Implement Zero-Configuration Resolution Scaling (#1365)

* Initial implementation of Render Target Scaling

Works with most games I have. No GUI option right now, it is hardcoded.

Missing handling for texelFetch operation.

* Realtime Configuration, refactoring.

* texelFetch scaling on fragment shader (WIP)

* Improve Shader-Side changes.

* Fix potential crash when no color/depth bound

* Workaround random uses of textures in compute.

This was blacklisting textures in a few games despite causing no bugs. Will eventually add full support so this doesn't break anything.

* Fix scales oscillating when changing between non-native scales.

* Scaled textures on compute, cleanup, lazier uniform update.

* Cleanup.

* Fix stupidity

* Address Thog Feedback.

* Cover most of GDK's feedback (two comments remain)

* Fix bad rename

* Move IsDepthStencil to FormatExtensions, add docs.

* Fix default config, square texture detection.

* Three final fixes:

- Nearest copy when texture is integer format.
- Texture2D -> Texture3D copy correctly blacklists the texture before trying an unscaled copy (caused driver error)
- Discount small textures.

* Remove scale threshold.

Not needed right now - we'll see if we run into problems.

* All CPU modification blacklists scale.

* Fix comment.

15 files changed, 185 insertions, 7 deletions

diff --git a/Ryujinx.Graphics.Shader/CodeGen/Glsl/CodeGenContext.cs b/Ryujinx.Graphics.Shader/CodeGen/Glsl/CodeGenContext.cs
index 6bef8e6c..91ab7ad5 100644
--- a/Ryujinx.Graphics.Shader/CodeGen/Glsl/CodeGenContext.cs
+++ b/Ryujinx.Graphics.Shader/CodeGen/Glsl/CodeGenContext.cs
@@ -1,3 +1,5 @@
+using Ryujinx.Graphics.Shader.IntermediateRepresentation;
+using Ryujinx.Graphics.Shader.StructuredIr;
 using Ryujinx.Graphics.Shader.Translation;
 using System.Collections.Generic;
 using System.Text;
@@ -75,6 +77,21 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Glsl
             AppendLine("}" + suffix);
         }
 
+        public int FindTextureDescriptorIndex(AstTextureOperation texOp)
+        {
+            AstOperand operand = texOp.GetSource(0) as AstOperand;
+            bool bindless = (texOp.Flags & TextureFlags.Bindless) > 0;
+
+            int cBufSlot = bindless ? operand.CbufSlot : 0;
+            int cBufOffset = bindless ? operand.CbufOffset : 0;
+
+            return TextureDescriptors.FindIndex(descriptor => 
+                descriptor.Type == texOp.Type && 
+                descriptor.HandleIndex == texOp.Handle && 
+                descriptor.CbufSlot == cBufSlot &&
+                descriptor.CbufOffset == cBufOffset);
+        }
+
         private void UpdateIndentation()
         {
             _indentation = GetIndentation(_level);
diff --git a/Ryujinx.Graphics.Shader/CodeGen/Glsl/Declarations.cs b/Ryujinx.Graphics.Shader/CodeGen/Glsl/Declarations.cs
index bd947ab7..f9d61928 100644
--- a/Ryujinx.Graphics.Shader/CodeGen/Glsl/Declarations.cs
+++ b/Ryujinx.Graphics.Shader/CodeGen/Glsl/Declarations.cs
@@ -137,6 +137,14 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Glsl
                 context.AppendLine();
             }
 
+            if (context.Config.Stage == ShaderStage.Fragment || context.Config.Stage == ShaderStage.Compute)
+            {
+                if (DeclareRenderScale(context))
+                {
+                    context.AppendLine();
+                }
+            }
+
             if ((info.HelperFunctionsMask & HelperFunctionsMask.MultiplyHighS32) != 0)
             {
                 AppendHelperFunction(context, "Ryujinx.Graphics.Shader/CodeGen/Glsl/HelperFunctions/MultiplyHighS32.glsl");
@@ -219,6 +227,33 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Glsl
             }
         }
 
+        private static bool DeclareRenderScale(CodeGenContext context)
+        {
+            if ((context.Config.UsedFeatures & (FeatureFlags.FragCoordXY | FeatureFlags.IntegerSampling)) != 0)
+            {
+                string stage = OperandManager.GetShaderStagePrefix(context.Config.Stage);
+
+                int scaleElements = context.TextureDescriptors.Count;
+                
+                if (context.Config.Stage == ShaderStage.Fragment)
+                {
+                    scaleElements++; // Also includes render target scale, for gl_FragCoord.
+                }
+
+                context.AppendLine($"uniform float {stage}_renderScale[{scaleElements}];");
+
+                if (context.Config.UsedFeatures.HasFlag(FeatureFlags.IntegerSampling))
+                {
+                    context.AppendLine();
+                    AppendHelperFunction(context, $"Ryujinx.Graphics.Shader/CodeGen/Glsl/HelperFunctions/TexelFetchScale_{stage}.glsl");
+                }
+
+                return true;
+            }
+
+            return false;
+        }
+
         private static void DeclareStorages(CodeGenContext context, StructuredProgramInfo info)
         {
             string sbName = OperandManager.GetShaderStagePrefix(context.Config.Stage);
diff --git a/Ryujinx.Graphics.Shader/CodeGen/Glsl/HelperFunctions/TexelFetchScale_cp.glsl b/Ryujinx.Graphics.Shader/CodeGen/Glsl/HelperFunctions/TexelFetchScale_cp.glsