipeline pipeline = (Pipeline)_renderer.Pipeline;
+
+            pipeline.RestoreProgram();
+            pipeline.RestoreImages1And2();
+        }
+
+        private static SizedInternalFormat GetFormat(int componentSize, int componentsCount)
+        {
+            if (componentSize == 1)
+            {
+                return componentsCount switch
+                {
+                    1 => SizedInternalFormat.R8ui,
+                    2 => SizedInternalFormat.Rg8ui,
+                    4 => SizedInternalFormat.Rgba8ui,
+                    _ => throw new ArgumentException($"Invalid components count {componentsCount}.")
+                };
+            }
+            else if (componentSize == 2)
+            {
+                return componentsCount switch
+                {
+                    1 => SizedInternalFormat.R16ui,
+                    2 => SizedInternalFormat.Rg16ui,
+                    4 => SizedInternalFormat.Rgba16ui,
+                    _ => throw new ArgumentException($"Invalid components count {componentsCount}.")
+                };
+            }
+            else if (componentSize == 4)
+            {
+                return componentsCount switch
+                {
+                    1 => SizedInternalFormat.R32ui,
+                    2 => SizedInternalFormat.Rg32ui,
+                    4 => SizedInternalFormat.Rgba32ui,
+                    _ => throw new ArgumentException($"Invalid components count {componentsCount}.")
+                };
+            }
+            else
+            {
+                throw new ArgumentException($"Invalid component size {componentSize}.");
+            }
+        }
+
+        private int GetShorteningShader(int componentSize, int srcComponentsCount, int dstComponentsCount)
+        {
+            return GetShader(ComputeShaderShortening, _shorteningProgramHandles, componentSize, srcComponentsCount, dstComponentsCount);
+        }
+
+        private int GetWideningShader(int componentSize, int srcComponentsCount, int dstComponentsCount)
+        {
+            return GetShader(ComputeShaderWidening, _wideningProgramHandles, componentSize, srcComponentsCount, dstComponentsCount);
+        }
+
+        private int GetShader(
+            string code,
+            Dictionary<int, int> programHandles,
+            int componentSize,
+            int srcComponentsCount,
+            int dstComponentsCount)
+        {
+            int componentSizeLog2 = BitOperations.Log2((uint)componentSize);
+
+            int srcIndex = componentSizeLog2 + BitOperations.Log2((uint)srcComponentsCount) * 3;
+            int dstIndex = componentSizeLog2 + BitOperations.Log2((uint)dstComponentsCount) * 3;
+
+            int key = srcIndex | (dstIndex << 8);
+
+            if (!programHandles.TryGetValue(key, out int programHandle))
+            {
+                int csHandle = GL.CreateShader(ShaderType.ComputeShader);
+
+                string[] formatTable = new[] { "r8ui", "r16ui", "r32ui", "rg8ui", "rg16ui", "rg32ui", "rgba8ui", "rgba16ui", "rgba32ui" };
+
+                string srcFormat = formatTable[srcIndex];
+                string dstFormat = formatTable[dstIndex];
+
+                int srcBpp = srcComponentsCount * componentSize;
+                int dstBpp = dstComponentsCount * componentSize;
+
+                int ratio = srcBpp < dstBpp ? dstBpp / srcBpp : srcBpp / dstBpp;
+                int ratioLog2 = BitOperations.Log2((uint)ratio);
+
+                GL.ShaderSource(csHandle, code
+                    .Replace("$SRC_FORMAT$", srcFormat)
+                    .Replace("$DST_FORMAT$", dstFormat)
+                    .Replace("$RATIO_LOG2$", ratioLog2.ToString(CultureInfo.InvariantCulture)));
+
+                GL.CompileShader(csHandle);
+
+                programHandle = GL.CreateProgram();
+
+                GL.AttachShader(programHandle, csHandle);
+                GL.LinkProgram(programHandle);
+                GL.DetachShader(programHandle, csHandle);
+                GL.DeleteShader(csHandle);
+
+                GL.GetProgram(programHandle, GetProgramParameterName.LinkStatus, out int status);
+
+                if (status == 0)
+                {
+                    throw new Exception(GL.GetProgramInfoLog(programHandle));
+                }
+
+                programHandles.Add(key, programHandle);
+            }
+
+            return programHandle;
+        }
+
+        public void Dispose()
+        {
+            foreach (int handle in _shorteningProgramHandles.Values)
+            {
+                GL.DeleteProgram(handle);
+            }
+
+            _shorteningProgramHandles.Clear();
+
+            foreach (int handle in _wideningProgramHandles.Values)
+            {
+                GL.DeleteProgram(handle);
+            }
+
+            _wideningProgramHandles.Clear();
+        }
+    }
+}
diff --git a/src/Ryujinx.Graphics.OpenGL/Image/TextureCopyMS.cs b/src/Ryujinx.Graphics.OpenGL/Image/TextureCopyMS.cs
new file mode 100644
index 00000000..9963dc66
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Image/TextureCopyMS.cs
@@ -0,0 +1,276 @@
+using OpenTK.Graphics.OpenGL;
+using Ryujinx.Graphics.GAL;
+using System;
+using System.Numerics;
+
+namespace Ryujinx.Graphics.OpenGL.Image
+{
+    class TextureCopyMS
+    {
+        private const string ComputeShaderMSToNonMS = @"#version 450 core
+
+layout (binding = 0, $FORMAT$) uniform uimage2DMS imgIn;
+layout (binding = 1, $FORMAT$) uniform uimage2D imgOut;
+
+layout (local_size_x = 32, local_size_y = 32, local_size_z = 1) in;
+
+void main()
+{
+    uvec2 coords = gl_GlobalInvocationID.xy;
+    ivec2 imageSz = imageSize(imgOut);
+    if (int(coords.x) >= imageSz.x || int(coords.y) >= imageSz.y)
+    {
+        return;
+    }
+    int inSamples = imageSamples(imgIn);
+    int samplesInXLog2 = 0;
+    int samplesInYLog2 = 0;
+    switch (inSamples)
+    {
+        case 2:
+            samplesInXLog2 = 1;
+            break;
+        case 4:
+            samplesInXLog2 = 1;
+            samplesInYLog2 = 1;
+            break;
+        case 8:
+            samplesInXLog2 = 2;
+            samplesInYLog2 = 1;
+            break;
+        case 16:
+            samplesInXLog2 = 2;
+            samplesInYLog2 = 2;
+            break;
+    }
+    int samplesInX = 1 << samplesInXLog2;
+    int samplesInY = 1 << samplesInYLog2;
+    int sampleIdx = (int(coords.x) & (samplesInX - 1)) | ((int(coords.y) & (samplesInY - 1)) << samplesInXLog2);
+    uvec4 value = imageLoad(imgIn, ivec2(int(coords.x) >> samplesInXLog2, int(coords.y) >> samplesInYLog2), sampleIdx);
+    imageStore(imgOut, ivec2(coords), value);
+}";
+
+        private const string ComputeShaderNonMSToMS = @"#version 450 core
+
+layout (binding = 0, $FORMAT$) uniform uimage2D imgIn;
+layout (binding = 1, $FORMAT$) uniform uimage2DMS imgOut;
+
+layout (local_size_x = 32, local_size_y = 32, local_size_z = 1) in;
+
+void main()
+{
+    uvec2 coords = gl_GlobalInvocationID.xy;
+    ivec2 imageSz = imageSize(imgIn);
+    if (int(coords.x) >= imageSz.x || int(coords.y) >= imageSz.y)
+    {
+        return;
+    }
+    int outSamples = imageSamples(imgOut);
+    int samplesInXLog2 = 0;
+    int samplesInYLog2 = 0;
+    switch (outSamples)
+    {
+        case 2:
+            samplesInXLog2 = 1;
+            break;
+        case 4:
+            samplesInXLog2 = 1;
+            samplesInYLog2 = 1;
+            break;
+        case 8:
+            samplesInXLog2 = 2;
+            samplesInYLog2 = 1;
+            break;
+        case 16:
+            samplesInXLog2 = 2;
+            samplesInYLog2 = 2;
+            break;
+    }
+    int samplesInX = 1 << samplesInXLog2;
+    int samplesInY = 1 << samplesInYLog2;
+    int sampleIdx = (int(coords.x) & (samplesInX - 1)) | ((int(coords.y) & (samplesInY - 1)) << samplesInXLog2);
+    uvec4 value = imageLoad(imgIn, ivec2(coords));
+    imageStore(imgOut, ivec2(int(coords.x) >> samplesInXLog2, int(coords.y) >> samplesInYLog2), sampleIdx, value);
+}";
+
+        private readonly OpenGLRenderer _renderer;
+        private int[] _msToNonMSProgramHandles;
+        private int[] _nonMSToMSProgramHandles;
+
+        public TextureCopyMS(OpenGLRenderer renderer)
+        {
+            _renderer = renderer;
+            _msToNonMSProgramHandles = new int[5];
+            _nonMSToMSProgramHandles = new int[5];
+        }
+
+        public void CopyMSToNonMS(ITextureInfo src, ITextureInfo dst, int srcLayer, int dstLayer, int depth)
+        {
+            TextureCreateInfo srcInfo = src.Info;
+            TextureCreateInfo dstInfo = dst.Info;
+
+            int srcHandle = CreateViewIfNeeded(src);
+            int dstHandle = CreateViewIfNeeded(dst);
+
+            int dstWidth = dstInfo.Width;
+            int dstHeight = dstInfo.Height;
+
+            GL.UseProgram(GetMSToNonMSShader(srcInfo.BytesPerPixel));
+
+            for (int z = 0; z < depth; z++)
+            {
+                GL.BindImageTexture(0, srcHandle, 0, false, srcLayer + z, TextureAccess.ReadOnly, GetFormat(srcInfo.BytesPerPixel));
+                GL.BindImageTexture(1, dstHandle, 0, false, dstLayer + z, TextureAccess.WriteOnly, GetFormat(dstInfo.BytesPerPixel));
+
+                GL.DispatchCompute((dstWidth + 31) / 32, (dstHeight + 31) / 32, 1);
+            }
+
+            Pipeline pipeline = (Pipeline)_renderer.Pipeline;
+
+            pipeline.RestoreProgram();
+            pipeline.RestoreImages1And2();
+
+            DestroyViewIfNeeded(src, srcHandle);
+            DestroyViewIfNeeded(dst, dstHandle);
+        }
+
+        public void CopyNonMSToMS(ITextureInfo src, ITextureInfo dst, int srcLayer, int dstLayer, int depth)
+        {
+            TextureCreateInfo srcInfo = src.Info;
+            TextureCreateInfo dstInfo = dst.Info;
+
+            int srcHandle = CreateViewIfNeeded(src);
+            int dstHandle = CreateViewIfNeeded(dst);
+
+            int srcWidth = srcInfo.Width;
+            int srcHeight = srcInfo.Height;
+
+            GL.UseProgram(GetNonMSToMSShader(srcInfo.BytesPerPixel));
+
+            for (int z = 0; z < depth; z++)
+            {
+                GL.BindImageTexture(0, srcHandle, 0, false, srcLayer + z, TextureAccess.ReadOnly, GetFormat(srcInfo.BytesPerPixel));
+                GL.BindImageTexture(1, dstHandle, 0, false, dstLayer + z, TextureAccess.WriteOnly, GetFormat(dstInfo.BytesPerPixel));
+
+                GL.DispatchCompute((srcWidth + 31) / 32, (srcHeight + 31) / 32, 1);
+            }
+
+            Pipeline pipeline = (Pipeline)_renderer.Pipeline;
+
+            pipeline.RestoreProgram();
+            pipeline.RestoreImages1And2();
+
+            DestroyViewIfNeeded(src, srcHandle);
+            DestroyViewIfNeeded(dst, dstHandle);
+        }
+
+        private static SizedInternalFormat GetFormat(int bytesPerPixel)
+        {
+            return bytesPerPixel switch
+            {
+                1 => SizedInternalFormat.R8ui,
+                2 => SizedInternalFormat.R16ui,
+                4 => SizedInternalFormat.R32ui,
+                8 => SizedInternalFormat.Rg32ui,
+                16 => SizedInternalFormat.Rgba32ui,
+                _ => throw new ArgumentException($"Invalid bytes per pixel {bytesPerPixel}.")
+            };
+        }
+
+        private static int CreateViewIfNeeded(ITextureInfo texture)
+        {
+            // Binding sRGB textures as images doesn't work on NVIDIA,
+            // we need to create and bind a RGBA view for it to work.
+            if (texture.Info.Format == Format.R8G8B8A8Srgb)
+            {
+                int handle = GL.GenTexture();
+
+                GL.TextureView(
+                    handle,
+                    texture.Info.Target.Convert(),
+                    texture.Storage.Handle,
+                    PixelInternalFormat.Rgba8,
+                    texture.FirstLevel,
+                    1,
+                    texture.FirstLayer,
+                    texture.Info.GetLayers());
+
+                return handle;
+            }
+
+            return texture.Handle;
+        }
+
+        private static void DestroyViewIfNeeded(ITextureInfo info, int handle)
+        {
+            if (info.Handle != handle)
+            {
+                GL.DeleteTexture(handle);
+            }
+        }
+
+        private int GetMSToNonMSShader(int bytesPerPixel)
+        {
+            return GetShader(ComputeShaderMSToNonMS, _msToNonMSProgramHandles, bytesPerPixel);
+        }
+
+        private int GetNonMSToMSShader(int bytesPerPixel)
+        {
+            return GetShader(ComputeShaderNonMSToMS, _nonMSToMSProgramHandles, bytesPerPixel);
+        }
+
+        private int GetShader(string code, int[] programHandles, int bytesPerPixel)
+        {
+            int index = BitOperations.Log2((uint)bytesPerPixel);
+
+            if (programHandles[index] == 0)
+            {
+                int csHandle = GL.CreateShader(ShaderType.ComputeShader);
+
+                string format = new[] { "r8ui", "r16ui", "r32ui", "rg32ui", "rgba32ui" }[index];
+
+                GL.ShaderSource(csHandle, code.Replace("$FORMAT$", format));
+                GL.CompileShader(csHandle);
+
+                int programHandle = GL.CreateProgram();
+
+                GL.AttachShader(programHandle, csHandle);
+                GL.LinkProgram(programHandle);
+                GL.DetachShader(programHandle, csHandle);
+                GL.DeleteShader(csHandle);
+
+                GL.GetProgram(programHandle, GetProgramParameterName.LinkStatus, out int status);
+
+                if (status == 0)
+                {
+                    throw new Exception(GL.GetProgramInfoLog(programHandle));
+                }
+
+                programHandles[index] = programHandle;
+            }
+
+            return programHandles[index];
+        }
+
+        public void Dispose()
+        {
+            for (int i = 0; i < _msToNonMSProgramHandles.Length; i++)
+            {
+                if (_msToNonMSProgramHandles[i] != 0)
+                {
+                    GL.DeleteProgram(_msToNonMSProgramHandles[i]);
+                    _msToNonMSProgramHandles[i] = 0;
+                }
+            }
+
+            for (int i = 0; i < _nonMSToMSProgramHandles.Length; i++)
+            {
+                if (_nonMSToMSProgramHandles[i] != 0)
+                {
+                    GL.DeleteProgram(_nonMSToMSProgramHandles[i]);
+                    _nonMSToMSProgramHandles[i] = 0;
+                }
+            }
+        }
+    }
+}
diff --git a/src/Ryujinx.Graphics.OpenGL/Image/TextureStorage.cs b/src/Ryujinx.Graphics.OpenGL/Image/TextureStorage.cs
new file mode 100644
index 00000000..c058ac88
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Image/TextureStorage.cs
@@ -0,0 +1,212 @@
+using OpenTK.Graphics.OpenGL;
+using Ryujinx.Common.Logging;
+using Ryujinx.Graphics.GAL;
+
+namespace Ryujinx.Graphics.OpenGL.Image
+{
+    class TextureStorage : ITextureInfo
+    {
+        public ITextureInfo Storage => this;
+        public int Handle { get; private set; }
+        public float ScaleFactor { get; private set; }
+
+        public TextureCreateInfo Info { get; }
+
+        private readonly OpenGLRenderer _renderer;
+
+        private int _viewsCount;
+
+        internal ITexture DefaultView { get; private set; }
+
+        public TextureStorage(OpenGLRenderer renderer, TextureCreateInfo info, float scaleFactor)
+        {
+            _renderer = renderer;
+            Info      = info;
+
+            Handle = GL.GenTexture();
+            ScaleFactor = scaleFactor;
+
+            CreateImmutableStorage();
+        }
+
+        private void CreateImmutableStorage()
+        {
+            TextureTarget target = Info.Target.Convert();
+
+            GL.ActiveTexture(TextureUnit.Texture0);
+
+            GL.BindTexture(target, Handle);
+
+            FormatInfo format = FormatTable.GetFormatInfo(Info.Format);
+
+            SizedInternalFormat internalFormat;
+
+            if (format.IsCompressed)
+            {
+                internalFormat = (SizedInternalFormat)format.PixelFormat;
+            }
+            else
+            {
+                internalFormat = (SizedInternalFormat)format.PixelInternalFormat;
+            }
+
+            int levels = Info.GetLevelsClamped();
+
+            switch (Info.Target)
+            {
+                case Target.Texture1D:
+                    GL.TexStorage1D(
+                        TextureTarget1d.Texture1D,
+                        levels,
+                        internalFormat,
+                        Info.Width);
+                    break;
+
+                case Target.Texture1DArray:
+                    GL.TexStorage2D(
+                        TextureTarget2d.Texture1DArray,
+                        levels,
+                        internalFormat,
+                        Info.Width,
+                        Info.Height);
+                    break;
+
+                case Target.Texture2D:
+                    GL.TexStorage2D(
+                        TextureTarget2d.Texture2D,
+                        levels,
+                        internalFormat,
+                        Info.Width,
+                        Info.Height);
+                    break;
+
+                case Target.Texture2DArray:
+                    GL.TexStorage3D(
+                        TextureTarget3d.Texture2DArray,
+                        levels,
+                        internalFormat,
+                        Info.Width,
+                        Info.Height,
+                        Info.Depth);
+                    break;
+
+                case Target.Texture2DMultisample:
+                    GL.TexStorage2DMultisample(
+                        TextureTargetMultisample2d.Texture2DMultisample,
+                        Info.Samples,
+                        internalFormat,
+                        Info.Width,
+                        Info.Height,
+                        true);
+                    break;
+
+                case Target.Texture2DMultisampleArray:
+                    GL.TexStorage3DMultisample(
+                        TextureTargetMultisample3d.Texture2DMultisampleArray,
+                        Info.Samples,
+                        internalFormat,
+                        Info.Width,
+                        Info.Height,
+                        Info.Depth,
+                        true);
+                    break;
+
+                case Target.Texture3D:
+                    GL.TexStorage3D(
+                        TextureTarget3d.Texture3D,
+                        levels,
+                        internalFormat,
+                        Info.Width,
+                        Info.Height,
+                        Info.Depth);
+                    break;
+
+                case Target.Cubemap:
+                    GL.TexStorage2D(
+                        TextureTarget2d.TextureCubeMap,
+                        levels,
+                        internalFormat,
+                        Info.Width,
+                        Info.Height);
+                    break;
+
+                case Target.CubemapArray:
+                    GL.TexStorage3D(
+                        (TextureTarget3d)All.TextureCubeMapArray,
+                        levels,
+                        internalFormat,
+                        Info.Width,
+                        Info.Height,
+                        Info.Depth);
+                    break;
+
+                default:
+                    Logger.Debug?.Print(LogClass.Gpu, $"Invalid or unsupported texture target: {target}.");
+                    break;
+            }
+        }
+
+        public ITexture CreateDefaultView()
+        {
+            DefaultView = CreateView(Info, 0, 0);
+
+            return DefaultView;
+        }
+
+        public ITexture CreateView(TextureCreateInfo info, int firstLayer, int firstLevel)
+        {
+            IncrementViewsCount();
+
+            return new TextureView(_renderer, this, info, firstLayer, firstLevel);
+        }
+
+        private void IncrementViewsCount()
+        {
+            _viewsCount++;
+        }
+
+        public void DecrementViewsCount()
+        {
+            // If we don't have any views, then the storage is now useless, delete it.
+            if (--_viewsCount == 0)
+            {
+                if (DefaultView == null)
+                {
+                    Dispose();
+                }
+                else
+                {
+                    // If the default view still exists, we can put it into the resource pool.
+                    Release();
+                }
+            }
+        }
+
+        /// <summary>
+        /// Release the TextureStorage to the resource pool without disposing its handle.
+        /// </summary>
+        public void Release()
+        {
+            _viewsCount = 1; // When we are used again, we will have the default view.
+
+            _renderer.ResourcePool.AddTexture((TextureView)DefaultView);
+        }
+
+        public void DeleteDefault()
+        {
+            DefaultView = null;
+        }
+
+        public void Dispose()
+        {
+            DefaultView = null;
+
+            if (Handle != 0)
+            {
+                GL.DeleteTexture(Handle);
+
+                Handle = 0;
+            }
+        }
+    }
+}
diff --git a/src/Ryujinx.Graphics.OpenGL/Image/TextureView.cs b/src/Ryujinx.Graphics.OpenGL/Image/TextureView.cs
new file mode 100644
index 00000000..804b3b03
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Image/TextureView.cs
@@ -0,0 +1,867 @@
+using OpenTK.Graphics.OpenGL;
+using Ryujinx.Common;
+using Ryujinx.Common.Memory;
+using Ryujinx.Graphics.GAL;
+using System;
+
+namespace Ryujinx.Graphics.OpenGL.Image
+{
+    class TextureView : TextureBase, ITexture, ITextureInfo
+    {
+        private readonly OpenGLRenderer _renderer;
+
+        private readonly TextureStorage _parent;
+
+        public ITextureInfo Storage => _parent;
+
+        public int FirstLayer { get; private set; }
+        public int FirstLevel { get; private set; }
+
+        public TextureView(
+            OpenGLRenderer renderer,
+            TextureStorage parent,
+            TextureCreateInfo info,
+            int firstLayer,
+            int firstLevel) : base(info, parent.ScaleFactor)
+        {
+            _renderer = renderer;
+            _parent = parent;
+
+            FirstLayer = firstLayer;
+            FirstLevel = firstLevel;
+
+            CreateView();
+        }
+
+        private void CreateView()
+        {
+            TextureTarget target = Target.Convert();
+
+            FormatInfo format = FormatTable.GetFormatInfo(Info.Format);
+
+            PixelInternalFormat pixelInternalFormat;
+
+            if (format.IsCompressed)
+            {
+                pixelInternalFormat = (PixelInternalFormat)format.PixelFormat;
+            }
+            else
+            {
+                pixelInternalFormat = format.PixelInternalFormat;
+            }
+
+            int levels = Info.GetLevelsClamped();
+
+            GL.TextureView(
+                Handle,
+                target,
+                _parent.Handle,
+                pixelInternalFormat,
+                FirstLevel,
+                levels,
+                FirstLayer,
+                Info.GetLayers());
+
+            GL.ActiveTexture(TextureUnit.Texture0);
+
+            GL.BindTexture(target, Handle);
+
+            int[] swizzleRgba = new int[]
+            {
+                (int)Info.SwizzleR.Convert(),
+                (int)Info.SwizzleG.Convert(),
+                (int)Info.SwizzleB.Convert(),
+                (int)Info.SwizzleA.Convert()
+            };
+
+            if (Info.Format == Format.A1B5G5R5Unorm)
+            {
+                int temp = swizzleRgba[0];
+                int temp2 = swizzleRgba[1];
+                swizzleRgba[0] = swizzleRgba[3];
+                swizzleRgba[1] = swizzleRgba[2];
+                swizzleRgba[2] = temp2;
+                swizzleRgba[3] = temp;
+            }
+            else if (Info.Format.IsBgr())
+            {
+                // Swap B <-> R for BGRA formats, as OpenGL has no support for them
+                // and we need to manually swap the components on read/write on the GPU.
+                int temp = swizzleRgba[0];
+                swizzleRgba[0] = swizzleRgba[2];
+                swizzleRgba[2] = temp;
+            }
+
+            GL.TexParameter(target, TextureParameterName.TextureSwizzleRgba, swizzleRgba);
+
+            int maxLevel = levels - 1;
+
+            if (maxLevel < 0)
+            {
+                maxLevel = 0;
+            }
+
+            GL.TexParameter(target, TextureParameterName.TextureMaxLevel, maxLevel);
+            GL.TexParameter(target, TextureParameterName.DepthStencilTextureMode, (int)Info.DepthStencilMode.Convert());
+        }
+
+        public ITexture CreateView(TextureCreateInfo info, int firstLayer, int firstLevel)
+        {
+            firstLayer += FirstLayer;
+            firstLevel += FirstLevel;
+
+            return _parent.CreateView(info, firstLayer, firstLevel);
+        }
+
+        public void CopyTo(ITexture destination, int firstLayer, int firstLevel)
+        {
+            TextureView destinationView = (TextureView)destination;
+
+            bool srcIsMultisample = Target.IsMultisample();
+            bool dstIsMultisample = destinationView.Target.IsMultisample();
+
+            if (dstIsMultisample != srcIsMultisample && Info.Format.IsDepthOrStencil())
+            {
+                int layers = Math.Min(Info.GetLayers(), destinationView.Info.GetLayers() - firstLayer);
+                CopyWithBlitForDepthMS(destinationView, 0, firstLayer, layers);
+            }
+            else if (!dstIsMultisample && srcIsMultisample)
+            {
+                int layers = Math.Min(Info.GetLayers(), destinationView.Info.GetLayers() - firstLayer);
+                _renderer.TextureCopyMS.CopyMSToNonMS(this, destinationView, 0, firstLayer, layers);
+            }
+            else if (dstIsMultisample && !srcIsMultisample)
+            {
+                int layers = Math.Min(Info.GetLayers(), destinationView.Info.GetLayers() - firstLayer);
+                _renderer.TextureCopyMS.CopyNonMSToMS(this, destinationView, 0, firstLayer, layers);
+            }
+            else if (destinationView.Info.BytesPerPixel != Info.BytesPerPixel)
+            {
+                int layers = Math.Min(Info.GetLayers(), destinationView.Info.GetLayers() - firstLayer);
+                int levels = Math.Min(Info.Levels, destinationView.Info.Levels - firstLevel);
+                _renderer.TextureCopyIncompatible.CopyIncompatibleFormats(this, destinationView, 0, firstLayer, 0, firstLevel, layers, levels);
+            }
+            else
+            {
+                _renderer.TextureCopy.CopyUnscaled(this, destinationView, 0, firstLayer, 0, firstLevel);
+            }
+        }
+
+        public void CopyTo(ITexture destination, int srcLayer, int dstLayer, int srcLevel, int dstLevel)
+        {
+            TextureView destinationView = (TextureView)destination;
+
+            bool srcIsMultisample = Target.IsMultisample();
+            bool dstIsMultisample = destinationView.Target.IsMultisample();
+
+            if (dstIsMultisample != srcIsMultisample && Info.Format.IsDepthOrStencil())
+            {
+                CopyWithBlitForDepthMS(destinationView, srcLayer, dstLayer, 1);
+            }
+            else if (!dstIsMultisample && srcIsMultisample)
+            {
+                _renderer.TextureCopyMS.CopyMSToNonMS(this, destinationView, srcLayer, dstLayer, 1);
+            }
+            else if (dstIsMultisample && !srcIsMultisample)
+            {
+                _renderer.TextureCopyMS.CopyNonMSToMS(this, destinationView, srcLayer, dstLayer, 1);
+            }
+            else if (destinationView.Info.BytesPerPixel != Info.BytesPerPixel)
+            {
+                _renderer.TextureCopyIncompatible.CopyIncompatibleFormats(this, destinationView, srcLayer, dstLayer, srcLevel, dstLevel, 1, 1);
+            }
+            else
+            {
+                _renderer.TextureCopy.CopyUnscaled(this, destinationView, srcLayer, dstLayer, srcLevel, dstLevel, 1, 1);
+            }
+        }
+
+        private void CopyWithBlitForDepthMS(TextureView destinationView, int srcLayer, int dstLayer, int layers)
+        {
+            // This is currently used for multisample <-> non-multisample copies.
+            // We can't do that with compute because it's not possible to write depth textures on compute.
+            // It can be done with draws, but we don't have support for saving and restoring the OpenGL state
+            // for a draw with different state right now.
+            // This approach uses blit, which causes a resolution loss since some samples will be lost
+            // in the process.
+
+            Extents2D srcRegion = new Extents2D(0, 0, Width, Height);
+            Extents2D dstRegion = new Extents2D(0, 0, destinationView.Width, destinationView.Height);
+
+            if (destinationView.Target.IsMultisample())
+            {
+                TextureView intermmediate = _renderer.TextureCopy.IntermediatePool.GetOrCreateWithAtLeast(
+                    Info.Target,
+                    Info.BlockWidth,
+                    Info.BlockHeight,
+                    Info.BytesPerPixel,
+                    Format,
+                    destinationView.Width,
+                    destinationView.Height,
+                    Info.Depth,
+                    1,
+                    1);
+
+                _renderer.TextureCopy.Copy(this, intermmediate, srcRegion, dstRegion, false);
+                _renderer.TextureCopy.Copy(intermmediate, destinationView, dstRegion, dstRegion, false, srcLayer, dstLayer, 0, 0, layers, 1);
+            }
+            else
+            {
+                Target target = Target switch
+                {
+                    Target.Texture2DMultisample => Target.Texture2D,
+                    Target.Texture2DMultisampleArray => Target.Texture2DArray,
+                    _ => Target
+                };
+
+                TextureView intermmediate = _renderer.TextureCopy.IntermediatePool.GetOrCreateWithAtLeast(
+                    target,
+                    Info.BlockWidth,
+                    Info.BlockHeight,
+                    Info.BytesPerPixel,
+                    Format,
+                    Width,
+                    Height,
+                    Info.Depth,
+                    1,
+                    1);
+
+                _renderer.TextureCopy.Copy(this, intermmediate, srcRegion, srcRegion, false);
+                _renderer.TextureCopy.Copy(intermmediate, destinationView, srcRegion, dstRegion, false, srcLayer, dstLayer, 0, 0, layers, 1);
+            }
+        }
+
+        public void CopyTo(ITexture destination, Extents2D srcRegion, Extents2D dstRegion, bool linearFilter)
+        {
+            _renderer.TextureCopy.Copy(this, (TextureView)destination, srcRegion, dstRegion, linearFilter);
+        }
+
+        public unsafe PinnedSpan<byte> GetData()
+        {
+            int size = 0;
+            int levels = Info.GetLevelsClamped();
+
+            for (int level = 0; level < levels; level++)
+            {
+                size += Info.GetMipSize(level);
+            }
+
+            ReadOnlySpan<byte> data;
+
+            if (HwCapabilities.UsePersistentBufferForFlush)
+            {
+                data = _renderer.PersistentBuffers.Default.GetTextureData(this, size);
+            }
+            else
+            {
+                IntPtr target = _renderer.PersistentBuffers.Default.GetHostArray(size);
+
+                WriteTo(target);
+
+                data = new ReadOnlySpan<byte>(target.ToPointer(), size);
+            }
+
+            if (Format == Format.S8UintD24Unorm)
+            {
+                data = FormatConverter.ConvertD24S8ToS8D24(data);
+            }
+
+            return PinnedSpan<byte>.UnsafeFromSpan(data);
+        }
+
+        public unsafe PinnedSpan<byte> GetData(int layer, int level)
+        {
+            int size = Info.GetMipSize(level);
+
+            if (HwCapabilities.UsePersistentBufferForFlush)
+            {
+                return PinnedSpan<byte>.UnsafeFromSpan(_renderer.PersistentBuffers.Default.GetTextureData(this, size, layer, level));
+            }
+            else
+            {
+                IntPtr target = _renderer.PersistentBuffers.Default.GetHostArray(size);
+
+                int offset = WriteTo2D(target, layer, level);
+
+                return new PinnedSpan<byte>((byte*)target.ToPointer() + offset, size);
+            }
+        }
+
+        public void WriteToPbo(int offset, bool forceBgra)
+        {
+            WriteTo(IntPtr.Zero + offset, forceBgra);
+        }
+
+        public int WriteToPbo2D(int offset, int layer, int level)
+        {
+            return WriteTo2D(IntPtr.Zero + offset, layer, level);
+        }
+
+        private int WriteTo2D(IntPtr data, int layer, int level)
+        {
+            TextureTarget target = Target.Convert();
+
+            Bind(target, 0);
+
+            FormatInfo format = FormatTable.GetFormatInfo(Info.Format);
+
+            PixelFormat pixelFormat = format.PixelFormat;
+            PixelType pixelType = format.PixelType;
+
+            if (target == TextureTarget.TextureCubeMap || target == TextureTarget.TextureCubeMapArray)
+            {
+                target = TextureTarget.TextureCubeMapPositiveX + (layer % 6);
+            }
+
+            int mipSize = Info.GetMipSize2D(level);
+
+            if (format.IsCompressed)
+            {
+                GL.GetCompressedTextureSubImage(Handle, level, 0, 0, layer, Math.Max(1, Info.Width >> level), Math.Max(1, Info.Height >> level), 1, mipSize, data);
+            }
+            else if (format.PixelFormat != PixelFormat.DepthStencil)
+            {
+                GL.GetTextureSubImage(Handle, level, 0, 0, layer, Math.Max(1, Info.Width >> level), Math.Max(1, Info.Height >> level), 1, pixelFormat, pixelType, mipSize, data);
+            }
+            else
+            {
+                GL.GetTexImage(target, level, pixelFormat, pixelType, data);
+
+                // The GL function returns all layers. Must return the offset of the layer we're interested in.
+                return target switch
+                {
+                    TextureTarget.TextureCubeMapArray => (layer / 6) * mipSize,
+                    TextureTarget.Texture1DArray => layer * mipSize,
+                    TextureTarget.Texture2DArray => layer * mipSize,
+                    _ => 0
+                };
+            }
+
+            return 0;
+        }
+
+        private void WriteTo(IntPtr data, bool forceBgra = false)
+        {
+            TextureTarget target = Target.Convert();
+
+            Bind(target, 0);
+
+            FormatInfo format = FormatTable.GetFormatInfo(Info.Format);
+
+            PixelFormat pixelFormat = format.PixelFormat;
+            PixelType pixelType = format.PixelType;
+
+            if (forceBgra)
+            {
+                if (pixelType == PixelType.UnsignedShort565)
+                {
+                    pixelType = PixelType.UnsignedShort565Reversed;
+                }
+                else if (pixelType == PixelType.UnsignedShort565Reversed)
+                {
+                    pixelType = PixelType.UnsignedShort565;
+                }
+                else
+                {
+                    pixelFormat = PixelFormat.Bgra;
+                }
+            }
+
+            int faces = 1;
+
+            if (target == TextureTarget.TextureCubeMap)
+            {
+                target = TextureTarget.TextureCubeMapPositiveX;
+
+                faces = 6;
+            }
+
+            int levels = Info.GetLevelsClamped();
+
+            for (int level = 0; level < levels; level++)
+            {
+                for (int face = 0; face < faces; face++)
+                {
+                    int faceOffset = face * Info.GetMipSize2D(level);
+
+                    if (format.IsCompressed)
+                    {
+                        GL.GetCompressedTexImage(target + face, level, data + faceOffset);
+                    }
+                    else
+                    {
+                        GL.GetTexImage(target + face, level, pixelFormat, pixelType, data + faceOffset);
+                    }
+                }
+
+                data += Info.GetMipSize(level);
+            }
+        }
+
+        public void SetData(SpanOrArray<byte> data)
+        {
+            var dataSpan = data.AsSpan();
+
+            if (Format == Format.S8UintD24Unorm)
+            {
+                dataSpan = FormatConverter.ConvertS8D24ToD24S8(dataSpan);
+            }
+
+            unsafe
+            {
+                fixed (byte* ptr = dataSpan)
+                {
+                    ReadFrom((IntPtr)ptr, dataSpan.Length);
+                }
+            }
+        }
+
+        public void SetData(SpanOrArray<byte> data, int layer, int level)
+        {
+            var dataSpan = data.AsSpan();
+
+            if (Format == Format.S8UintD24Unorm)
+            {
+                dataSpan = FormatConverter.ConvertS8D24ToD24S8(dataSpan);
+            }
+
+            unsafe
+            {
+                fixed (byte* ptr = dataSpan)
+                {
+                    int width = Math.Max(Info.Width >> level, 1);
+                    int height = Math.Max(Info.Height >> level, 1);
+
+                    ReadFrom2D((IntPtr)ptr, layer, level, 0, 0, width, height);
+                }
+            }
+        }
+
+        public void SetData(SpanOrArray<byte> data, int layer, int level, Rectangle<int> region)
+        {
+            var dataSpan = data.AsSpan();
+
+            if (Format == Format.S8UintD24Unorm)
+            {
+                dataSpan = FormatConverter.ConvertS8D24ToD24S8(dataSpan);
+            }
+
+            int wInBlocks = BitUtils.DivRoundUp(region.Width, Info.BlockWidth);
+            int hInBlocks = BitUtils.DivRoundUp(region.Height, Info.BlockHeight);
+
+            unsafe
+            {
+                fixed (byte* ptr = dataSpan)
+                {
+                    ReadFrom2D(
+                        (IntPtr)ptr,
+                        layer,
+                        level,
+                        region.X,
+                        region.Y,
+                        region.Width,
+                        region.Height,
+                        BitUtils.AlignUp(wInBlocks * Info.BytesPerPixel, 4) * hInBlocks);
+                }
+            }
+        }
+
+        public void ReadFromPbo(int offset, int size)
+        {
+            ReadFrom(IntPtr.Zero + offset, size);
+        }
+
+        public void ReadFromPbo2D(int offset, int layer, int level, int width, int height)
+        {
+            ReadFrom2D(IntPtr.Zero + offset, layer, level, 0, 0, width, height);
+        }
+
+        private void ReadFrom2D(IntPtr data, int layer, int level, int x, int y, int width, int height)
+        {
+            int mipSize = Info.GetMipSize2D(level);
+
+            ReadFrom2D(data, layer, level, x, y, width, height, mipSize);
+        }
+
+        private void ReadFrom2D(IntPtr data, int layer, int level, int x, int y, int width, int height, int mipSize)
+        {
+            TextureTarget target = Target.Convert();
+
+            Bind(target, 0);
+
+            FormatInfo format = FormatTable.GetFormatInfo(Info.Format);
+
+            switch (Target)
+            {
+                case Target.Texture1D:
+                    if (format.IsCompressed)
+                    {
+                        GL.CompressedTexSubImage1D(
+                            target,
+                            level,
+                            x,
+                            width,
+                            format.PixelFormat,
+                            mipSize,
+                            data);
+                    }
+                    else
+                    {
+                        GL.TexSubImage1D(
+                            target,
+                            level,
+                            x,
+                            width,
+                            format.PixelFormat,
+                            format.PixelType,
+                            data);
+                    }
+                    break;
+
+                case Target.Texture1DArray:
+                    if (format.IsCompressed)
+                    {
+                        GL.CompressedTexSubImage2D(
+                            target,
+                            level,
+                            x,
+                            layer,
+                            width,
+                            1,
+                            format.PixelFormat,
+                            mipSize,
+                            data);
+                    }
+                    else
+                    {
+                        GL.TexSubImage2D(
+                            target,
+                            level,
+                            x,
+                            layer,
+                            width,
+                            1,
+                            format.PixelFormat,
+                            format.PixelType,
+                            data);
+                    }
+                    break;
+
+                case Target.Texture2D:
+                    if (format.IsCompressed)
+                    {
+                        GL.CompressedTexSubImage2D(
+                            target,
+                            level,
+                            x,
+                            y,
+                            width,
+                            height,
+                            format.PixelFormat,
+                            mipSize,
+                            data);
+                    }
+                    else
+                    {
+                        GL.TexSubImage2D(
+                            target,
+                            level,
+                            x,
+                            y,
+                            width,
+                            height,
+                            format.PixelFormat,
+                            format.PixelType,
+                            data);
+                    }
+                    break;
+
+                case Target.Texture2DArray:
+                case Target.Texture3D:
+                case Target.CubemapArray:
+                    if (format.IsCompressed)
+                    {
+                        GL.CompressedTexSubImage3D(
+                            target,
+                            level,
+                            x,
+                            y,
+                            layer,
+                            width,
+                            height,
+                            1,
+                            format.PixelFormat,
+                            mipSize,
+                            data);
+                    }
+                    else
+                    {
+                        GL.TexSubImage3D(
+                            target,
+                            level,
+                            x,
+                            y,
+                            layer,
+                            width,
+                            height,
+                            1,
+                            format.PixelFormat,
+                            format.PixelType,
+                            data);
+                    }
+                    break;
+
+                case Target.Cubemap:
+                    if (format.IsCompressed)
+                    {
+                        GL.CompressedTexSubImage2D(
+                            TextureTarget.TextureCubeMapPositiveX + layer,
+                            level,
+                            x,
+                            y,
+                            width,
+                            height,
+                            format.PixelFormat,
+                            mipSize,
+                            data);
+                    }
+                    else
+                    {
+                        GL.TexSubImage2D(
+                            TextureTarget.TextureCubeMapPositiveX + layer,
+                            level,
+                            x,
+                            y,
+                            width,
+                            height,
+                            format.PixelFormat,
+                            format.PixelType,
+                            data);
+                    }
+                    break;
+            }
+        }
+
+        private void ReadFrom(IntPtr data, int size)
+        {
+            TextureTarget target = Target.Convert();
+            int baseLevel = 0;
+
+            // glTexSubImage on cubemap views is broken on Intel, we have to use the storage instead.
+            if (Target == Target.Cubemap && HwCapabilities.Vendor == HwCapabilities.GpuVendor.IntelWindows)
+            {
+                GL.ActiveTexture(TextureUnit.Texture0);
+                GL.BindTexture(target, Storage.Handle);
+                baseLevel = FirstLevel;
+            }
+            else
+            {
+                Bind(target, 0);
+            }
+
+            FormatInfo format = FormatTable.GetFormatInfo(Info.Format);
+
+            int width = Info.Width;
+            int height = Info.Height;
+            int depth = Info.Depth;
+            int levels = Info.GetLevelsClamped();
+
+            int offset = 0;
+
+            for (int level = 0; level < levels; level++)
+            {
+                int mipSize = Info.GetMipSize(level);
+
+                int endOffset = offset + mipSize;
+
+                if ((uint)endOffset > (uint)size)
+                {
+                    return;
+                }
+
+                switch (Target)
+                {
+                    case Target.Texture1D:
+                        if (format.IsCompressed)
+                        {
+                            GL.CompressedTexSubImage1D(
+                                target,
+                                level,
+                                0,
+                                width,
+                                format.PixelFormat,
+                                mipSize,
+                                data);
+                        }
+                        else
+                        {
+                            GL.TexSubImage1D(
+                                target,
+                                level,
+                                0,
+                                width,
+                                format.PixelFormat,
+                                format.PixelType,
+                                data);
+                        }
+                        break;
+
+                    case Target.Texture1DArray:
+                    case Target.Texture2D:
+                        if (format.IsCompressed)
+                        {
+                            GL.CompressedTexSubImage2D(
+                                target,
+                                level,
+                                0,
+                                0,
+                                width,
+                                height,
+                                format.PixelFormat,
+                                mipSize,
+                                data);
+                        }
+                        else
+                        {
+                            GL.TexSubImage2D(
+                                target,
+                                level,
+                                0,
+                                0,
+                                width,
+                                height,
+                                format.PixelFormat,
+                                format.PixelType,
+                                data);
+                        }
+                        break;
+
+                    case Target.Texture2DArray:
+                    case Target.Texture3D:
+                    case Target.CubemapArray:
+                        if (format.IsCompressed)
+                        {
+                            GL.CompressedTexSubImage3D(
+                                target,
+                                level,
+                                0,
+                                0,
+                                0,
+                                width,
+                                height,
+                                depth,
+                                format.PixelFormat,
+                                mipSize,
+                                data);
+                        }
+                        else
+                        {
+                            GL.TexSubImage3D(
+                                target,
+                                level,
+                                0,
+                                0,
+                                0,
+                                width,
+                                height,
+                                depth,
+                                format.PixelFormat,
+                                format.PixelType,
+                                data);
+                        }
+                        break;
+
+                    case Target.Cubemap:
+                        int faceOffset = 0;
+
+                        for (int face = 0; face < 6; face++, faceOffset += mipSize / 6)
+                        {
+                            if (format.IsCompressed)
+                            {
+                                GL.CompressedTexSubImage2D(
+                                    TextureTarget.TextureCubeMapPositiveX + face,
+                                    baseLevel + level,
+                                    0,
+                                    0,
+                                    width,
+                                    height,
+                                    format.PixelFormat,
+                                    mipSize / 6,
+                                    data + faceOffset);
+                            }
+                            else
+                            {
+                                GL.TexSubImage2D(
+                                    TextureTarget.TextureCubeMapPositiveX + face,
+                                    baseLevel + level,
+                                    0,
+                                    0,
+                                    width,
+                                    height,
+                                    format.PixelFormat,
+                                    format.PixelType,
+                                    data + faceOffset);
+                            }
+                        }
+                        break;
+                }
+
+                data += mipSize;
+                offset += mipSize;
+
+                width = Math.Max(1, width >> 1);
+                height = Math.Max(1, height >> 1);
+
+                if (Target == Target.Texture3D)
+                {
+                    depth = Math.Max(1, depth >> 1);
+                }
+            }
+        }
+
+        public void SetStorage(BufferRange buffer)
+        {
+            throw new NotSupportedException();
+        }
+
+        private void DisposeHandles()
+        {
+            if (Handle != 0)
+            {
+                GL.DeleteTexture(Handle);
+
+                Handle = 0;
+            }
+        }
+
+        /// <summary>
+        /// Release the view without necessarily disposing the parent if we are the default view.
+        /// This allows it to be added to the resource pool and reused later.
+        /// </summary>
+        public void Release()
+        {
+            bool hadHandle = Handle != 0;
+
+            if (_parent.DefaultView != this)
+            {
+                DisposeHandles();
+            }
+
+            if (hadHandle)
+            {
+                _parent.DecrementViewsCount();
+            }
+        }
+
+        public void Dispose()
+        {
+            if (_parent.DefaultView == this)
+            {
+                // Remove the default view (us), so that the texture cannot be released to the cache.
+                _parent.DeleteDefault();
+            }
+
+            Release();
+        }
+    }
+}
diff --git a/src/Ryujinx.Graphics.OpenGL/OpenGLRenderer.cs b/src/Ryujinx.Graphics.OpenGL/OpenGLRenderer.cs
new file mode 100644
index 00000000..3903b4d4
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/OpenGLRenderer.cs
@@ -0,0 +1,276 @@
+using OpenTK.Graphics.OpenGL;
+using Ryujinx.Common.Configuration;
+using Ryujinx.Common.Logging;
+using Ryujinx.Graphics.GAL;
+using Ryujinx.Graphics.OpenGL.Image;
+using Ryujinx.Graphics.OpenGL.Queries;
+using Ryujinx.Graphics.Shader.Translation;
+using System;
+
+namespace Ryujinx.Graphics.OpenGL
+{
+    public sealed class OpenGLRenderer : IRenderer
+    {
+        private readonly Pipeline _pipeline;
+
+        public IPipeline Pipeline => _pipeline;
+
+        private readonly Counters _counters;
+
+        private readonly Window _window;
+
+        public IWindow Window => _window;
+
+        private TextureCopy _textureCopy;
+        private TextureCopy _backgroundTextureCopy;
+        internal TextureCopy TextureCopy => BackgroundContextWorker.InBackground ? _backgroundTextureCopy : _textureCopy;
+        internal TextureCopyIncompatible TextureCopyIncompatible { get; }
+        internal TextureCopyMS TextureCopyMS { get; }
+
+        private Sync _sync;
+
+        public event EventHandler<ScreenCaptureImageInfo> ScreenCaptured;
+
+        internal PersistentBuffers PersistentBuffers { get; }
+
+        internal ResourcePool ResourcePool { get; }
+
+        internal int BufferCount { get; private set; }
+
+        public string GpuVendor { get; private set; }
+        public string GpuRenderer { get; private set; }
+        public string GpuVersion { get; private set; }
+
+        public bool PreferThreading => true;
+
+        public OpenGLRenderer()
+        {
+            _pipeline = new Pipeline();
+            _counters = new Counters();
+            _window = new Window(this);
+            _textureCopy = new TextureCopy(this);
+            _backgroundTextureCopy = new TextureCopy(this);
+            TextureCopyIncompatible = new TextureCopyIncompatible(this);
+            TextureCopyMS = new TextureCopyMS(this);
+            _sync = new Sync();
+            PersistentBuffers = new PersistentBuffers();
+            ResourcePool = new ResourcePool();
+        }
+
+        public BufferHandle CreateBuffer(int size, BufferHandle storageHint)
+        {
+            BufferCount++;
+
+            return Buffer.Create(size);
+        }
+
+        public IProgram CreateProgram(ShaderSource[] shaders, ShaderInfo info)
+        {
+            return new Program(shaders, info.FragmentOutputMap);
+        }
+
+        public ISampler CreateSampler(SamplerCreateInfo info)
+        {
+            return new Sampler(info);
+        }
+
+        public ITexture CreateTexture(TextureCreateInfo info, float scaleFactor)
+        {
+            if (info.Target == Target.TextureBuffer)
+            {
+                return new TextureBuffer(this, info);
+            }
+            else
+            {
+                return ResourcePool.GetTextureOrNull(info, scaleFactor) ?? new TextureStorage(this, info, scaleFactor).CreateDefaultView();
+            }
+        }
+
+        public void DeleteBuffer(BufferHandle buffer)
+        {
+            Buffer.Delete(buffer);
+        }
+
+        public HardwareInfo GetHardwareInfo()
+        {
+            return new HardwareInfo(GpuVendor, GpuRenderer);
+        }
+
+        public PinnedSpan<byte> GetBufferData(BufferHandle buffer, int offset, int size)
+        {
+            return Buffer.GetData(this, buffer, offset, size);
+        }
+
+        public Capabilities GetCapabilities()
+        {
+            bool intelWindows = HwCapabilities.Vendor == HwCapabilities.GpuVendor.IntelWindows;
+            bool amdWindows = HwCapabilities.Vendor == HwCapabilities.GpuVendor.AmdWindows;
+
+            return new Capabilities(
+                api: TargetApi.OpenGL,
+                vendorName: GpuVendor,
+                hasFrontFacingBug: intelWindows,
+                hasVectorIndexingBug: amdWindows,
+                needsFragmentOutputSpecialization: false,
+                reduceShaderPrecision: false,
+                supportsAstcCompression: HwCapabilities.SupportsAstcCompression,
+                supportsBc123Compression: HwCapabilities.SupportsTextureCompressionS3tc,
+                supportsBc45Compression: HwCapabilities.SupportsTextureCompressionRgtc,
+                supportsBc67Compression: true, // Should check BPTC extension, but for some reason NVIDIA is not exposing the extension.
+                supportsEtc2Compression: true,
+                supports3DTextureCompression: false,
+                supportsBgraFormat: false,
+                supportsR4G4Format: false,
+                supportsR4G4B4A4Format: true,
+                supportsSnormBufferTextureFormat: false,
+                supports5BitComponentFormat: true,
+                supportsBlendEquationAdvanced: HwCapabilities.SupportsBlendEquationAdvanced,
+                supportsFragmentShaderInterlock: HwCapabilities.SupportsFragmentShaderInterlock,
+                supportsFragmentShaderOrderingIntel: HwCapabilities.SupportsFragmentShaderOrdering,
+                supportsGeometryShader: true,
+                supportsGeometryShaderPassthrough: HwCapabilities.SupportsGeometryShaderPassthrough,
+                supportsImageLoadFormatted: HwCapabilities.SupportsImageLoadFormatted,
+                supportsLayerVertexTessellation: HwCapabilities.SupportsShaderViewportLayerArray,
+                supportsMismatchingViewFormat: HwCapabilities.SupportsMismatchingViewFormat,
+                supportsCubemapView: true,
+                supportsNonConstantTextureOffset: HwCapabilities.SupportsNonConstantTextureOffset,
+                supportsShaderBallot: HwCapabilities.SupportsShaderBallot,
+                supportsTextureShadowLod: HwCapabilities.SupportsTextureShadowLod,
+                supportsViewportIndexVertexTessellation: HwCapabilities.SupportsShaderViewportLayerArray,
+                supportsViewportMask: HwCapabilities.SupportsViewportArray2,
+                supportsViewportSwizzle: HwCapabilities.SupportsViewportSwizzle,
+                supportsIndirectParameters: HwCapabilities.SupportsIndirectParameters,
+                maximumUniformBuffersPerStage: 13, // TODO: Avoid hardcoding those limits here and get from driver?
+                maximumStorageBuffersPerStage: 16,
+                maximumTexturesPerStage: 32,
+                maximumImagesPerStage: 8,
+                maximumComputeSharedMemorySize: HwCapabilities.MaximumComputeSharedMemorySize,
+                maximumSupportedAnisotropy: HwCapabilities.MaximumSupportedAnisotropy,
+                storageBufferOffsetAlignment: HwCapabilities.StorageBufferOffsetAlignment,
+                gatherBiasPrecision: intelWindows || amdWindows ? 8 : 0); // Precision is 8 for these vendors on Vulkan.
+        }
+
+        public void SetBufferData(BufferHandle buffer, int offset, ReadOnlySpan<byte> data)
+        {
+            Buffer.SetData(buffer, offset, data);
+        }
+
+        public void UpdateCounters()
+        {
+            _counters.Update();
+        }
+
+        public void PreFrame()
+        {
+            _sync.Cleanup();
+            ResourcePool.Tick();
+        }
+
+        public ICounterEvent ReportCounter(CounterType type, EventHandler<ulong> resultHandler, bool hostReserved)
+        {
+            return _counters.QueueReport(type, resultHandler, _pipeline.DrawCount, hostReserved);
+        }
+
+        public void Initialize(GraphicsDebugLevel glLogLevel)
+        {
+            Debugger.Initialize(glLogLevel);
+
+            PrintGpuInformation();
+
+            if (HwCapabilities.SupportsParallelShaderCompile)
+            {
+                GL.Arb.MaxShaderCompilerThreads(Math.Min(Environment.ProcessorCount, 8));
+            }
+
+            _pipeline.Initialize(this);
+            _counters.Initialize(_pipeline);
+
+            // This is required to disable [0, 1] clamping for SNorm outputs on compatibility profiles.
+            // This call is expected to fail if we're running with a core profile,
+            // as this clamp target was deprecated, but that's fine as a core profile
+            // should already have the desired behaviour were outputs are not clamped.
+            GL.ClampColor(ClampColorTarget.ClampFragmentColor, ClampColorMode.False);
+        }
+
+        private void PrintGpuInformation()
+        {
+            GpuVendor   = GL.GetString(StringName.Vendor);
+            GpuRenderer = GL.GetString(StringName.Renderer);
+            GpuVersion  = GL.GetString(StringName.Version);
+
+            Logger.Notice.Print(LogClass.Gpu, $"{GpuVendor} {GpuRenderer} ({GpuVersion})");
+        }
+
+        public void ResetCounter(CounterType type)
+        {
+            _counters.QueueReset(type);
+        }
+
+        public void BackgroundContextAction(Action action, bool alwaysBackground = false)
+        {
+            // alwaysBackground is ignored, since we cannot switch from the current context.
+
+            if (IOpenGLContext.HasContext())
+            {
+                action(); // We have a context already - use that (assuming it is the main one).
+            }
+            else
+            {
+                _window.BackgroundContext.Invoke(action);
+            }
+        }
+
+        public void InitializeBackgroundContext(IOpenGLContext baseContext)
+        {
+            _window.InitializeBackgroundContext(baseContext);
+        }
+
+        public void Dispose()
+        {
+            _textureCopy.Dispose();
+            _backgroundTextureCopy.Dispose();
+            TextureCopyMS.Dispose();
+            PersistentBuffers.Dispose();
+            ResourcePool.Dispose();
+            _pipeline.Dispose();
+            _window.Dispose();
+            _counters.Dispose();
+            _sync.Dispose();
+        }
+
+        public IProgram LoadProgramBinary(byte[] programBinary, bool hasFragmentShader, ShaderInfo info)
+        {
+            return new Program(programBinary, hasFragmentShader, info.FragmentOutputMap);
+        }
+
+        public void CreateSync(ulong id, bool strict)
+        {
+            _sync.Create(id);
+        }
+
+        public void WaitSync(ulong id)
+        {
+            _sync.Wait(id);
+        }
+
+        public ulong GetCurrentSync()
+        {
+            return _sync.GetCurrent();
+        }
+
+        public void SetInterruptAction(Action<Action> interruptAction)
+        {
+            // Currently no need for an interrupt action.
+        }
+
+        public void Screenshot()
+        {
+            _window.ScreenCaptureRequested = true;
+        }
+
+        public void OnScreenCaptured(ScreenCaptureImageInfo bitmap)
+        {
+            ScreenCaptured?.Invoke(this, bitmap);
+        }
+    }
+}
diff --git a/src/Ryujinx.Graphics.OpenGL/PersistentBuffers.cs b/src/Ryujinx.Graphics.OpenGL/PersistentBuffers.cs
new file mode 100644
index 00000000..654e25b9
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/PersistentBuffers.cs
@@ -0,0 +1,136 @@
+using OpenTK.Graphics.OpenGL;
+using Ryujinx.Common.Logging;
+using Ryujinx.Graphics.GAL;
+using Ryujinx.Graphics.OpenGL.Image;
+using System;
+using System.Runtime.CompilerServices;
+using System.Runtime.InteropServices;
+
+namespace Ryujinx.Graphics.OpenGL
+{
+    class PersistentBuffers : IDisposable
+    {
+        private PersistentBuffer _main = new PersistentBuffer();
+        private PersistentBuffer _background = new PersistentBuffer();
+
+        public PersistentBuffer Default => BackgroundContextWorker.InBackground ? _background : _main;
+
+        public void Dispose()
+        {
+            _main?.Dispose();
+            _background?.Dispose();
+        }
+    }
+
+    class PersistentBuffer : IDisposable
+    {
+        private IntPtr _bufferMap;
+        private int _copyBufferHandle;
+        private int _copyBufferSize;
+
+        private byte[] _data;
+        private IntPtr _dataMap;
+
+        private void EnsureBuffer(int requiredSize)
+        {
+            if (_copyBufferSize < requiredSize && _copyBufferHandle != 0)
+            {
+                GL.DeleteBuffer(_copyBufferHandle);
+
+                _copyBufferHandle = 0;
+            }
+
+            if (_copyBufferHandle == 0)
+            {
+                _copyBufferHandle = GL.GenBuffer();
+                _copyBufferSize = requiredSize;
+
+                GL.BindBuffer(BufferTarget.CopyWriteBuffer, _copyBufferHandle);
+                GL.BufferStorage(BufferTarget.CopyWriteBuffer, requiredSize, IntPtr.Zero, BufferStorageFlags.MapReadBit | BufferStorageFlags.MapPersistentBit);
+
+                _bufferMap = GL.MapBufferRange(BufferTarget.CopyWriteBuffer, IntPtr.Zero, requiredSize, BufferAccessMask.MapReadBit | BufferAccessMask.MapPersistentBit);
+            }
+        }
+
+        public unsafe IntPtr GetHostArray(int requiredSize)
+        {
+            if (_data == null || _data.Length < requiredSize)
+            {
+                _data = GC.AllocateUninitializedArray<byte>(requiredSize, true);
+
+                _dataMap = (IntPtr)Unsafe.AsPointer(ref MemoryMarshal.GetArrayDataReference(_data));
+            }
+
+            return _dataMap;
+        }
+
+        private void Sync()
+        {
+            GL.MemoryBarrier(MemoryBarrierFlags.ClientMappedBufferBarrierBit);
+
+            IntPtr sync = GL.FenceSync(SyncCondition.SyncGpuCommandsComplete, WaitSyncFlags.None);
+            WaitSyncStatus syncResult = GL.ClientWaitSync(sync, ClientWaitSyncFlags.SyncFlushCommandsBit, 1000000000);
+
+            if (syncResult == WaitSyncStatus.TimeoutExpired)
+            {
+                Logger.Error?.PrintMsg(LogClass.Gpu, $"Failed to sync persistent buffer state within 1000ms. Continuing...");
+            }
+
+            GL.DeleteSync(sync);
+        }
+
+        public unsafe ReadOnlySpan<byte> GetTextureData(TextureView view, int size)
+        {
+            EnsureBuffer(size);
+
+            GL.BindBuffer(BufferTarget.PixelPackBuffer, _copyBufferHandle);
+
+            view.WriteToPbo(0, false);
+
+            GL.BindBuffer(BufferTarget.PixelPackBuffer, 0);
+
+            Sync();
+
+            return new ReadOnlySpan<byte>(_bufferMap.ToPointer(), size);
+        }
+
+        public unsafe ReadOnlySpan<byte> GetTextureData(TextureView view, int size, int layer, int level)
+        {
+            EnsureBuffer(size);
+
+            GL.BindBuffer(BufferTarget.PixelPackBuffer, _copyBufferHandle);
+
+            int offset = view.WriteToPbo2D(0, layer, level);
+
+            GL.BindBuffer(BufferTarget.PixelPackBuffer, 0);
+
+            Sync();
+
+            return new ReadOnlySpan<byte>(_bufferMap.ToPointer(), size).Slice(offset);
+        }
+
+        public unsafe ReadOnlySpan<byte> GetBufferData(BufferHandle buffer, int offset, int size)
+        {
+            EnsureBuffer(size);
+
+            GL.BindBuffer(BufferTarget.CopyReadBuffer, buffer.ToInt32());
+            GL.BindBuffer(BufferTarget.CopyWriteBuffer, _copyBufferHandle);
+
+            GL.CopyBufferSubData(BufferTarget.CopyReadBuffer, BufferTarget.CopyWriteBuffer, (IntPtr)offset, IntPtr.Zero, size);
+
+            GL.BindBuffer(BufferTarget.CopyWriteBuffer, 0);
+
+            Sync();
+
+            return new ReadOnlySpan<byte>(_bufferMap.ToPointer(), size);
+        }
+
+        public void Dispose()
+        {
+            if (_copyBufferHandle != 0)
+            {
+                GL.DeleteBuffer(_copyBufferHandle);
+            }
+        }
+    }
+}
diff --git a/src/Ryujinx.Graphics.OpenGL/Pipeline.cs b/src/Ryujinx.Graphics.OpenGL/Pipeline.cs
new file mode 100644
index 00000000..6b6d0289
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Pipeline.cs
@@ -0,0 +1,1770 @@
+using OpenTK.Graphics.OpenGL;
+using Ryujinx.Common.Logging;
+using Ryujinx.Graphics.GAL;
+using Ryujinx.Graphics.OpenGL.Image;
+using Ryujinx.Graphics.OpenGL.Queries;
+using Ryujinx.Graphics.Shader;
+using System;
+using System.Runtime.CompilerServices;
+
+namespace Ryujinx.Graphics.OpenGL
+{
+    class Pipeline : IPipeline, IDisposable
+    {
+        private const int SavedImages = 2;
+
+        private readonly DrawTextureEmulation _drawTexture;
+
+        internal ulong DrawCount { get; private set; }
+
+        private Program _program;
+
+        private bool _rasterizerDiscard;
+
+        private VertexArray _vertexArray;
+        private Framebuffer _framebuffer;
+
+        private IntPtr _indexBaseOffset;
+
+        private DrawElementsType _elementsType;
+
+        private PrimitiveType _primitiveType;
+
+        private int _stencilFrontMask;
+        private bool _depthMask;
+        private bool _depthTestEnable;
+        private bool _stencilTestEnable;
+        private bool _cullEnable;
+
+        private float[] _viewportArray = Array.Empty<float>();
+        private double[] _depthRangeArray = Array.Empty<double>();
+
+        private int _boundDrawFramebuffer;
+        private int _boundReadFramebuffer;
+
+        private CounterQueueEvent _activeConditionalRender;
+
+        private Vector4<int>[] _fpIsBgra = new Vector4<int>[SupportBuffer.FragmentIsBgraCount];
+        private Vector4<float>[] _renderScale = new Vector4<float>[73];
+        private int _fragmentScaleCount;
+
+        private (TextureBase, Format)[] _images;
+        private TextureBase _unit0Texture;
+        private Sampler _unit0Sampler;
+
+        private FrontFaceDirection _frontFace;
+        private ClipOrigin _clipOrigin;
+        private ClipDepthMode _clipDepthMode;
+
+        private uint _fragmentOutputMap;
+        private uint _componentMasks;
+        private uint _currentComponentMasks;
+        private bool _advancedBlendEnable;
+
+        private uint _scissorEnables;
+
+        private bool _tfEnabled;
+        private TransformFeedbackPrimitiveType _tfTopology;
+
+        private SupportBufferUpdater _supportBuffer;
+        private readonly BufferHandle[] _tfbs;
+        private readonly BufferRange[] _tfbTargets;
+
+        private ColorF _blendConstant;
+
+        internal Pipeline()
+        {
+            _drawTexture = new DrawTextureEmulation();
+            _rasterizerDiscard = false;
+            _clipOrigin = ClipOrigin.LowerLeft;
+            _clipDepthMode = ClipDepthMode.NegativeOneToOne;
+
+            _fragmentOutputMap = uint.MaxValue;
+            _componentMasks = uint.MaxValue;
+
+            _images = new (TextureBase, Format)[SavedImages];
+
+            var defaultScale = new Vector4<float> { X = 1f, Y = 0f, Z = 0f, W = 0f };
+            new Span<Vector4<float>>(_renderScale).Fill(defaultScale);
+
+            _tfbs = new BufferHandle[Constants.MaxTransformFeedbackBuffers];
+            _tfbTargets = new BufferRange[Constants.MaxTransformFeedbackBuffers];
+        }
+
+        public void Initialize(OpenGLRenderer renderer)
+        {
+            _supportBuffer = new SupportBufferUpdater(renderer);
+            GL.BindBufferBase(BufferRangeTarget.UniformBuffer, 0, Unsafe.As<BufferHandle, int>(ref _supportBuffer.Handle));
+
+            _supportBuffer.UpdateFragmentIsBgra(_fpIsBgra, 0, SupportBuffer.FragmentIsBgraCount);
+            _supportBuffer.UpdateRenderScale(_renderScale, 0, SupportBuffer.RenderScaleMaxCount);
+        }
+
+        public void Barrier()
+        {
+            GL.MemoryBarrier(MemoryBarrierFlags.AllBarrierBits);
+        }
+
+        public void BeginTransformFeedback(PrimitiveTopology topology)
+        {
+            GL.BeginTransformFeedback(_tfTopology = topology.ConvertToTfType());
+            _tfEnabled = true;
+        }
+
+        public void ClearBuffer(BufferHandle destination, int offset, int size, uint value)
+        {
+            Buffer.Clear(destination, offset, size, value);
+        }
+
+        public void ClearRenderTargetColor(int index, int layer, int layerCount, uint componentMask, ColorF color)
+        {
+            EnsureFramebuffer();
+
+            GL.ColorMask(
+                index,
+                (componentMask & 1) != 0,
+                (componentMask & 2) != 0,
+                (componentMask & 4) != 0,
+                (componentMask & 8) != 0);
+
+            float[] colors = new float[] { color.Red, color.Green, color.Blue, color.Alpha };
+
+            if (layer != 0 || layerCount != _framebuffer.GetColorLayerCount(index))
+            {
+                for (int l = layer; l < layer + layerCount; l++)
+                {
+                    _framebuffer.AttachColorLayerForClear(index, l);
+
+                    GL.ClearBuffer(OpenTK.Graphics.OpenGL.ClearBuffer.Color, index, colors);
+                }
+
+                _framebuffer.DetachColorLayerForClear(index);
+            }
+            else
+            {
+                GL.ClearBuffer(OpenTK.Graphics.OpenGL.ClearBuffer.Color, index, colors);
+            }
+
+            RestoreComponentMask(index);
+        }
+
+        public void ClearRenderTargetDepthStencil(int layer, int layerCount, float depthValue, bool depthMask, int stencilValue, int stencilMask)
+        {
+            EnsureFramebuffer();
+
+            bool stencilMaskChanged =
+                stencilMask != 0 &&
+                stencilMask != _stencilFrontMask;
+
+            bool depthMaskChanged = depthMask && depthMask != _depthMask;
+
+            if (stencilMaskChanged)
+            {
+                GL.StencilMaskSeparate(StencilFace.Front, stencilMask);
+            }
+
+            if (depthMaskChanged)
+            {
+                GL.DepthMask(depthMask);
+            }
+
+            if (layer != 0 || layerCount != _framebuffer.GetDepthStencilLayerCount())
+            {
+                for (int l = layer; l < layer + layerCount; l++)
+                {
+                    _framebuffer.AttachDepthStencilLayerForClear(l);
+
+                    ClearDepthStencil(depthValue, depthMask, stencilValue, stencilMask);
+                }
+
+                _framebuffer.DetachDepthStencilLayerForClear();
+            }
+            else
+            {
+                ClearDepthStencil(depthValue, depthMask, stencilValue, stencilMask);
+            }
+
+            if (stencilMaskChanged)
+            {
+                GL.StencilMaskSeparate(StencilFace.Front, _stencilFrontMask);
+            }
+
+            if (depthMaskChanged)
+            {
+                GL.DepthMask(_depthMask);
+            }
+        }
+
+        private static void ClearDepthStencil(float depthValue, bool depthMask, int stencilValue, int stencilMask)
+        {
+            if (depthMask && stencilMask != 0)
+            {
+                GL.ClearBuffer(ClearBufferCombined.DepthStencil, 0, depthValue, stencilValue);
+            }
+            else if (depthMask)
+            {
+                GL.ClearBuffer(OpenTK.Graphics.OpenGL.ClearBuffer.Depth, 0, ref depthValue);
+            }
+            else if (stencilMask != 0)
+            {
+                GL.ClearBuffer(OpenTK.Graphics.OpenGL.ClearBuffer.Stencil, 0, ref stencilValue);
+            }
+        }
+
+        public void CommandBufferBarrier()
+        {
+            GL.MemoryBarrier(MemoryBarrierFlags.CommandBarrierBit);
+        }
+
+        public void CopyBuffer(BufferHandle source, BufferHandle destination, int srcOffset, int dstOffset, int size)
+        {
+            Buffer.Copy(source, destination, srcOffset, dstOffset, size);
+        }
+
+        public void DispatchCompute(int groupsX, int groupsY, int groupsZ)
+        {
+            if (!_program.IsLinked)
+            {
+                Logger.Debug?.Print(LogClass.Gpu, "Dispatch error, shader not linked.");
+                return;
+            }
+
+            PrepareForDispatch();
+
+            GL.DispatchCompute(groupsX, groupsY, groupsZ);
+        }
+
+        public void Draw(int vertexCount, int instanceCount, int firstVertex, int firstInstance)
+        {
+            if (!_program.IsLinked)
+            {
+                Logger.Debug?.Print(LogClass.Gpu, "Draw error, shader not linked.");
+                return;
+            }
+
+            PreDraw(vertexCount);
+
+            if (_primitiveType == PrimitiveType.Quads && !HwCapabilities.SupportsQuads)
+            {
+                DrawQuadsImpl(vertexCount, instanceCount, firstVertex, firstInstance);
+            }
+            else if (_primitiveType == PrimitiveType.QuadStrip && !HwCapabilities.SupportsQuads)
+            {
+                DrawQuadStripImpl(vertexCount, instanceCount, firstVertex, firstInstance);
+            }
+            else
+            {
+                DrawImpl(vertexCount, instanceCount, firstVertex, firstInstance);
+            }
+
+            PostDraw();
+        }
+
+        private void DrawQuadsImpl(
+            int vertexCount,
+            int instanceCount,
+            int firstVertex,
+            int firstInstance)
+        {
+            // TODO: Instanced rendering.
+            int quadsCount = vertexCount / 4;
+
+            int[] firsts = new int[quadsCount];
+            int[] counts = new int[quadsCount];
+
+            for (int quadIndex = 0; quadIndex < quadsCount; quadIndex++)
+            {
+                firsts[quadIndex] = firstVertex + quadIndex * 4;
+                counts[quadIndex] = 4;
+            }
+
+            GL.MultiDrawArrays(
+                PrimitiveType.TriangleFan,
+                firsts,
+                counts,
+                quadsCount);
+        }
+
+        private void DrawQuadStripImpl(
+            int vertexCount,
+            int instanceCount,
+            int firstVertex,
+            int firstInstance)
+        {
+            int quadsCount = (vertexCount - 2) / 2;
+
+            if (firstInstance != 0 || instanceCount != 1)
+            {
+                for (int quadIndex = 0; quadIndex < quadsCount; quadIndex++)
+                {
+                    GL.DrawArraysInstancedBaseInstance(PrimitiveType.TriangleFan, firstVertex + quadIndex * 2, 4, instanceCount, firstInstance);
+                }
+            }
+            else
+            {
+                int[] firsts = new int[quadsCount];
+                int[] counts = new int[quadsCount];
+
+                firsts[0] = firstVertex;
+                counts[0] = 4;
+
+                for (int quadIndex = 1; quadIndex < quadsCount; quadIndex++)
+                {
+                    firsts[quadIndex] = firstVertex + quadIndex * 2;
+                    counts[quadIndex] = 4;
+                }
+
+                GL.MultiDrawArrays(
+                    PrimitiveType.TriangleFan,
+                    firsts,
+                    counts,
+                    quadsCount);
+            }
+        }
+
+        private void DrawImpl(
+            int vertexCount,
+            int instanceCount,
+            int firstVertex,
+            int firstInstance)
+        {
+            if (firstInstance == 0 && instanceCount == 1)
+            {
+                GL.DrawArrays(_primitiveType, firstVertex, vertexCount);
+            }
+            else if (firstInstance == 0)
+            {
+                GL.DrawArraysInstanced(_primitiveType, firstVertex, vertexCount, instanceCount);
+            }
+            else
+            {
+                GL.DrawArraysInstancedBaseInstance(
+                    _primitiveType,
+                    firstVertex,
+                    vertexCount,
+                    instanceCount,
+                    firstInstance);
+            }
+        }
+
+        public void DrawIndexed(
+            int indexCount,
+            int instanceCount,
+            int firstIndex,
+            int firstVertex,
+            int firstInstance)
+        {
+            if (!_program.IsLinked)
+            {
+                Logger.Debug?.Print(LogClass.Gpu, "Draw error, shader not linked.");
+                return;
+            }
+
+            PreDrawVbUnbounded();
+
+            int indexElemSize = 1;
+
+            switch (_elementsType)
+            {
+                case DrawElementsType.UnsignedShort: indexElemSize = 2; break;
+                case DrawElementsType.UnsignedInt: indexElemSize = 4; break;
+            }
+
+            IntPtr indexBaseOffset = _indexBaseOffset + firstIndex * indexElemSize;
+
+            if (_primitiveType == PrimitiveType.Quads && !HwCapabilities.SupportsQuads)
+            {
+                DrawQuadsIndexedImpl(
+                    indexCount,
+                    instanceCount,
+                    indexBaseOffset,
+                    indexElemSize,
+                    firstVertex,
+                    firstInstance);
+            }
+            else if (_primitiveType == PrimitiveType.QuadStrip && !HwCapabilities.SupportsQuads)
+            {
+                DrawQuadStripIndexedImpl(
+                    indexCount,
+                    instanceCount,
+                    indexBaseOffset,
+                    indexElemSize,
+                    firstVertex,
+                    firstInstance);
+            }
+            else
+            {
+                DrawIndexedImpl(
+                    indexCount,
+                    instanceCount,
+                    indexBaseOffset,
+                    firstVertex,
+                    firstInstance);
+            }
+
+            PostDraw();
+        }
+
+        private void DrawQuadsIndexedImpl(
+            int indexCount,
+            int instanceCount,
+            IntPtr indexBaseOffset,
+            int indexElemSize,
+            int firstVertex,
+            int firstInstance)
+        {
+            int quadsCount = indexCount / 4;
+
+            if (firstInstance != 0 || instanceCount != 1)
+            {
+                if (firstVertex != 0 && firstInstance != 0)
+                {
+                    for (int quadIndex = 0; quadIndex < quadsCount; quadIndex++)
+                    {
+                        GL.DrawElementsInstancedBaseVertexBaseInstance(
+                            PrimitiveType.TriangleFan,
+                            4,
+                            _elementsType,
+                            indexBaseOffset + quadIndex * 4 * indexElemSize,
+                            instanceCount,
+                            firstVertex,
+                            firstInstance);
+                    }
+                }
+                else if (firstInstance != 0)
+                {
+                    for (int quadIndex = 0; quadIndex < quadsCount; quadIndex++)
+                    {
+                        GL.DrawElementsInstancedBaseInstance(
+                            PrimitiveType.TriangleFan,
+                            4,
+                            _elementsType,
+                            indexBaseOffset + quadIndex * 4 * indexElemSize,
+                            instanceCount,
+                            firstInstance);
+                    }
+                }
+                else
+                {
+                    for (int quadIndex = 0; quadIndex < quadsCount; quadIndex++)
+                    {
+                        GL.DrawElementsInstanced(
+                            PrimitiveType.TriangleFan,
+                            4,
+                            _elementsType,
+                            indexBaseOffset + quadIndex * 4 * indexElemSize,
+                            instanceCount);
+                    }
+                }
+            }
+            else
+            {
+                IntPtr[] indices = new IntPtr[quadsCount];
+
+                int[] counts = new int[quadsCount];
+
+                int[] baseVertices = new int[quadsCount];
+
+                for (int quadIndex = 0; quadIndex < quadsCount; quadIndex++)
+                {
+                    indices[quadIndex] = indexBaseOffset + quadIndex * 4 * indexElemSize;
+
+                    counts[quadIndex] = 4;
+
+                    baseVertices[quadIndex] = firstVertex;
+                }
+
+                GL.MultiDrawElementsBaseVertex(
+                    PrimitiveType.TriangleFan,
+                    counts,
+                    _elementsType,
+                    indices,
+                    quadsCount,
+                    baseVertices);
+            }
+        }
+
+        private void DrawQuadStripIndexedImpl(
+            int indexCount,
+            int instanceCount,
+            IntPtr indexBaseOffset,
+            int indexElemSize,
+            int firstVertex,
+            int firstInstance)
+        {
+            // TODO: Instanced rendering.
+            int quadsCount = (indexCount - 2) / 2;
+
+            IntPtr[] indices = new IntPtr[quadsCount];
+
+            int[] counts = new int[quadsCount];
+
+            int[] baseVertices = new int[quadsCount];
+
+            indices[0] = indexBaseOffset;
+
+            counts[0] = 4;
+
+            baseVertices[0] = firstVertex;
+
+            for (int quadIndex = 1; quadIndex < quadsCount; quadIndex++)
+            {
+                indices[quadIndex] = indexBaseOffset + quadIndex * 2 * indexElemSize;
+
+                counts[quadIndex] = 4;
+
+                baseVertices[quadIndex] = firstVertex;
+            }
+
+            GL.MultiDrawElementsBaseVertex(
+                PrimitiveType.TriangleFan,
+                counts,
+                _elementsType,
+                indices,
+                quadsCount,
+                baseVertices);
+        }
+
+        private void DrawIndexedImpl(
+            int indexCount,
+            int instanceCount,
+            IntPtr indexBaseOffset,
+            int firstVertex,
+            int firstInstance)
+        {
+            if (firstInstance == 0 && firstVertex == 0 && instanceCount == 1)
+            {
+                GL.DrawElements(_primitiveType, indexCount, _elementsType, indexBaseOffset);
+            }
+            else if (firstInstance == 0 && instanceCount == 1)
+            {
+                GL.DrawElementsBaseVertex(
+                    _primitiveType,
+                    indexCount,
+                    _elementsType,
+                    indexBaseOffset,
+                    firstVertex);
+            }
+            else if (firstInstance == 0 && firstVertex == 0)
+            {
+                GL.DrawElementsInstanced(
+                    _primitiveType,
+                    indexCount,
+                    _elementsType,
+                    indexBaseOffset,
+                    instanceCount);
+            }
+            else if (firstInstance == 0)
+            {
+                GL.DrawElementsInstancedBaseVertex(
+                    _primitiveType,
+                    indexCount,
+                    _elementsType,
+                    indexBaseOffset,
+                    instanceCount,
+                    firstVertex);
+            }
+            else if (firstVertex == 0)
+            {
+                GL.DrawElementsInstancedBaseInstance(
+                    _primitiveType,
+                    indexCount,
+                    _elementsType,
+                    indexBaseOffset,
+                    instanceCount,
+                    firstInstance);
+            }
+            else
+            {
+                GL.DrawElementsInstancedBaseVertexBaseInstance(
+                    _primitiveType,
+                    indexCount,
+                    _elementsType,
+                    indexBaseOffset,
+                    instanceCount,
+                    firstVertex,
+                    firstInstance);
+            }
+        }
+
+        public void DrawIndexedIndirect(BufferRange indirectBuffer)
+        {
+            if (!_program.IsLinked)
+            {
+                Logger.Debug?.Print(LogClass.Gpu, "Draw error, shader not linked.");
+                return;
+            }
+
+            PreDrawVbUnbounded();
+
+            _vertexArray.SetRangeOfIndexBuffer();
+
+            GL.BindBuffer((BufferTarget)All.DrawIndirectBuffer, indirectBuffer.Handle.ToInt32());
+
+            GL.DrawElementsIndirect(_primitiveType, _elementsType, (IntPtr)indirectBuffer.Offset);
+
+            _vertexArray.RestoreIndexBuffer();
+
+            PostDraw();
+        }
+
+        public void DrawIndexedIndirectCount(BufferRange indirectBuffer, BufferRange parameterBuffer, int maxDrawCount, int stride)
+        {
+            if (!_program.IsLinked)
+            {
+                Logger.Debug?.Print(LogClass.Gpu, "Draw error, shader not linked.");
+                return;
+            }
+
+            PreDrawVbUnbounded();
+
+            _vertexArray.SetRangeOfIndexBuffer();
+
+            GL.BindBuffer((BufferTarget)All.DrawIndirectBuffer, indirectBuffer.Handle.ToInt32());
+            GL.BindBuffer((BufferTarget)All.ParameterBuffer, parameterBuffer.Handle.ToInt32());
+
+            GL.MultiDrawElementsIndirectCount(
+                _primitiveType,
+                (All)_elementsType,
+                (IntPtr)indirectBuffer.Offset,
+                (IntPtr)parameterBuffer.Offset,
+                maxDrawCount,
+                stride);
+
+            _vertexArray.RestoreIndexBuffer();
+
+            PostDraw();
+        }
+
+        public void DrawIndirect(BufferRange indirectBuffer)
+        {
+            if (!_program.IsLinked)
+            {
+                Logger.Debug?.Print(LogClass.Gpu, "Draw error, shader not linked.");
+                return;
+            }
+
+            PreDrawVbUnbounded();
+
+            GL.BindBuffer((BufferTarget)All.DrawIndirectBuffer, indirectBuffer.Handle.ToInt32());
+
+            GL.DrawArraysIndirect(_primitiveType, (IntPtr)indirectBuffer.Offset);
+
+            PostDraw();
+        }
+
+        public void DrawIndirectCount(BufferRange indirectBuffer, BufferRange parameterBuffer, int maxDrawCount, int stride)
+        {
+            if (!_program.IsLinked)
+            {
+                Logger.Debug?.Print(LogClass.Gpu, "Draw error, shader not linked.");
+                return;
+            }
+
+            PreDrawVbUnbounded();
+
+            GL.BindBuffer((BufferTarget)All.DrawIndirectBuffer, indirectBuffer.Handle.ToInt32());
+            GL.BindBuffer((BufferTarget)All.ParameterBuffer, parameterBuffer.Handle.ToInt32());
+
+            GL.MultiDrawArraysIndirectCount(
+                _primitiveType,
+                (IntPtr)indirectBuffer.Offset,
+                (IntPtr)parameterBuffer.Offset,
+                maxDrawCount,
+                stride);
+
+            PostDraw();
+        }
+
+        public void DrawTexture(ITexture texture, ISampler sampler, Extents2DF srcRegion, Extents2DF dstRegion)
+        {
+            if (texture is TextureView view && sampler is Sampler samp)
+            {
+                _supportBuffer.Commit();
+
+                if (HwCapabilities.SupportsDrawTexture)
+                {
+                    GL.NV.DrawTexture(
+                        view.Handle,
+                        samp.Handle,
+                        dstRegion.X1,
+                        dstRegion.Y1,
+                        dstRegion.X2,
+                        dstRegion.Y2,
+                        0,
+                        srcRegion.X1 / view.Width,
+                        srcRegion.Y1 / view.Height,
+                        srcRegion.X2 / view.Width,
+                        srcRegion.Y2 / view.Height);
+                }
+                else
+                {
+                    static void Disable(EnableCap cap, bool enabled)
+                    {
+                        if (enabled)
+                        {
+                            GL.Disable(cap);
+                        }
+                    }
+
+                    static void Enable(EnableCap cap, bool enabled)
+                    {
+                        if (enabled)
+                        {
+                            GL.Enable(cap);
+                        }
+                    }
+
+                    Disable(EnableCap.CullFace, _cullEnable);
+                    Disable(EnableCap.StencilTest, _stencilTestEnable);
+                    Disable(EnableCap.DepthTest, _depthTestEnable);
+
+                    if (_depthMask)
+                    {
+                        GL.DepthMask(false);
+                    }
+
+                    if (_tfEnabled)
+                    {
+                        GL.EndTransformFeedback();
+                    }
+
+                    GL.ClipControl(ClipOrigin.UpperLeft, ClipDepthMode.NegativeOneToOne);
+
+                    _drawTexture.Draw(
+                        view,
+                        samp,
+                        dstRegion.X1,
+                        dstRegion.Y1,
+                        dstRegion.X2,
+                        dstRegion.Y2,
+                        srcRegion.X1 / view.Width,
+                        srcRegion.Y1 / view.Height,
+                        srcRegion.X2 / view.Width,
+                        srcRegion.Y2 / view.Height);
+
+                    _program?.Bind();
+                    _unit0Sampler?.Bind(0);
+
+                    RestoreViewport0();
+
+                    Enable(EnableCap.CullFace, _cullEnable);
+                    Enable(EnableCap.StencilTest, _stencilTestEnable);
+                    Enable(EnableCap.DepthTest, _depthTestEnable);
+
+                    if (_depthMask)
+                    {
+                        GL.DepthMask(true);
+                    }
+
+                    if (_tfEnabled)
+                    {
+                        GL.BeginTransformFeedback(_tfTopology);
+                    }
+
+                    RestoreClipControl();
+                }
+            }
+        }
+
+        public void EndTransformFeedback()
+        {
+            GL.EndTransformFeedback();
+            _tfEnabled = false;
+        }
+
+        public double GetCounterDivisor(CounterType type)
+        {
+            if (type == CounterType.SamplesPassed)
+            {
+                return _renderScale[0].X * _renderScale[0].X;
+            }
+
+            return 1;
+        }
+
+        public void SetAlphaTest(bool enable, float reference, CompareOp op)
+        {
+            if (!enable)
+            {
+                GL.Disable(EnableCap.AlphaTest);
+                return;
+            }
+
+            GL.AlphaFunc((AlphaFunction)op.Convert(), reference);
+            GL.Enable(EnableCap.AlphaTest);
+        }
+
+        public void SetBlendState(AdvancedBlendDescriptor blend)
+        {
+            if (HwCapabilities.SupportsBlendEquationAdvanced)
+            {
+                GL.BlendEquation((BlendEquationMode)blend.Op.Convert());
+                GL.NV.BlendParameter(NvBlendEquationAdvanced.BlendOverlapNv, (int)blend.Overlap.Convert());
+                GL.NV.BlendParameter(NvBlendEquationAdvanced.BlendPremultipliedSrcNv, blend.SrcPreMultiplied ? 1 : 0);
+                GL.Enable(EnableCap.Blend);
+                _advancedBlendEnable = true;
+            }
+        }
+
+        public void SetBlendState(int index, BlendDescriptor blend)
+        {
+            if (_advancedBlendEnable)
+            {
+                GL.Disable(EnableCap.Blend);
+                _advancedBlendEnable = false;
+            }
+
+            if (!blend.Enable)
+            {
+                GL.Disable(IndexedEnableCap.Blend, index);
+                return;
+            }
+
+            GL.BlendEquationSeparate(
+                index,
+                blend.ColorOp.Convert(),
+                blend.AlphaOp.Convert());
+
+            GL.BlendFuncSeparate(
+                index,
+                (BlendingFactorSrc)blend.ColorSrcFactor.Convert(),
+                (BlendingFactorDest)blend.ColorDstFactor.Convert(),
+                (BlendingFactorSrc)blend.AlphaSrcFactor.Convert(),
+                (BlendingFactorDest)blend.AlphaDstFactor.Convert());
+
+            EnsureFramebuffer();
+
+            _framebuffer.SetDualSourceBlend(
+                blend.ColorSrcFactor.IsDualSource() ||
+                blend.ColorDstFactor.IsDualSource() ||
+                blend.AlphaSrcFactor.IsDualSource() ||
+                blend.AlphaDstFactor.IsDualSource());
+
+            if (_blendConstant != blend.BlendConstant)
+            {
+                _blendConstant = blend.BlendConstant;
+
+                GL.BlendColor(
+                    blend.BlendConstant.Red,
+                    blend.BlendConstant.Green,
+                    blend.BlendConstant.Blue,
+                    blend.BlendConstant.Alpha);
+            }
+
+            GL.Enable(IndexedEnableCap.Blend, index);
+        }
+
+        public void SetDepthBias(PolygonModeMask enables, float factor, float units, float clamp)
+        {
+            if ((enables & PolygonModeMask.Point) != 0)
+            {
+                GL.Enable(EnableCap.PolygonOffsetPoint);
+            }
+            else
+            {
+                GL.Disable(EnableCap.PolygonOffsetPoint);
+            }
+
+            if ((enables & PolygonModeMask.Line) != 0)
+            {
+                GL.Enable(EnableCap.PolygonOffsetLine);
+            }
+            else
+            {
+                GL.Disable(EnableCap.PolygonOffsetLine);
+            }
+
+            if ((enables & PolygonModeMask.Fill) != 0)
+            {
+                GL.Enable(EnableCap.PolygonOffsetFill);
+            }
+            else
+            {
+                GL.Disable(EnableCap.PolygonOffsetFill);
+            }
+
+            if (enables == 0)
+            {
+                return;
+            }
+
+            if (HwCapabilities.SupportsPolygonOffsetClamp)
+            {
+                GL.PolygonOffsetClamp(factor, units, clamp);
+            }
+            else
+            {
+                GL.PolygonOffset(factor, units);
+            }
+        }
+
+        public void SetDepthClamp(bool clamp)
+        {
+            if (!clamp)
+            {
+                GL.Disable(EnableCap.DepthClamp);
+                return;
+            }
+
+            GL.Enable(EnableCap.DepthClamp);
+        }
+
+        public void SetDepthMode(DepthMode mode)
+        {
+            ClipDepthMode depthMode = mode.Convert();
+
+            if (_clipDepthMode != depthMode)
+            {
+                _clipDepthMode = depthMode;
+
+                GL.ClipControl(_clipOrigin, depthMode);
+            }
+        }
+
+        public void SetDepthTest(DepthTestDescriptor depthTest)
+        {
+            if (depthTest.TestEnable)
+            {
+                GL.Enable(EnableCap.DepthTest);
+                GL.DepthFunc((DepthFunction)depthTest.Func.Convert());
+            }
+            else
+            {
+                GL.Disable(EnableCap.DepthTest);
+            }
+
+            GL.DepthMask(depthTest.WriteEnable);
+            _depthMask = depthTest.WriteEnable;
+            _depthTestEnable = depthTest.TestEnable;
+        }
+
+        public void SetFaceCulling(bool enable, Face face)
+        {
+            _cullEnable = enable;
+
+            if (!enable)
+            {
+                GL.Disable(EnableCap.CullFace);
+                return;
+            }
+
+            GL.CullFace(face.Convert());
+
+            GL.Enable(EnableCap.CullFace);
+        }
+
+        public void SetFrontFace(FrontFace frontFace)
+        {
+            SetFrontFace(_frontFace = frontFace.Convert());
+        }
+
+        public void SetImage(int binding, ITexture texture, Format imageFormat)
+        {
+            if ((uint)binding < SavedImages)
+            {
+                _images[binding] = (texture as TextureBase, imageFormat);
+            }
+
+            if (texture == null)
+            {
+                GL.BindImageTexture(binding, 0, 0, true, 0, TextureAccess.ReadWrite, SizedInternalFormat.Rgba8);
+                return;
+            }
+
+            TextureBase texBase = (TextureBase)texture;
+
+            SizedInternalFormat format = FormatTable.GetImageFormat(imageFormat);
+
+            if (format != 0)
+            {
+                GL.BindImageTexture(binding, texBase.Handle, 0, true, 0, TextureAccess.ReadWrite, format);
+            }
+        }
+
+        public void SetIndexBuffer(BufferRange buffer, IndexType type)
+        {
+            _elementsType = type.Convert();
+
+            _indexBaseOffset = (IntPtr)buffer.Offset;
+
+            EnsureVertexArray();
+
+            _vertexArray.SetIndexBuffer(buffer);
+        }
+
+        public void SetLogicOpState(bool enable, LogicalOp op)
+        {
+            if (enable)
+            {
+                GL.Enable(EnableCap.ColorLogicOp);
+
+                GL.LogicOp((LogicOp)op.Convert());
+            }
+            else
+            {
+                GL.Disable(EnableCap.ColorLogicOp);
+            }
+        }
+
+        public void SetMultisampleState(MultisampleDescriptor multisample)
+        {
+            if (multisample.AlphaToCoverageEnable)
+            {
+                GL.Enable(EnableCap.SampleAlphaToCoverage);
+
+                if (multisample.AlphaToOneEnable)
+                {
+                    GL.Enable(EnableCap.SampleAlphaToOne);
+                }
+                else
+                {
+                    GL.Disable(EnableCap.SampleAlphaToOne);
+                }
+
+                if (HwCapabilities.SupportsAlphaToCoverageDitherControl)
+                {
+                    GL.NV.AlphaToCoverageDitherControl(multisample.AlphaToCoverageDitherEnable
+                        ? NvAlphaToCoverageDitherControl.AlphaToCoverageDitherEnableNv
+                        : NvAlphaToCoverageDitherControl.AlphaToCoverageDitherDisableNv);
+                }
+            }
+            else
+            {
+                GL.Disable(EnableCap.SampleAlphaToCoverage);
+            }
+        }
+
+        public void SetLineParameters(float width, bool smooth)
+        {
+            if (smooth)
+            {
+                GL.Enable(EnableCap.LineSmooth);
+            }
+            else
+            {
+                GL.Disable(EnableCap.LineSmooth);
+            }
+
+            GL.LineWidth(width);
+        }
+
+        public unsafe void SetPatchParameters(int vertices, ReadOnlySpan<float> defaultOuterLevel, ReadOnlySpan<float> defaultInnerLevel)
+        {
+            GL.PatchParameter(PatchParameterInt.PatchVertices, vertices);
+
+            fixed (float* pOuterLevel = defaultOuterLevel)
+            {
+                GL.PatchParameter(PatchParameterFloat.PatchDefaultOuterLevel, pOuterLevel);
+            }
+
+            fixed (float* pInnerLevel = defaultInnerLevel)
+            {
+                GL.PatchParameter(PatchParameterFloat.PatchDefaultInnerLevel, pInnerLevel);
+            }
+        }
+
+        public void SetPointParameters(float size, bool isProgramPointSize, bool enablePointSprite, Origin origin)
+        {
+            // GL_POINT_SPRITE was deprecated in core profile 3.2+ and causes GL_INVALID_ENUM when set.
+            // As we don't know if the current context is core or compat, it's safer to keep this code.
+            if (enablePointSprite)
+            {
+                GL.Enable(EnableCap.PointSprite);
+            }
+            else
+            {
+                GL.Disable(EnableCap.PointSprite);
+            }
+
+            if (isProgramPointSize)
+            {
+                GL.Enable(EnableCap.ProgramPointSize);
+            }
+            else
+            {
+                GL.Disable(EnableCap.ProgramPointSize);
+            }
+
+            GL.PointParameter(origin == Origin.LowerLeft
+                ? PointSpriteCoordOriginParameter.LowerLeft
+                : PointSpriteCoordOriginParameter.UpperLeft);
+
+            // Games seem to set point size to 0 which generates a GL_INVALID_VALUE
+            // From the spec, GL_INVALID_VALUE is generated if size is less than or equal to 0.
+            GL.PointSize(Math.Max(float.Epsilon, size));
+        }
+
+        public void SetPolygonMode(GAL.PolygonMode frontMode, GAL.PolygonMode backMode)
+        {
+            if (frontMode == backMode)
+            {
+                GL.PolygonMode(MaterialFace.FrontAndBack, frontMode.Convert());
+            }
+            else
+            {
+                GL.PolygonMode(MaterialFace.Front, frontMode.Convert());
+                GL.PolygonMode(MaterialFace.Back, backMode.Convert());
+            }
+        }
+
+        public void SetPrimitiveRestart(bool enable, int index)
+        {
+            if (!enable)
+            {
+                GL.Disable(EnableCap.PrimitiveRestart);
+                return;
+            }
+
+            GL.PrimitiveRestartIndex(index);
+
+            GL.Enable(EnableCap.PrimitiveRestart);
+        }
+
+        public void SetPrimitiveTopology(PrimitiveTopology topology)
+        {
+            _primitiveType = topology.Convert();
+        }
+
+        public void SetProgram(IProgram program)
+        {
+            Program prg = (Program)program;
+
+            if (_tfEnabled)
+            {
+                GL.EndTransformFeedback();
+                prg.Bind();
+                GL.BeginTransformFeedback(_tfTopology);
+            }
+            else
+            {
+                prg.Bind();
+            }
+
+            if (prg.HasFragmentShader && _fragmentOutputMap != (uint)prg.FragmentOutputMap)
+            {
+                _fragmentOutputMap = (uint)prg.FragmentOutputMap;
+
+                for (int index = 0; index < Constants.MaxRenderTargets; index++)
+                {
+                    RestoreComponentMask(index, force: false);
+                }
+            }
+
+            _program = prg;
+        }
+
+        public void SetRasterizerDiscard(bool discard)
+        {
+            if (discard)
+            {
+                GL.Enable(EnableCap.RasterizerDiscard);
+            }
+            else
+            {
+                GL.Disable(EnableCap.RasterizerDiscard);
+            }
+
+            _rasterizerDiscard = discard;
+        }
+
+        public void SetRenderTargetScale(float scale)
+        {
+            _renderScale[0].X = scale;
+            _supportBuffer.UpdateRenderScale(_renderScale, 0, 1); // Just the first element.
+        }
+
+        public void SetRenderTargetColorMasks(ReadOnlySpan<uint> componentMasks)
+        {
+            _componentMasks = 0;
+
+            for (int index = 0; index < componentMasks.Length; index++)
+            {
+                _componentMasks |= componentMasks[index] << (index * 4);
+
+                RestoreComponentMask(index, force: false);
+            }
+        }
+
+        public void SetRenderTargets(ITexture[] colors, ITexture depthStencil)
+        {
+            EnsureFramebuffer();
+
+            bool isBgraChanged = false;
+
+            for (int index = 0; index < colors.Length; index++)
+            {
+                TextureView color = (TextureView)colors[index];
+
+                _framebuffer.AttachColor(index, color);
+
+                if (color != null)
+                {
+                    int isBgra = color.Format.IsBgr() ? 1 : 0;
+
+                    if (_fpIsBgra[index].X != isBgra)
+                    {
+                        _fpIsBgra[index].X = isBgra;
+                        isBgraChanged = true;
+
+                        RestoreComponentMask(index);
+                    }
+                }
+            }
+
+            if (isBgraChanged)
+            {
+                _supportBuffer.UpdateFragmentIsBgra(_fpIsBgra, 0, SupportBuffer.FragmentIsBgraCount);
+            }
+
+            TextureView depthStencilView = (TextureView)depthStencil;
+
+            _framebuffer.AttachDepthStencil(depthStencilView);
+            _framebuffer.SetDrawBuffers(colors.Length);
+        }
+
+        public void SetScissors(ReadOnlySpan<Rectangle<int>> regions)
+        {
+            int count = Math.Min(regions.Length, Constants.MaxViewports);
+
+            Span<int> v = stackalloc int[count * 4];
+
+            for (int index = 0; index < count; index++)
+            {
+                int vIndex = index * 4;
+
+                var region = regions[index];
+
+                bool enabled = (region.X | region.Y) != 0 || region.Width != 0xffff || region.Height != 0xffff;
+                uint mask = 1u << index;
+
+                if (enabled)
+                {
+                    v[vIndex] = region.X;
+                    v[vIndex + 1] = region.Y;
+                    v[vIndex + 2] = region.Width;
+                    v[vIndex + 3] = region.Height;
+
+                    if ((_scissorEnables & mask) == 0)
+                    {
+                        _scissorEnables |= mask;
+                        GL.Enable(IndexedEnableCap.ScissorTest, index);
+                    }
+                }
+                else
+                {
+                    if ((_scissorEnables & mask) != 0)
+                    {
+                        _scissorEnables &= ~mask;
+                        GL.Disable(IndexedEnableCap.ScissorTest, index);
+                    }
+                }
+            }
+
+            GL.ScissorArray(0, count, ref v[0]);
+        }
+
+        public void SetStencilTest(StencilTestDescriptor stencilTest)
+        {
+            _stencilTestEnable = stencilTest.TestEnable;
+
+            if (!stencilTest.TestEnable)
+            {
+                GL.Disable(EnableCap.StencilTest);
+                return;
+            }
+
+            GL.StencilOpSeparate(
+                StencilFace.Front,
+                stencilTest.FrontSFail.Convert(),
+                stencilTest.FrontDpFail.Convert(),
+                stencilTest.FrontDpPass.Convert());
+
+            GL.StencilFuncSeparate(
+                StencilFace.Front,
+                (StencilFunction)stencilTest.FrontFunc.Convert(),
+                stencilTest.FrontFuncRef,
+                stencilTest.FrontFuncMask);
+
+            GL.StencilMaskSeparate(StencilFace.Front, stencilTest.FrontMask);
+
+            GL.StencilOpSeparate(
+                StencilFace.Back,
+                stencilTest.BackSFail.Convert(),
+                stencilTest.BackDpFail.Convert(),
+                stencilTest.BackDpPass.Convert());
+
+            GL.StencilFuncSeparate(
+                StencilFace.Back,
+                (StencilFunction)stencilTest.BackFunc.Convert(),
+                stencilTest.BackFuncRef,
+                stencilTest.BackFuncMask);
+
+            GL.StencilMaskSeparate(StencilFace.Back, stencilTest.BackMask);
+
+            GL.Enable(EnableCap.StencilTest);
+
+            _stencilFrontMask = stencilTest.FrontMask;
+        }
+
+        public void SetStorageBuffers(ReadOnlySpan<BufferAssignment> buffers)
+        {
+            SetBuffers(buffers, isStorage: true);
+        }
+
+        public void SetTextureAndSampler(ShaderStage stage, int binding, ITexture texture, ISampler sampler)
+        {
+            if (texture != null)
+            {
+                if (binding == 0)
+                {
+                    _unit0Texture = (TextureBase)texture;
+                }
+                else
+                {
+                    ((TextureBase)texture).Bind(binding);
+                }
+            }
+            else
+            {
+                TextureBase.ClearBinding(binding);
+            }
+
+            Sampler glSampler = (Sampler)sampler;
+
+            glSampler?.Bind(binding);
+
+            if (binding == 0)
+            {
+                _unit0Sampler = glSampler;
+            }
+        }
+
+
+        public void SetTransformFeedbackBuffers(ReadOnlySpan<BufferRange> buffers)
+        {
+            if (_tfEnabled)
+            {
+                GL.EndTransformFeedback();
+            }
+
+            int count = Math.Min(buffers.Length, Constants.MaxTransformFeedbackBuffers);
+
+            for (int i = 0; i < count; i++)
+            {
+                BufferRange buffer = buffers[i];
+                _tfbTargets[i] = buffer;
+
+                if (buffer.Handle == BufferHandle.Null)
+                {
+                    GL.BindBufferBase(BufferRangeTarget.TransformFeedbackBuffer, i, 0);
+                    continue;
+                }
+
+                if (_tfbs[i] == BufferHandle.Null)
+                {
+                    _tfbs[i] = Buffer.Create();
+                }
+
+                Buffer.Resize(_tfbs[i], buffer.Size);
+                Buffer.Copy(buffer.Handle, _tfbs[i], buffer.Offset, 0, buffer.Size);
+                GL.BindBufferBase(BufferRangeTarget.TransformFeedbackBuffer, i, _tfbs[i].ToInt32());
+            }
+
+            if (_tfEnabled)
+            {
+                GL.BeginTransformFeedback(_tfTopology);
+            }
+        }
+
+        public void SetUniformBuffers(ReadOnlySpan<BufferAssignment> buffers)
+        {
+            SetBuffers(buffers, isStorage: false);
+        }
+
+        public void SetUserClipDistance(int index, bool enableClip)
+        {
+            if (!enableClip)
+            {
+                GL.Disable(EnableCap.ClipDistance0 + index);
+                return;
+            }
+
+            GL.Enable(EnableCap.ClipDistance0 + index);
+        }
+
+        public void SetVertexAttribs(ReadOnlySpan<VertexAttribDescriptor> vertexAttribs)
+        {
+            EnsureVertexArray();
+
+            _vertexArray.SetVertexAttributes(vertexAttribs);
+        }
+
+        public void SetVertexBuffers(ReadOnlySpan<VertexBufferDescriptor> vertexBuffers)
+        {
+            EnsureVertexArray();
+
+            _vertexArray.SetVertexBuffers(vertexBuffers);
+        }
+
+        public void SetViewports(ReadOnlySpan<Viewport> viewports, bool disableTransform)
+        {
+            Array.Resize(ref _viewportArray, viewports.Length * 4);
+            Array.Resize(ref _depthRangeArray, viewports.Length * 2);
+
+            float[] viewportArray = _viewportArray;
+            double[] depthRangeArray = _depthRangeArray;
+
+            for (int index = 0; index < viewports.Length; index++)
+            {
+                int viewportElemIndex = index * 4;
+
+                Viewport viewport = viewports[index];
+
+                viewportArray[viewportElemIndex + 0] = viewport.Region.X;
+                viewportArray[viewportElemIndex + 1] = viewport.Region.Y + (viewport.Region.Height < 0 ? viewport.Region.Height : 0);
+                viewportArray[viewportElemIndex + 2] = viewport.Region.Width;
+                viewportArray[viewportElemIndex + 3] = MathF.Abs(viewport.Region.Height);
+
+                if (HwCapabilities.SupportsViewportSwizzle)
+                {
+                    GL.NV.ViewportSwizzle(
+                        index,
+                        viewport.SwizzleX.Convert(),
+                        viewport.SwizzleY.Convert(),
+                        viewport.SwizzleZ.Convert(),
+                        viewport.SwizzleW.Convert());
+                }
+
+                depthRangeArray[index * 2 + 0] = viewport.DepthNear;
+                depthRangeArray[index * 2 + 1] = viewport.DepthFar;
+            }
+
+            bool flipY = viewports.Length != 0 && viewports[0].Region.Height < 0;
+
+            SetOrigin(flipY ? ClipOrigin.UpperLeft : ClipOrigin.LowerLeft);
+
+            GL.ViewportArray(0, viewports.Length, viewportArray);
+            GL.DepthRangeArray(0, viewports.Length, depthRangeArray);
+
+            float disableTransformF = disableTransform ? 1.0f : 0.0f;
+            if (_supportBuffer.Data.ViewportInverse.W != disableTransformF || disableTransform)
+            {
+                float scale = _renderScale[0].X;
+                _supportBuffer.UpdateViewportInverse(new Vector4<float>
+                {
+                    X = scale * 2f / viewports[0].Region.Width,
+                    Y = scale * 2f / viewports[0].Region.Height,
+                    Z = 1,
+                    W = disableTransformF
+                });
+            }
+        }
+
+        public void TextureBarrier()
+        {
+            GL.MemoryBarrier(MemoryBarrierFlags.TextureFetchBarrierBit);
+        }
+
+        public void TextureBarrierTiled()
+        {
+            GL.MemoryBarrier(MemoryBarrierFlags.TextureFetchBarrierBit);
+        }
+
+        private void SetBuffers(ReadOnlySpan<BufferAssignment> buffers, bool isStorage)
+        {
+            BufferRangeTarget target = isStorage ? BufferRangeTarget.ShaderStorageBuffer : BufferRangeTarget.UniformBuffer;
+
+            for (int index = 0; index < buffers.Length; index++)
+            {
+                BufferAssignment assignment = buffers[index];
+                BufferRange buffer = assignment.Range;
+
+                if (buffer.Handle == BufferHandle.Null)
+                {
+                    GL.BindBufferRange(target, assignment.Binding, 0, IntPtr.Zero, 0);
+                    continue;
+                }
+
+                GL.BindBufferRange(target, assignment.Binding, buffer.Handle.ToInt32(), (IntPtr)buffer.Offset, buffer.Size);
+            }
+        }
+
+        private void SetOrigin(ClipOrigin origin)
+        {
+            if (_clipOrigin != origin)
+            {
+                _clipOrigin = origin;
+
+                GL.ClipControl(origin, _clipDepthMode);
+
+                SetFrontFace(_frontFace);
+            }
+        }
+
+        private void SetFrontFace(FrontFaceDirection frontFace)
+        {
+            // Changing clip origin will also change the front face to compensate
+            // for the flipped viewport, we flip it again here to compensate as
+            // this effect is undesirable for us.
+            if (_clipOrigin == ClipOrigin.UpperLeft)
+            {
+                frontFace = frontFace == FrontFaceDirection.Ccw ? FrontFaceDirection.Cw : FrontFaceDirection.Ccw;
+            }
+
+            GL.FrontFace(frontFace);
+        }
+
+        private void EnsureVertexArray()
+        {
+            if (_vertexArray == null)
+            {
+                _vertexArray = new VertexArray();
+
+                _vertexArray.Bind();
+            }
+        }
+
+        private void EnsureFramebuffer()
+        {
+            if (_framebuffer == null)
+            {
+                _framebuffer = new Framebuffer();
+
+                int boundHandle = _framebuffer.Bind();
+                _boundDrawFramebuffer = _boundReadFramebuffer = boundHandle;
+
+                GL.Enable(EnableCap.FramebufferSrgb);
+            }
+        }
+
+        internal (int drawHandle, int readHandle) GetBoundFramebuffers()
+        {
+            if (BackgroundContextWorker.InBackground)
+            {
+                return (0, 0);
+            }
+
+            return (_boundDrawFramebuffer, _boundReadFramebuffer);
+        }
+
+        public void UpdateRenderScale(ReadOnlySpan<float> scales, int totalCount, int fragmentCount)
+        {
+            bool changed = false;
+
+            for (int index = 0; index < totalCount; index++)
+            {
+                if (_renderScale[1 + index].X != scales[index])
+                {
+                    _renderScale[1 + index].X = scales[index];
+                    changed = true;
+                }
+            }
+
+            // Only update fragment count if there are scales after it for the vertex stage.
+            if (fragmentCount != totalCount && fragmentCount != _fragmentScaleCount)
+            {
+                _fragmentScaleCount = fragmentCount;
+                _supportBuffer.UpdateFragmentRenderScaleCount(_fragmentScaleCount);
+            }
+
+            if (changed)
+            {
+                _supportBuffer.UpdateRenderScale(_renderScale, 0, 1 + totalCount);
+            }
+        }
+
+        private void PrepareForDispatch()
+        {
+            _unit0Texture?.Bind(0);
+            _supportBuffer.Commit();
+        }
+
+        private void PreDraw(int vertexCount)
+        {
+            _vertexArray.PreDraw(vertexCount);
+            PreDraw();
+        }
+
+        private void PreDrawVbUnbounded()
+        {
+            _vertexArray.PreDrawVbUnbounded();
+            PreDraw();
+        }
+
+        private void PreDraw()
+        {
+            DrawCount++;
+
+            _unit0Texture?.Bind(0);
+            _supportBuffer.Commit();
+        }
+
+        private void PostDraw()
+        {
+            if (_tfEnabled)
+            {
+                for (int i = 0; i < Constants.MaxTransformFeedbackBuffers; i++)
+                {
+                    if (_tfbTargets[i].Handle != BufferHandle.Null)
+                    {
+                        Buffer.Copy(_tfbs[i], _tfbTargets[i].Handle, 0, _tfbTargets[i].Offset, _tfbTargets[i].Size);
+                    }
+                }
+            }
+        }
+
+        public void RestoreComponentMask(int index, bool force = true)
+        {
+            // If the bound render target is bgra, swap the red and blue masks.
+            uint redMask = _fpIsBgra[index].X == 0 ? 1u : 4u;
+            uint blueMask = _fpIsBgra[index].X == 0 ? 4u : 1u;
+
+            int shift = index * 4;
+            uint componentMask = _componentMasks & _fragmentOutputMap;
+            uint checkMask = 0xfu << shift;
+            uint componentMaskAtIndex = componentMask & checkMask;
+
+            if (!force && componentMaskAtIndex == (_currentComponentMasks & checkMask))
+            {
+                return;
+            }
+
+            componentMask >>= shift;
+            componentMask &= 0xfu;
+
+            GL.ColorMask(
+                index,
+                (componentMask & redMask) != 0,
+                (componentMask & 2u) != 0,
+                (componentMask & blueMask) != 0,
+                (componentMask & 8u) != 0);
+
+            _currentComponentMasks &= ~checkMask;
+            _currentComponentMasks |= componentMaskAtIndex;
+        }
+
+        public void RestoreClipControl()
+        {
+            GL.ClipControl(_clipOrigin, _clipDepthMode);
+        }
+
+        public void RestoreScissor0Enable()
+        {
+            if ((_scissorEnables & 1u) != 0)
+            {
+                GL.Enable(IndexedEnableCap.ScissorTest, 0);
+            }
+        }
+
+        public void RestoreRasterizerDiscard()
+        {
+            if (_rasterizerDiscard)
+            {
+                GL.Enable(EnableCap.RasterizerDiscard);
+            }
+        }
+
+        public void RestoreViewport0()
+        {
+            if (_viewportArray.Length > 0)
+            {
+                GL.ViewportArray(0, 1, _viewportArray);
+            }
+        }
+
+        public void RestoreProgram()
+        {
+            _program?.Bind();
+        }
+
+        public void RestoreImages1And2()
+        {
+            for (int i = 0; i < SavedImages; i++)
+            {
+                (TextureBase texBase, Format imageFormat) = _images[i];
+
+                if (texBase != null)
+                {
+                    SizedInternalFormat format = FormatTable.GetImageFormat(imageFormat);
+
+                    if (format != 0)
+                    {
+                        GL.BindImageTexture(i, texBase.Handle, 0, true, 0, TextureAccess.ReadWrite, format);
+                        continue;
+                    }
+                }
+
+                GL.BindImageTexture(i, 0, 0, true, 0, TextureAccess.ReadWrite, SizedInternalFormat.Rgba8);
+            }
+        }
+
+        public bool TryHostConditionalRendering(ICounterEvent value, ulong compare, bool isEqual)
+        {
+            if (value is CounterQueueEvent)
+            {
+                // Compare an event and a constant value.
+                CounterQueueEvent evt = (CounterQueueEvent)value;
+
+                // Easy host conditional rendering when the check matches what GL can do:
+                //  - Event is of type samples passed.
+                //  - Result is not a combination of multiple queries.
+                //  - Comparing against 0.
+                //  - Event has not already been flushed.
+
+                if (compare == 0 && evt.Type == QueryTarget.SamplesPassed && evt.ClearCounter)
+                {
+                    if (!value.ReserveForHostAccess())
+                    {
+                        // If the event has been flushed, then just use the values on the CPU.
+                        // The query object may already be repurposed for another draw (eg. begin + end).
+                        return false;
+                    }
+
+                    GL.BeginConditionalRender(evt.Query, isEqual ? ConditionalRenderType.QueryNoWaitInverted : ConditionalRenderType.QueryNoWait);
+                    _activeConditionalRender = evt;
+
+                    return true;
+                }
+            }
+
+            // The GPU will flush the queries to CPU and evaluate the condition there instead.
+
+            GL.Flush(); // The thread will be stalled manually flushing the counter, so flush GL commands now.
+            return false;
+        }
+
+        public bool TryHostConditionalRendering(ICounterEvent value, ICounterEvent compare, bool isEqual)
+        {
+            GL.Flush(); // The GPU thread will be stalled manually flushing the counter, so flush GL commands now.
+            return false; // We don't currently have a way to compare two counters for conditional rendering.
+        }
+
+        public void EndHostConditionalRendering()
+        {
+            GL.EndConditionalRender();
+
+            _activeConditionalRender?.ReleaseHostAccess();
+            _activeConditionalRender = null;
+        }
+
+        public void Dispose()
+        {
+            _supportBuffer?.Dispose();
+
+            for (int i = 0; i < Constants.MaxTransformFeedbackBuffers; i++)
+            {
+                if (_tfbs[i] != BufferHandle.Null)
+                {
+                    Buffer.Delete(_tfbs[i]);
+                    _tfbs[i] = BufferHandle.Null;
+                }
+            }
+
+            _activeConditionalRender?.ReleaseHostAccess();
+            _framebuffer?.Dispose();
+            _vertexArray?.Dispose();
+            _drawTexture.Dispose();
+        }
+    }
+}
diff --git a/src/Ryujinx.Graphics.OpenGL/Program.cs b/src/Ryujinx.Graphics.OpenGL/Program.cs
new file mode 100644
index 00000000..a6009108
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Program.cs
@@ -0,0 +1,177 @@
+using OpenTK.Graphics.OpenGL;
+using Ryujinx.Common.Logging;
+using Ryujinx.Graphics.GAL;
+using Ryujinx.Graphics.Shader;
+using Ryujinx.Graphics.Shader.Translation;
+using System;
+using System.Buffers.Binary;
+
+namespace Ryujinx.Graphics.OpenGL
+{
+    class Program : IProgram
+    {
+        private const int MaxShaderLogLength = 2048;
+
+        public int Handle { get; private set; }
+
+        public bool IsLinked
+        {
+            get
+            {
+                if (_status == ProgramLinkStatus.Incomplete)
+                {
+                    CheckProgramLink(true);
+                }
+
+                return _status == ProgramLinkStatus.Success;
+            }
+        }
+
+        private ProgramLinkStatus _status = ProgramLinkStatus.Incomplete;
+        private int[] _shaderHandles;
+
+        public bool HasFragmentShader;
+        public int FragmentOutputMap { get; }
+
+        public Program(ShaderSource[] shaders, int fragmentOutputMap)
+        {
+            Handle = GL.CreateProgram();
+
+            GL.ProgramParameter(Handle, ProgramParameterName.ProgramBinaryRetrievableHint, 1);
+
+            _shaderHandles = new int[shaders.Length];
+
+            for (int index = 0; index < shaders.Length; index++)
+            {
+                ShaderSource shader = shaders[index];
+
+                if (shader.Stage == ShaderStage.Fragment)
+                {
+                    HasFragmentShader = true;
+                }
+
+                int shaderHandle = GL.CreateShader(shader.Stage.Convert());
+
+                switch (shader.Language)
+                {
+                    case TargetLanguage.Glsl:
+                        GL.ShaderSource(shaderHandle, shader.Code);
+                        GL.CompileShader(shaderHandle);
+                        break;
+                    case TargetLanguage.Spirv:
+                        GL.ShaderBinary(1, ref shaderHandle, (BinaryFormat)All.ShaderBinaryFormatSpirVArb, shader.BinaryCode, shader.BinaryCode.Length);
+                        GL.SpecializeShader(shaderHandle, "main", 0, (int[])null, (int[])null);
+                        break;
+                }
+
+                GL.AttachShader(Handle, shaderHandle);
+
+                _shaderHandles[index] = shaderHandle;
+            }
+
+            GL.LinkProgram(Handle);
+
+            FragmentOutputMap = fragmentOutputMap;
+        }
+
+        public Program(ReadOnlySpan<byte> code, bool hasFragmentShader, int fragmentOutputMap)
+        {
+            Handle = GL.CreateProgram();
+
+            if (code.Length >= 4)
+            {
+                BinaryFormat binaryFormat = (BinaryFormat)BinaryPrimitives.ReadInt32LittleEndian(code.Slice(code.Length - 4, 4));
+
+                unsafe
+                {
+                    fixed (byte* ptr = code)
+                    {
+                        GL.ProgramBinary(Handle, binaryFormat, (IntPtr)ptr, code.Length - 4);
+                    }
+                }
+            }
+
+            HasFragmentShader = hasFragmentShader;
+            FragmentOutputMap = fragmentOutputMap;
+        }
+
+        public void Bind()
+        {
+            GL.UseProgram(Handle);
+        }
+
+        public ProgramLinkStatus CheckProgramLink(bool blocking)
+        {
+            if (!blocking && HwCapabilities.SupportsParallelShaderCompile)
+            {
+                GL.GetProgram(Handle, (GetProgramParameterName)ArbParallelShaderCompile.CompletionStatusArb, out int completed);
+
+                if (completed == 0)
+                {
+                    return ProgramLinkStatus.Incomplete;
+                }
+            }
+
+            GL.GetProgram(Handle, GetProgramParameterName.LinkStatus, out int status);
+            DeleteShaders();
+
+            if (status == 0)
+            {
+                _status = ProgramLinkStatus.Failure;
+
+                string log = GL.GetProgramInfoLog(Handle);
+
+                if (log.Length > MaxShaderLogLength)
+                {
+                    log = log.Substring(0, MaxShaderLogLength) + "...";
+                }
+
+                Logger.Warning?.Print(LogClass.Gpu, $"Shader linking failed: \n{log}");
+            }
+            else
+            {
+                _status = ProgramLinkStatus.Success;
+            }
+
+            return _status;
+        }
+
+        public byte[] GetBinary()
+        {
+            GL.GetProgram(Handle, (GetProgramParameterName)All.ProgramBinaryLength, out int size);
+
+            byte[] data = new byte[size + 4];
+
+            GL.GetProgramBinary(Handle, size, out _, out BinaryFormat binFormat, data);
+
+            BinaryPrimitives.WriteInt32LittleEndian(data.AsSpan(size, 4), (int)binFormat);
+
+            return data;
+        }
+
+        private void DeleteShaders()
+        {
+            if (_shaderHandles != null)
+            {
+                foreach (int shaderHandle in _shaderHandles)
+                {
+                    GL.DetachShader(Handle, shaderHandle);
+                    GL.DeleteShader(shaderHandle);
+                }
+
+                _shaderHandles = null;
+            }
+        }
+
+        public void Dispose()
+        {
+            if (Handle != 0)
+            {
+                DeleteShaders();
+                GL.DeleteProgram(Handle);
+
+                Handle = 0;
+            }
+        }
+    }
+}
diff --git a/src/Ryujinx.Graphics.OpenGL/Queries/BufferedQuery.cs b/src/Ryujinx.Graphics.OpenGL/Queries/BufferedQuery.cs
new file mode 100644
index 00000000..9d43f6c3
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Queries/BufferedQuery.cs
@@ -0,0 +1,120 @@
+using OpenTK.Graphics.OpenGL;
+using Ryujinx.Common.Logging;
+using System;
+using System.Runtime.InteropServices;
+using System.Threading;
+
+namespace Ryujinx.Graphics.OpenGL.Queries
+{
+    class BufferedQuery : IDisposable
+    {
+        private const int MaxQueryRetries = 5000;
+        private const long DefaultValue = -1;
+        private const ulong HighMask = 0xFFFFFFFF00000000;
+
+        public int Query { get; }
+
+        private int _buffer;
+        private IntPtr _bufferMap;
+        private QueryTarget _type;
+
+        public BufferedQuery(QueryTarget type)
+        {
+            _buffer = GL.GenBuffer();
+            Query = GL.GenQuery();
+            _type = type;
+
+            GL.BindBuffer(BufferTarget.QueryBuffer, _buffer);
+
+            unsafe
+            {
+                long defaultValue = DefaultValue;
+                GL.BufferStorage(BufferTarget.QueryBuffer, sizeof(long), (IntPtr)(&defaultValue), BufferStorageFlags.MapReadBit | BufferStorageFlags.MapWriteBit | BufferStorageFlags.MapPersistentBit);
+            }
+            _bufferMap = GL.MapBufferRange(BufferTarget.QueryBuffer, IntPtr.Zero, sizeof(long), BufferAccessMask.MapReadBit | BufferAccessMask.MapWriteBit | BufferAccessMask.MapPersistentBit);
+        }
+
+        public void Reset()
+        {
+            GL.EndQuery(_type);
+            GL.BeginQuery(_type, Query);
+        }
+
+        public void Begin()
+        {
+            GL.BeginQuery(_type, Query);
+        }
+
+        public unsafe void End(bool withResult)
+        {
+            GL.EndQuery(_type);
+
+            if (withResult)
+            {
+                GL.BindBuffer(BufferTarget.QueryBuffer, _buffer);
+
+                Marshal.WriteInt64(_bufferMap, -1L);
+                GL.GetQueryObject(Query, GetQueryObjectParam.QueryResult, (long*)0);
+                GL.MemoryBarrier(MemoryBarrierFlags.QueryBufferBarrierBit | MemoryBarrierFlags.ClientMappedBufferBarrierBit);
+            }
+            else
+            {
+                // Dummy result, just return 0.
+                Marshal.WriteInt64(_bufferMap, 0L);
+            }
+        }
+
+        private bool WaitingForValue(long data)
+        {
+            return data == DefaultValue ||
+                ((ulong)data & HighMask) == (unchecked((ulong)DefaultValue) & HighMask);
+        }
+
+        public bool TryGetResult(out long result)
+        {
+            result = Marshal.ReadInt64(_bufferMap);
+
+            return !WaitingForValue(result);
+        }
+
+        public long AwaitResult(AutoResetEvent wakeSignal = null)
+        {
+            long data = DefaultValue;
+
+            if (wakeSignal == null)
+            {
+                while (WaitingForValue(data))
+                {
+                    data = Marshal.ReadInt64(_bufferMap);
+                }
+            }
+            else
+            {
+                int iterations = 0;
+                while (WaitingForValue(data) && iterations++ < MaxQueryRetries)
+                {
+                    data = Marshal.ReadInt64(_bufferMap);
+                    if (WaitingForValue(data))
+                    {
+                        wakeSignal.WaitOne(1);
+                    }
+                }
+
+                if (iterations >= MaxQueryRetries)
+                {
+                    Logger.Error?.Print(LogClass.Gpu, $"Error: Query result timed out. Took more than {MaxQueryRetries} tries.");
+                }
+            }
+
+            return data;
+        }
+
+        public void Dispose()
+        {
+            GL.BindBuffer(BufferTarget.QueryBuffer, _buffer);
+            GL.UnmapBuffer(BufferTarget.QueryBuffer);
+            GL.DeleteBuffer(_buffer);
+            GL.DeleteQuery(Query);
+        }
+    }
+}
diff --git a/src/Ryujinx.Graphics.OpenGL/Queries/CounterQueue.cs b/src/Ryujinx.Graphics.OpenGL/Queries/CounterQueue.cs
new file mode 100644
index 00000000..e0aafa6f
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Queries/CounterQueue.cs
@@ -0,0 +1,229 @@
+using OpenTK.Graphics.OpenGL;
+using Ryujinx.Graphics.GAL;
+using System;
+using System.Collections.Generic;
+using System.Threading;
+
+namespace Ryujinx.Graphics.OpenGL.Queries
+{
+    class CounterQueue : IDisposable
+    {
+        private const int QueryPoolInitialSize = 100;
+
+        public CounterType Type { get; }
+        public bool Disposed { get; private set; }
+
+        private readonly Pipeline _pipeline;
+
+        private Queue<CounterQueueEvent> _events = new Queue<CounterQueueEvent>();
+        private CounterQueueEvent _current;
+
+        private ulong _accumulatedCounter;
+        private int _waiterCount;
+
+        private object _lock = new object();
+
+        private Queue<BufferedQuery> _queryPool;
+        private AutoResetEvent _queuedEvent = new AutoResetEvent(false);
+        private AutoResetEvent _wakeSignal = new AutoResetEvent(false);
+        private AutoResetEvent _eventConsumed = new AutoResetEvent(false);
+
+        private Thread _consumerThread;
+
+        internal CounterQueue(Pipeline pipeline, CounterType type)
+        {
+            Type = type;
+
+            _pipeline = pipeline;
+
+            QueryTarget glType = GetTarget(Type);
+
+            _queryPool = new Queue<BufferedQuery>(QueryPoolInitialSize);
+            for (int i = 0; i < QueryPoolInitialSize; i++)
+            {
+                _queryPool.Enqueue(new BufferedQuery(glType));
+            }
+
+            _current = new CounterQueueEvent(this, glType, 0);
+
+            _consumerThread = new Thread(EventConsumer);
+            _consumerThread.Start();
+        }
+
+        private void EventConsumer()
+        {
+            while (!Disposed)
+            {
+                CounterQueueEvent evt = null;
+                lock (_lock)
+                {
+                    if (_events.Count > 0)
+                    {
+                        evt = _events.Dequeue();
+                    }
+                }
+
+                if (evt == null)
+                {
+                    _queuedEvent.WaitOne(); // No more events to go through, wait for more.
+                }
+                else
+                {
+                    // Spin-wait rather than sleeping if there are any waiters, by passing null instead of the wake signal.
+                    evt.TryConsume(ref _accumulatedCounter, true, _waiterCount == 0 ? _wakeSignal : null);
+                }
+
+                if (_waiterCount > 0)
+                {
+                    _eventConsumed.Set();
+                }
+            }
+        }
+
+        internal BufferedQuery GetQueryObject()
+        {
+            // Creating/disposing query objects on a context we're sharing with will cause issues.
+            // So instead, make a lot of query objects on the main thread and reuse them.
+
+            lock (_lock)
+            {
+                if (_queryPool.Count > 0)
+                {
+                    BufferedQuery result = _queryPool.Dequeue();
+                    return result;
+                }
+                else
+                {
+                    return new BufferedQuery(GetTarget(Type));
+                }
+            }
+        }
+
+        internal void ReturnQueryObject(BufferedQuery query)
+        {
+            lock (_lock)
+            {
+                _queryPool.Enqueue(query);
+            }
+        }
+
+        public CounterQueueEvent QueueReport(EventHandler<ulong> resultHandler, ulong lastDrawIndex, bool hostReserved)
+        {
+            CounterQueueEvent result;
+            ulong draws = lastDrawIndex - _current.DrawIndex;
+
+            lock (_lock)
+            {
+                // A query's result only matters if more than one draw was performed during it.
+                // Otherwise, dummy it out and return 0 immediately.
+
+                if (hostReserved)
+                {
+                    // This counter event is guaranteed to be available for host conditional rendering.
+                    _current.ReserveForHostAccess();
+                }
+
+                _current.Complete(draws > 0, _pipeline.GetCounterDivisor(Type));
+                _events.Enqueue(_current);
+
+                _current.OnResult += resultHandler;
+
+                result = _current;
+
+                _current = new CounterQueueEvent(this, GetTarget(Type), lastDrawIndex);
+            }
+
+            _queuedEvent.Set();
+
+            return result;
+        }
+
+        public void QueueReset()
+        {
+            lock (_lock)
+            {
+                _current.Clear();
+            }
+        }
+
+        private static QueryTarget GetTarget(CounterType type)
+        {
+            switch (type)
+            {
+                case CounterType.SamplesPassed: return QueryTarget.SamplesPassed;
+                case CounterType.PrimitivesGenerated: return QueryTarget.PrimitivesGenerated;
+                case CounterType.TransformFeedbackPrimitivesWritten: return QueryTarget.TransformFeedbackPrimitivesWritten;
+            }
+
+            return QueryTarget.SamplesPassed;
+        }
+
+        public void Flush(bool blocking)
+        {
+            if (!blocking)
+            {
+                // Just wake the consumer thread - it will update the queries.
+                _wakeSignal.Set();
+                return;
+            }
+
+            lock (_lock)
+            {
+                // Tell the queue to process all events.
+                while (_events.Count > 0)
+                {
+                    CounterQueueEvent flush = _events.Peek();
+                    if (!flush.TryConsume(ref _accumulatedCounter, true))
+                    {
+                        return; // If not blocking, then return when we encounter an event that is not ready yet.
+                    }
+                    _events.Dequeue();
+                }
+            }
+        }
+
+        public void FlushTo(CounterQueueEvent evt)
+        {
+            // Flush the counter queue on the main thread.
+
+            Interlocked.Increment(ref _waiterCount);
+
+            _wakeSignal.Set();
+
+            while (!evt.Disposed)
+            {
+                _eventConsumed.WaitOne(1);
+            }
+
+            Interlocked.Decrement(ref _waiterCount);
+        }
+
+        public void Dispose()
+        {
+            lock (_lock)
+            {
+                while (_events.Count > 0)
+                {
+                    CounterQueueEvent evt = _events.Dequeue();
+
+                    evt.Dispose();
+                }
+
+                Disposed = true;
+            }
+
+            _queuedEvent.Set();
+
+            _consumerThread.Join();
+
+            foreach (BufferedQuery query in _queryPool)
+            {
+                query.Dispose();
+            }
+
+            _queuedEvent.Dispose();
+            _wakeSignal.Dispose();
+            _eventConsumed.Dispose();
+        }
+    }
+}
diff --git a/src/Ryujinx.Graphics.OpenGL/Queries/CounterQueueEvent.cs b/src/Ryujinx.Graphics.OpenGL/Queries/CounterQueueEvent.cs
new file mode 100644
index 00000000..7297baab
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Queries/CounterQueueEvent.cs
@@ -0,0 +1,163 @@
+using OpenTK.Graphics.OpenGL;
+using Ryujinx.Graphics.GAL;
+using System;
+using System.Threading;
+
+namespace Ryujinx.Graphics.OpenGL.Queries
+{
+    class CounterQueueEvent : ICounterEvent
+    {
+        public event EventHandler<ulong> OnResult;
+
+        public QueryTarget Type { get; }
+        public bool ClearCounter { get; private set; }
+        public int Query => _counter.Query;
+
+        public bool Disposed { get; private set; }
+        public bool Invalid { get; set; }
+
+        public ulong DrawIndex { get; }
+
+        private CounterQueue _queue;
+        private BufferedQuery _counter;
+
+        private bool _hostAccessReserved = false;
+        private int _refCount = 1; // Starts with a reference from the counter queue.
+
+        private object _lock = new object();
+        private ulong _result = ulong.MaxValue;
+        private double _divisor = 1f;
+
+        public CounterQueueEvent(CounterQueue queue, QueryTarget type, ulong drawIndex)
+        {
+            _queue = queue;
+
+            _counter = queue.GetQueryObject();
+            Type = type;
+
+            DrawIndex = drawIndex;
+
+            _counter.Begin();
+        }
+
+        internal void Clear()
+        {
+            _counter.Reset();
+            ClearCounter = true;
+        }
+
+        internal void Complete(bool withResult, double divisor)
+        {
+            _counter.End(withResult);
+
+            _divisor = divisor;
+        }
+
+        internal bool TryConsume(ref ulong result, bool block, AutoResetEvent wakeSignal = null)
+        {
+            lock (_lock)
+            {
+                if (Disposed)
+                {
+                    return true;
+                }
+
+                if (ClearCounter || Type == QueryTarget.Timestamp)
+                {
+                    result = 0;
+                }
+
+                long queryResult;
+
+                if (block)
+                {
+                    queryResult = _counter.AwaitResult(wakeSignal);
+                }
+                else
+                {
+                    if (!_counter.TryGetResult(out queryResult))
+                    {
+                        return false;
+                    }
+                }
+
+                result += _divisor == 1 ? (ulong)queryResult : (ulong)Math.Ceiling(queryResult / _divisor);
+
+                _result = result;
+
+                OnResult?.Invoke(this, result);
+
+                Dispose(); // Return the our resources to the pool.
+
+                return true;
+            }
+        }
+
+        public void Flush()
+        {
+            if (Disposed)
+            {
+                return;
+            }
+
+            // Tell the queue to process all events up to this one.
+            _queue.FlushTo(this);
+        }
+
+        public void DecrementRefCount()
+        {
+            if (Interlocked.Decrement(ref _refCount) == 0)
+            {
+                DisposeInternal();
+            }
+        }
+
+        public bool ReserveForHostAccess()
+        {
+            if (_hostAccessReserved)
+            {
+                return true;
+            }
+
+            if (IsValueAvailable())
+            {
+                return false;
+            }
+
+            if (Interlocked.Increment(ref _refCount) == 1)
+            {
+                Interlocked.Decrement(ref _refCount);
+
+                return false;
+            }
+
+            _hostAccessReserved = true;
+
+            return true;
+        }
+
+        public void ReleaseHostAccess()
+        {
+            _hostAccessReserved = false;
+
+            DecrementRefCount();
+        }
+
+        private void DisposeInternal()
+        {
+            _queue.ReturnQueryObject(_counter);
+        }
+
+        private bool IsValueAvailable()
+        {
+            return _result != ulong.MaxValue || _counter.TryGetResult(out _);
+        }
+
+        public void Dispose()
+        {
+            Disposed = true;
+
+            DecrementRefCount();
+        }
+    }
+}
diff --git a/src/Ryujinx.Graphics.OpenGL/Queries/Counters.cs b/src/Ryujinx.Graphics.OpenGL/Queries/Counters.cs
new file mode 100644
index 00000000..e6c7ab2b
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Queries/Counters.cs
@@ -0,0 +1,57 @@
+using Ryujinx.Graphics.GAL;
+using System;
+
+namespace Ryujinx.Graphics.OpenGL.Queries
+{
+    class Counters : IDisposable
+    {
+        private CounterQueue[] _counterQueues;
+
+        public Counters()
+        {
+            int count = Enum.GetNames<CounterType>().Length;
+
+            _counterQueues = new CounterQueue[count];
+        }
+
+        public void Initialize(Pipeline pipeline)
+        {
+            for (int index = 0; index < _counterQueues.Length; index++)
+            {
+                CounterType type = (CounterType)index;
+                _counterQueues[index] = new CounterQueue(pipeline, type);
+            }
+        }
+
+        public CounterQueueEvent QueueReport(CounterType type, EventHandler<ulong> resultHandler, ulong lastDrawIndex, bool hostReserved)
+        {
+            return _counterQueues[(int)type].QueueReport(resultHandler, lastDrawIndex, hostReserved);
+        }
+
+        public void QueueReset(CounterType type)
+        {
+            _counterQueues[(int)type].QueueReset();
+        }
+
+        public void Update()
+        {
+            foreach (var queue in _counterQueues)
+            {
+                queue.Flush(false);
+            }
+        }
+
+        public void Flush(CounterType type)
+        {
+            _counterQueues[(int)type].Flush(true);
+        }
+
+        public void Dispose()
+        {
+            foreach (var queue in _counterQueues)
+            {
+                queue.Dispose();
+            }
+        }
+    }
+}
\ No newline at end of file
diff --git a/src/Ryujinx.Graphics.OpenGL/ResourcePool.cs b/src/Ryujinx.Graphics.OpenGL/ResourcePool.cs
new file mode 100644
index 00000000..57231cd6
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/ResourcePool.cs
@@ -0,0 +1,122 @@
+using Ryujinx.Graphics.GAL;
+using Ryujinx.Graphics.OpenGL.Image;
+using System;
+using System.Collections.Generic;
+
+namespace Ryujinx.Graphics.OpenGL
+{
+    class DisposedTexture
+    {
+        public TextureCreateInfo Info;
+        public TextureView View;
+        public float ScaleFactor;
+        public int RemainingFrames;
+    }
+
+    /// <summary>
+    /// A structure for pooling resources that can be reused without recreation, such as textures.
+    /// </summary>
+    class ResourcePool : IDisposable
+    {
+        private const int DisposedLiveFrames = 2;
+
+        private readonly object _lock = new object();
+        private readonly Dictionary<TextureCreateInfo, List<DisposedTexture>> _textures = new Dictionary<TextureCreateInfo, List<DisposedTexture>>();
+
+        /// <summary>
+        /// Add a texture that is not being used anymore to the resource pool to be used later.
+        /// Both the texture's view and storage should be completely unused.
+        /// </summary>
+        /// <param name="view">The texture's view</param>
+        public void AddTexture(TextureView view)
+        {
+            lock (_lock)
+            {
+                List<DisposedTexture> list;
+                if (!_textures.TryGetValue(view.Info, out list))
+                {
+                    list = new List<DisposedTexture>();
+                    _textures.Add(view.Info, list);
+                }
+
+                list.Add(new DisposedTexture()
+                {
+                    Info = view.Info,
+                    View = view,
+                    ScaleFactor = view.ScaleFactor,
+                    RemainingFrames = DisposedLiveFrames
+                });
+            }
+        }
+
+        /// <summary>
+        /// Attempt to obtain a texture from the resource cache with the desired parameters.
+        /// </summary>
+        /// <param name="info">The creation info for the desired texture</param>
+        /// <param name="scaleFactor">The scale factor for the desired texture</param>
+        /// <returns>A TextureView with the description specified, or null if one was not found.</returns>
+        public TextureView GetTextureOrNull(TextureCreateInfo info, float scaleFactor)
+        {
+            lock (_lock)
+            {
+                List<DisposedTexture> list;
+                if (!_textures.TryGetValue(info, out list))
+                {
+                    return null;
+                }
+
+                foreach (DisposedTexture texture in list)
+                {
+                    if (scaleFactor == texture.ScaleFactor)
+                    {
+                        list.Remove(texture);
+                        return texture.View;
+                    }
+                }
+
+                return null;
+            }
+        }
+
+        /// <summary>
+        /// Update the pool, removing any resources that have expired.
+        /// </summary>
+        public void Tick()
+        {
+            lock (_lock)
+            {
+                foreach (List<DisposedTexture> list in _textures.Values)
+                {
+                    for (int i = 0; i < list.Count; i++)
+                    {
+                        DisposedTexture tex = list[i];
+
+                        if (--tex.RemainingFrames < 0)
+                        {
+                            tex.View.Dispose();
+                            list.RemoveAt(i--);
+                        }
+                    }
+                }
+            }
+        }
+
+        /// <summary>
+        /// Disposes the resource pool.
+        /// </summary>
+        public void Dispose()
+        {
+            lock (_lock)
+            {
+                foreach (List<DisposedTexture> list in _textures.Values)
+                {
+                    foreach (DisposedTexture texture in list)
+                    {
+                        texture.View.Dispose();
+                    }
+                }
+                _textures.Clear();
+            }
+        }
+    }
+}
diff --git a/src/Ryujinx.Graphics.OpenGL/Ryujinx.Graphics.OpenGL.csproj b/src/Ryujinx.Graphics.OpenGL/Ryujinx.Graphics.OpenGL.csproj
new file mode 100644
index 00000000..2313cc68
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Ryujinx.Graphics.OpenGL.csproj
@@ -0,0 +1,32 @@
+<Project Sdk="Microsoft.NET.Sdk">
+
+  <PropertyGroup>
+    <TargetFramework>net7.0</TargetFramework>
+    <AllowUnsafeBlocks>true</AllowUnsafeBlocks>
+  </PropertyGroup>
+
+  <ItemGroup>
+    <PackageReference Include="OpenTK.Graphics" />
+  </ItemGroup>
+
+  <ItemGroup>
+    <EmbeddedResource Include="Effects\Textures\SmaaAreaTexture.bin" />
+    <EmbeddedResource Include="Effects\Textures\SmaaSearchTexture.bin" />
+    <EmbeddedResource Include="Effects\Shaders\fsr_sharpening.glsl" />
+    <EmbeddedResource Include="Effects\Shaders\fxaa.glsl" />
+    <EmbeddedResource Include="Effects\Shaders\smaa.hlsl" />
+    <EmbeddedResource Include="Effects\Shaders\smaa_blend.glsl" />
+    <EmbeddedResource Include="Effects\Shaders\smaa_edge.glsl" />
+    <EmbeddedResource Include="Effects\Shaders\smaa_neighbour.glsl" />
+    <EmbeddedResource Include="Effects\Shaders\ffx_fsr1.h" />
+    <EmbeddedResource Include="Effects\Shaders\ffx_a.h" />
+    <EmbeddedResource Include="Effects\Shaders\fsr_scaling.glsl" />
+  </ItemGroup>
+
+  <ItemGroup>
+    <ProjectReference Include="..\Ryujinx.Common\Ryujinx.Common.csproj" />
+    <ProjectReference Include="..\Ryujinx.Graphics.GAL\Ryujinx.Graphics.GAL.csproj" />
+    <ProjectReference Include="..\Ryujinx.Graphics.Shader\Ryujinx.Graphics.Shader.csproj" />
+  </ItemGroup>
+
+</Project>
diff --git a/src/Ryujinx.Graphics.OpenGL/Sync.cs b/src/Ryujinx.Graphics.OpenGL/Sync.cs
new file mode 100644
index 00000000..58818e6a
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Sync.cs
@@ -0,0 +1,168 @@
+using OpenTK.Graphics.OpenGL;
+using Ryujinx.Common.Logging;
+using System;
+using System.Collections.Generic;
+using System.Linq;
+
+namespace Ryujinx.Graphics.OpenGL
+{
+    class Sync : IDisposable
+    {
+        private class SyncHandle
+        {
+            public ulong ID;
+            public IntPtr Handle;
+        }
+
+        private ulong _firstHandle = 0;
+        private ClientWaitSyncFlags _syncFlags => HwCapabilities.RequiresSyncFlush ? ClientWaitSyncFlags.None : ClientWaitSyncFlags.SyncFlushCommandsBit;
+
+        private List<SyncHandle> _handles = new List<SyncHandle>();
+
+        public void Create(ulong id)
+        {
+            SyncHandle handle = new SyncHandle
+            {
+                ID = id,
+                Handle = GL.FenceSync(SyncCondition.SyncGpuCommandsComplete, WaitSyncFlags.None)
+            };
+
+
+            if (HwCapabilities.RequiresSyncFlush)
+            {
+                // Force commands to flush up to the syncpoint.
+                GL.ClientWaitSync(handle.Handle, ClientWaitSyncFlags.SyncFlushCommandsBit, 0);
+            }
+
+            lock (_handles)
+            {
+                _handles.Add(handle);
+            }
+        }
+
+        public ulong GetCurrent()
+        {
+            lock (_handles)
+            {
+                ulong lastHandle = _firstHandle;
+
+                foreach (SyncHandle handle in _handles)
+                {
+                    lock (handle)
+                    {
+                        if (handle.Handle == IntPtr.Zero)
+                        {
+                            continue;
+                        }
+
+                        if (handle.ID > lastHandle)
+                        {
+                            WaitSyncStatus syncResult = GL.ClientWaitSync(handle.Handle, _syncFlags, 0);
+
+                            if (syncResult == WaitSyncStatus.AlreadySignaled)
+                            {
+                                lastHandle = handle.ID;
+                            }
+                        }
+                    }
+                }
+
+                return lastHandle;
+            }
+        }
+
+        public void Wait(ulong id)
+        {
+            SyncHandle result = null;
+
+            lock (_handles)
+            {
+                if ((long)(_firstHandle - id) > 0)
+                {
+                    return; // The handle has already been signalled or deleted.
+                }
+
+                foreach (SyncHandle handle in _handles)
+                {
+                    if (handle.ID == id)
+                    {
+                        result = handle;
+                        break;
+                    }
+                }
+            }
+
+            if (result != null)
+            {
+                lock (result)
+                {
+                    if (result.Handle == IntPtr.Zero)
+                    {
+                        return;
+                    }
+
+                    WaitSyncStatus syncResult = GL.ClientWaitSync(result.Handle, _syncFlags, 1000000000);
+                    
+                    if (syncResult == WaitSyncStatus.TimeoutExpired)
+                    {
+                        Logger.Error?.PrintMsg(LogClass.Gpu, $"GL Sync Object {result.ID} failed to signal within 1000ms. Continuing...");
+                    }
+                }
+            }
+        }
+
+        public void Cleanup()
+        {
+            // Iterate through handles and remove any that have already been signalled.
+
+            while (true)
+            {
+                SyncHandle first = null;
+                lock (_handles)
+                {
+                    first = _handles.FirstOrDefault();
+                }
+
+                if (first == null) break;
+
+                WaitSyncStatus syncResult = GL.ClientWaitSync(first.Handle, _syncFlags, 0);
+
+                if (syncResult == WaitSyncStatus.AlreadySignaled)
+                {
+                    // Delete the sync object.
+                    lock (_handles)
+                    {
+                        lock (first)
+                        {
+                            _firstHandle = first.ID + 1;
+                            _handles.RemoveAt(0);
+                            GL.DeleteSync(first.Handle);
+                            first.Handle = IntPtr.Zero;
+                        }
+                    }
+                } else
+                {
+                    // This sync handle and any following have not been reached yet.
+                    break;
+                }
+            }
+        }
+
+        public void Dispose()
+        {
+            lock (_handles)
+            {
+                foreach (SyncHandle handle in _handles)
+                {
+                    lock (handle)
+                    {
+                        GL.DeleteSync(handle.Handle);
+                        handle.Handle = IntPtr.Zero;
+                    }
+                }
+
+                _handles.Clear();
+            }
+        }
+    }
+}
diff --git a/src/Ryujinx.Graphics.OpenGL/VertexArray.cs b/src/Ryujinx.Graphics.OpenGL/VertexArray.cs
new file mode 100644
index 00000000..7d22033e
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/VertexArray.cs
@@ -0,0 +1,280 @@
+using OpenTK.Graphics.OpenGL;
+using Ryujinx.Graphics.GAL;
+using System;
+using System.Numerics;
+using System.Runtime.CompilerServices;
+
+namespace Ryujinx.Graphics.OpenGL
+{
+    class VertexArray : IDisposable
+    {
+        public int Handle { get; private set; }
+
+        private readonly VertexAttribDescriptor[] _vertexAttribs;
+        private readonly VertexBufferDescriptor[] _vertexBuffers;
+
+        private int _minVertexCount;
+
+        private uint _vertexAttribsInUse;
+        private uint _vertexBuffersInUse;
+        private uint _vertexBuffersLimited;
+
+        private BufferRange _indexBuffer;
+        private BufferHandle _tempIndexBuffer;
+        private BufferHandle _tempVertexBuffer;
+        private int _tempVertexBufferSize;
+
+        public VertexArray()
+        {
+            Handle = GL.GenVertexArray();
+
+            _vertexAttribs = new VertexAttribDescriptor[Constants.MaxVertexAttribs];
+            _vertexBuffers = new VertexBufferDescriptor[Constants.MaxVertexBuffers];
+
+            _tempIndexBuffer = Buffer.Create();
+        }
+
+        public void Bind()
+        {
+            GL.BindVertexArray(Handle);
+        }
+
+        public void SetVertexBuffers(ReadOnlySpan<VertexBufferDescriptor> vertexBuffers)
+        {
+            int minVertexCount = int.MaxValue;
+
+            int bindingIndex;
+            for (bindingIndex = 0; bindingIndex < vertexBuffers.Length; bindingIndex++)
+            {
+                VertexBufferDescriptor vb = vertexBuffers[bindingIndex];
+
+                if (vb.Buffer.Handle != BufferHandle.Null)
+                {
+                    int vertexCount = vb.Stride <= 0 ? 0 : vb.Buffer.Size / vb.Stride;
+                    if (minVertexCount > vertexCount)
+                    {
+                        minVertexCount = vertexCount;
+                    }
+
+                    GL.BindVertexBuffer(bindingIndex, vb.Buffer.Handle.ToInt32(), (IntPtr)vb.Buffer.Offset, vb.Stride);
+                    GL.VertexBindingDivisor(bindingIndex, vb.Divisor);
+                    _vertexBuffersInUse |= 1u << bindingIndex;
+                }
+                else
+                {
+                    if ((_vertexBuffersInUse & (1u << bindingIndex)) != 0)
+                    {
+                        GL.BindVertexBuffer(bindingIndex, 0, IntPtr.Zero, 0);
+                        _vertexBuffersInUse &= ~(1u << bindingIndex);
+                    }
+                }
+
+                _vertexBuffers[bindingIndex] = vb;
+            }
+
+            _minVertexCount = minVertexCount;
+        }
+
+        public void SetVertexAttributes(ReadOnlySpan<VertexAttribDescriptor> vertexAttribs)
+        {
+            int index = 0;
+
+            for (; index < vertexAttribs.Length; index++)
+            {
+                VertexAttribDescriptor attrib = vertexAttribs[index];
+
+                if (attrib.Equals(_vertexAttribs[index]))
+                {
+                    continue;
+                }
+
+                FormatInfo fmtInfo = FormatTable.GetFormatInfo(attrib.Format);
+
+                if (attrib.IsZero)
+                {
+                    // Disabling the attribute causes the shader to read a constant value.
+                    // We currently set the constant to (0, 0, 0, 0).
+                    DisableVertexAttrib(index);
+                }
+                else
+                {
+                    EnableVertexAttrib(index);
+                }
+
+                int offset = attrib.Offset;
+                int size   = fmtInfo.Components;
+
+                bool isFloat = fmtInfo.PixelType == PixelType.Float ||
+                               fmtInfo.PixelType == PixelType.HalfFloat;
+
+                if (isFloat || fmtInfo.Normalized || fmtInfo.Scaled)
+                {
+                    VertexAttribType type = (VertexAttribType)fmtInfo.PixelType;
+
+                    GL.VertexAttribFormat(index, size, type, fmtInfo.Normalized, offset);
+                }
+                else
+                {
+                    VertexAttribIntegerType type = (VertexAttribIntegerType)fmtInfo.PixelType;
+
+                    GL.VertexAttribIFormat(index, size, type, offset);
+                }
+
+                GL.VertexAttribBinding(index, attrib.BufferIndex);
+
+                _vertexAttribs[index] = attrib;
+            }
+
+            for (; index < Constants.MaxVertexAttribs; index++)
+            {
+                DisableVertexAttrib(index);
+            }
+        }
+
+        public void SetIndexBuffer(BufferRange range)
+        {
+            _indexBuffer = range;
+            GL.BindBuffer(BufferTarget.ElementArrayBuffer, range.Handle.ToInt32());
+        }
+
+        public void SetRangeOfIndexBuffer()
+        {
+            Buffer.Resize(_tempIndexBuffer, _indexBuffer.Size);
+            Buffer.Copy(_indexBuffer.Handle, _tempIndexBuffer, _indexBuffer.Offset, 0, _indexBuffer.Size);
+            GL.BindBuffer(BufferTarget.ElementArrayBuffer, _tempIndexBuffer.ToInt32());
+        }
+
+        public void RestoreIndexBuffer()
+        {
+            GL.BindBuffer(BufferTarget.ElementArrayBuffer, _indexBuffer.Handle.ToInt32());
+        }
+
+        public void PreDraw(int vertexCount)
+        {
+            LimitVertexBuffers(vertexCount);
+        }
+
+        public void PreDrawVbUnbounded()
+        {
+            UnlimitVertexBuffers();
+        }
+
+        public void LimitVertexBuffers(int vertexCount)
+        {
+            // Is it possible for the draw to fetch outside the bounds of any vertex buffer currently bound?
+
+            if (vertexCount <= _minVertexCount)
+            {
+                return;
+            }
+
+            // If the draw can fetch out of bounds, let's ensure that it will only fetch zeros rather than memory garbage.
+
+            int currentTempVbOffset = 0;
+            uint buffersInUse = _vertexBuffersInUse;
+
+            while (buffersInUse != 0)
+            {
+                int vbIndex = BitOperations.TrailingZeroCount(buffersInUse);
+
+                ref var vb = ref _vertexBuffers[vbIndex];
+
+                int requiredSize = vertexCount * vb.Stride;
+
+                if (vb.Buffer.Size < requiredSize)
+                {
+                    BufferHandle tempVertexBuffer = EnsureTempVertexBufferSize(currentTempVbOffset + requiredSize);
+
+                    Buffer.Copy(vb.Buffer.Handle, tempVertexBuffer, vb.Buffer.Offset, currentTempVbOffset, vb.Buffer.Size);
+                    Buffer.Clear(tempVertexBuffer, currentTempVbOffset + vb.Buffer.Size, requiredSize - vb.Buffer.Size, 0);
+
+                    GL.BindVertexBuffer(vbIndex, tempVertexBuffer.ToInt32(), (IntPtr)currentTempVbOffset, vb.Stride);
+
+                    currentTempVbOffset += requiredSize;
+                    _vertexBuffersLimited |= 1u << vbIndex;
+                }
+
+                buffersInUse &= ~(1u << vbIndex);
+            }
+        }
+
+        private BufferHandle EnsureTempVertexBufferSize(int size)
+        {
+            BufferHandle tempVertexBuffer = _tempVertexBuffer;
+
+            if (_tempVertexBufferSize < size)
+            {
+                _tempVertexBufferSize = size;
+
+                if (tempVertexBuffer == BufferHandle.Null)
+                {
+                    tempVertexBuffer = Buffer.Create(size);
+                    _tempVertexBuffer = tempVertexBuffer;
+                    return tempVertexBuffer;
+                }
+
+                Buffer.Resize(_tempVertexBuffer, size);
+            }
+
+            return tempVertexBuffer;
+        }
+
+        public void UnlimitVertexBuffers()
+        {
+            uint buffersLimited = _vertexBuffersLimited;
+
+            if (buffersLimited == 0)
+            {
+                return;
+            }
+
+            while (buffersLimited != 0)
+            {
+                int vbIndex = BitOperations.TrailingZeroCount(buffersLimited);
+
+                ref var vb = ref _vertexBuffers[vbIndex];
+
+                GL.BindVertexBuffer(vbIndex, vb.Buffer.Handle.ToInt32(), (IntPtr)vb.Buffer.Offset, vb.Stride);
+
+                buffersLimited &= ~(1u << vbIndex);
+            }
+
+            _vertexBuffersLimited = 0;
+        }
+
+        [MethodImpl(MethodImplOptions.AggressiveInlining)]
+        private void EnableVertexAttrib(int index)
+        {
+            uint mask = 1u << index;
+
+            if ((_vertexAttribsInUse & mask) == 0)
+            {
+                _vertexAttribsInUse |= mask;
+                GL.EnableVertexAttribArray(index);
+            }
+        }
+
+        [MethodImpl(MethodImplOptions.AggressiveInlining)]
+        private void DisableVertexAttrib(int index)
+        {
+            uint mask = 1u << index;
+
+            if ((_vertexAttribsInUse & mask) != 0)
+            {
+                _vertexAttribsInUse &= ~mask;
+                GL.DisableVertexAttribArray(index);
+                GL.VertexAttrib4(index, 0f, 0f, 0f, 1f);
+            }
+        }
+
+        public void Dispose()
+        {
+            if (Handle != 0)
+            {
+                GL.DeleteVertexArray(Handle);
+
+                Handle = 0;
+            }
+        }
+    }
+}
diff --git a/src/Ryujinx.Graphics.OpenGL/Window.cs b/src/Ryujinx.Graphics.OpenGL/Window.cs
new file mode 100644
index 00000000..b37ec375
--- /dev/null
+++ b/src/Ryujinx.Graphics.OpenGL/Window.cs
@@ -0,0 +1,420 @@
+using OpenTK.Graphics.OpenGL;
+using Ryujinx.Graphics.GAL;
+using Ryujinx.Graphics.OpenGL.Effects;
+using Ryujinx.Graphics.OpenGL.Effects.Smaa;
+using Ryujinx.Graphics.OpenGL.Image;
+using System;
+
+namespace Ryujinx.Graphics.OpenGL
+{
+    class Window : IWindow, IDisposable
+    {
+        private readonly OpenGLRenderer _renderer;
+
+        private bool _initialized;
+
+        private int _width;
+        private int _height;
+        private bool _updateSize;
+        private int _copyFramebufferHandle;
+        private IPostProcessingEffect _antiAliasing;
+        private IScalingFilter _scalingFilter;
+        private bool _isLinear;
+        private AntiAliasing _currentAntiAliasing;
+        private bool _updateEffect;
+        private ScalingFilter _currentScalingFilter;
+        private float _scalingFilterLevel;
+        private bool _updateScalingFilter;
+        private bool _isBgra;
+        private TextureView _upscaledTexture;
+
+        internal BackgroundContextWorker BackgroundContext { get; private set; }
+
+        internal bool ScreenCaptureRequested { get; set; }
+
+        public Window(OpenGLRenderer renderer)
+        {
+            _renderer = renderer;
+        }
+
+        public void Present(ITexture texture, ImageCrop crop, Action swapBuffersCallback)
+        {
+            GL.Disable(EnableCap.FramebufferSrgb);
+
+            (int oldDrawFramebufferHandle, int oldReadFramebufferHandle) = ((Pipeline)_renderer.Pipeline).GetBoundFramebuffers();
+
+            CopyTextureToFrameBufferRGB(0, GetCopyFramebufferHandleLazy(), (TextureView)texture, crop, swapBuffersCallback);
+
+            GL.BindFramebuffer(FramebufferTarget.ReadFramebuffer, oldReadFramebufferHandle);
+            GL.BindFramebuffer(FramebufferTarget.DrawFramebuffer, oldDrawFramebufferHandle);
+
+            GL.Enable(EnableCap.FramebufferSrgb);
+
+            // Restore unpack alignment to 4, as performance overlays such as RTSS may change this to load their resources.
+            GL.PixelStore(PixelStoreParameter.UnpackAlignment, 4);
+        }
+
+        public void ChangeVSyncMode(bool vsyncEnabled) { }
+
+        public void SetSize(int width, int height)
+        {
+            _width = width;
+            _height = height;
+
+            _updateSize = true;
+        }
+
+        private void CopyTextureToFrameBufferRGB(int drawFramebuffer, int readFramebuffer, TextureView view, ImageCrop crop, Action swapBuffersCallback)
+        {
+            GL.BindFramebuffer(FramebufferTarget.DrawFramebuffer, drawFramebuffer);
+            GL.BindFramebuffer(FramebufferTarget.ReadFramebuffer, readFramebuffer);
+
+            TextureView viewConverted = view.Format.IsBgr() ? _renderer.TextureCopy.BgraSwap(view) : view;
+
+            UpdateEffect();
+
+            if (_antiAliasing != null)
+            {
+                var oldView = viewConverted;
+
+                viewConverted = _antiAliasing.Run(viewConverted, _width, _height);
+
+                if (viewConverted.Format.IsBgr())
+                {
+                    var swappedView = _renderer.TextureCopy.BgraSwap(viewConverted);
+
+                    viewConverted?.Dispose();
+
+                    viewConverted = swappedView;
+                }
+
+                if (viewConverted != oldView && oldView != view)
+                {
+                    oldView.Dispose();
+                }
+            }
+            
+            GL.BindFramebuffer(FramebufferTarget.DrawFramebuffer, drawFramebuffer);
+            GL.BindFramebuffer(FramebufferTarget.ReadFramebuffer, readFramebuffer);
+
+            GL.FramebufferTexture(
+                FramebufferTarget.ReadFramebuffer,
+                FramebufferAttachment.ColorAttachment0,
+                viewConverted.Handle,
+                0);
+
+            GL.ReadBuffer(ReadBufferMode.ColorAttachment0);
+
+            GL.Disable(EnableCap.RasterizerDiscard);
+            GL.Disable(IndexedEnableCap.ScissorTest, 0);
+
+            GL.Clear(ClearBufferMask.ColorBufferBit);
+
+            int srcX0, srcX1, srcY0, srcY1;
+            float scale = viewConverted.ScaleFactor;
+
+            if (crop.Left == 0 && crop.Right == 0)
+            {
+                srcX0 = 0;
+                srcX1 = (int)(viewConverted.Width / scale);
+            }
+            else
+            {
+                srcX0 = crop.Left;
+                srcX1 = crop.Right;
+            }
+
+            if (crop.Top == 0 && crop.Bottom == 0)
+            {
+                srcY0 = 0;
+                srcY1 = (int)(viewConverted.Height / scale);
+            }
+            else
+            {
+                srcY0 = crop.Top;
+                srcY1 = crop.Bottom;
+            }
+
+            if (scale != 1f)
+            {
+                srcX0 = (int)(srcX0 * scale);
+                srcY0 = (int)(srcY0 * scale);
+                srcX1 = (int)Math.Ceiling(srcX1 * scale);
+                srcY1 = (int)Math.Ceiling(srcY1 * scale);
+            }
+
+            float ratioX = crop.IsStretched ? 1.0f : MathF.Min(1.0f, _height * crop.AspectRatioX / (_width * crop.AspectRatioY));
+            float ratioY = crop.IsStretched ? 1.0f : MathF.Min(1.0f, _width * crop.AspectRatioY / (_height * crop.AspectRatioX));
+
+            int dstWidth = (int)(_width * ratioX);
+            int dstHeight = (int)(_height * ratioY);
+
+            int dstPaddingX = (_width - dstWidth) / 2;
+            int dstPaddingY = (_height - dstHeight) / 2;
+
+            int dstX0 = crop.FlipX ? _width - dstPaddingX : dstPaddingX;
+            int dstX1 = crop.FlipX ? dstPaddingX : _width - dstPaddingX;
+
+            int dstY0 = crop.FlipY ? dstPaddingY : _height - dstPaddingY;
+            int dstY1 = crop.FlipY ? _height - dstPaddingY : dstPaddingY;
+
+            if (ScreenCaptureRequested)
+            {
+                CaptureFrame(srcX0, srcY0, srcX1, srcY1, view.Format.IsBgr(), crop.FlipX, crop.FlipY);
+
+                ScreenCaptureRequested = false;
+            }
+
+            if (_scalingFilter != null)
+            {
+                if (viewConverted.Format.IsBgr() && !_isBgra)
+                {
+                    RecreateUpscalingTexture(true);
+                }
+
+                _scalingFilter.Run(
+                    viewConverted,
+                    _upscaledTexture,
+                    _width,
+                    _height,
+                    new Extents2D(
+                        srcX0,
+                        srcY0,
+                        srcX1,
+                        srcY1),
+                    new Extents2D(
+                        dstX0,
+                        dstY0,
+                        dstX1,
+                        dstY1)
+                    );
+
+                srcX0 = dstX0;
+                srcY0 = dstY0;
+                srcX1 = dstX1;
+                srcY1 = dstY1;
+
+                GL.FramebufferTexture(
+                    FramebufferTarget.ReadFramebuffer,
+                    FramebufferAttachment.ColorAttachment0,
+                    _upscaledTexture.Handle,
+                    0);
+            }
+
+            GL.BlitFramebuffer(
+                srcX0,
+                srcY0,
+                srcX1,
+                srcY1,
+                dstX0,
+                dstY0,
+                dstX1,
+                dstY1,
+                ClearBufferMask.ColorBufferBit,
+                _isLinear ? BlitFramebufferFilter.Linear : BlitFramebufferFilter.Nearest);
+
+            // Remove Alpha channel
+            GL.ColorMask(false, false, false, true);
+            GL.ClearColor(0.0f, 0.0f, 0.0f, 1.0f);
+            GL.Clear(ClearBufferMask.ColorBufferBit);
+
+            for (int i = 0; i < Constants.MaxRenderTargets; i++)
+            {
+                ((Pipeline)_renderer.Pipeline).RestoreComponentMask(i);
+            }
+
+            // Set clip control, viewport and the framebuffer to the output to placate overlays and OBS capture.
+            GL.ClipControl(ClipOrigin.LowerLeft, ClipDepthMode.NegativeOneToOne);
+            GL.Viewport(0, 0, _width, _height);
+            GL.BindFramebuffer(FramebufferTarget.Framebuffer, drawFramebuffer);
+
+            swapBuffersCallback();
+
+            ((Pipeline)_renderer.Pipeline).RestoreClipControl();
+            ((Pipeline)_renderer.Pipeline).RestoreScissor0Enable();
+            ((Pipeline)_renderer.Pipeline).RestoreRasterizerDiscard();
+            ((Pipeline)_renderer.Pipeline).RestoreViewport0();
+
+            if (viewConverted != view)
+            {
+                viewConverted.Dispose();
+            }
+        }
+
+        private int GetCopyFramebufferHandleLazy()
+        {
+            int handle = _copyFramebufferHandle;
+
+            if (handle == 0)
+            {
+                handle = GL.GenFramebuffer();
+
+                _copyFramebufferHandle = handle;
+            }
+
+            return handle;
+        }
+
+        public void InitializeBackgroundContext(IOpenGLContext baseContext)
+        {
+            BackgroundContext = new BackgroundContextWorker(baseContext);
+            _initialized = true;
+        }
+
+        public void CaptureFrame(int x, int y, int width, int height, bool isBgra, bool flipX, bool flipY)
+        {
+            long size = Math.Abs(4 * width * height);
+            byte[] bitmap = new byte[size];
+
+            GL.ReadPixels(x, y, width, height, isBgra ? PixelFormat.Bgra : PixelFormat.Rgba, PixelType.UnsignedByte, bitmap);
+
+            _renderer.OnScreenCaptured(new ScreenCaptureImageInfo(width, height, isBgra, bitmap, flipX, flipY));
+        }
+
+        public void Dispose()
+        {
+            if (!_initialized)
+            {
+                return;
+            }
+
+            BackgroundContext.Dispose();
+
+            if (_copyFramebufferHandle != 0)
+            {
+                GL.DeleteFramebuffer(_copyFramebufferHandle);
+
+                _copyFramebufferHandle = 0;
+            }
+
+            _antiAliasing?.Dispose();
+            _scalingFilter?.Dispose();
+            _upscaledTexture?.Dispose();
+        }
+
+        public void SetAntiAliasing(AntiAliasing effect)
+        {
+            if (_currentAntiAliasing == effect && _antiAliasing != null)
+            {
+                return;
+            }
+
+            _currentAntiAliasing = effect;
+
+            _updateEffect = true;
+        }
+
+        public void SetScalingFilter(ScalingFilter type)
+        {
+            if (_currentScalingFilter == type && _antiAliasing != null)
+            {
+                return;
+            }
+
+            _currentScalingFilter = type;
+
+            _updateScalingFilter = true;
+        }
+
+        private void UpdateEffect()
+        {
+            if (_updateEffect)
+            {
+                _updateEffect = false;
+
+                switch (_currentAntiAliasing)
+                {
+                    case AntiAliasing.Fxaa:
+                        _antiAliasing?.Dispose();
+                        _antiAliasing = new FxaaPostProcessingEffect(_renderer);
+                        break;
+                    case AntiAliasing.None:
+                        _antiAliasing?.Dispose();
+                        _antiAliasing = null;
+                        break;
+                    case AntiAliasing.SmaaLow:
+                    case AntiAliasing.SmaaMedium:
+                    case AntiAliasing.SmaaHigh:
+                    case AntiAliasing.SmaaUltra:
+                        var quality = _currentAntiAliasing - AntiAliasing.SmaaLow;
+                        if (_antiAliasing is SmaaPostProcessingEffect smaa)
+                        {
+                            smaa.Quality = quality;
+                        }
+                        else
+                        {
+                            _antiAliasing?.Dispose();
+                            _antiAliasing = new SmaaPostProcessingEffect(_renderer, quality);
+                        }
+                        break;
+                }
+            }
+
+            if (_updateSize && !_updateScalingFilter)
+            {
+                RecreateUpscalingTexture();
+            }
+
+            _updateSize = false;
+
+            if (_updateScalingFilter)
+            {
+                _updateScalingFilter = false;
+
+                switch (_currentScalingFilter)
+                {
+                    case ScalingFilter.Bilinear:
+                    case ScalingFilter.Nearest:
+                        _scalingFilter?.Dispose();
+                        _scalingFilter = null;
+                        _isLinear = _currentScalingFilter == ScalingFilter.Bilinear;
+                        _upscaledTexture?.Dispose();
+                        _upscaledTexture = null;
+                        break;
+                    case ScalingFilter.Fsr:
+                        if (_scalingFilter is not FsrScalingFilter)
+                        {
+                            _scalingFilter?.Dispose();
+                            _scalingFilter = new FsrScalingFilter(_renderer, _antiAliasing);
+                        }
+                        _isLinear = false;
+                        _scalingFilter.Level = _scalingFilterLevel;
+
+                        RecreateUpscalingTexture();
+                        break;
+                }
+            }
+        }
+
+        private void RecreateUpscalingTexture(bool forceBgra = false)
+        {
+            _upscaledTexture?.Dispose();
+
+            var info = new TextureCreateInfo(
+                _width,
+                _height,
+                1,
+                1,
+                1,
+                1,
+                1,
+                1,
+                Format.R8G8B8A8Unorm,
+                DepthStencilMode.Depth,
+                Target.Texture2D,
+                forceBgra ? SwizzleComponent.Blue : SwizzleComponent.Red,
+                SwizzleComponent.Green,
+                forceBgra ? SwizzleComponent.Red : SwizzleComponent.Blue,
+                SwizzleComponent.Alpha);
+
+            _isBgra = forceBgra;
+            _upscaledTexture = _renderer.CreateTexture(info, 1) as TextureView;
+        }
+
+        public void SetScalingFilterLevel(float level)
+        {
+            _scalingFilterLevel = level;
+            _updateScalingFilter = true;
+        }
+    }
+}
\ No newline at end of file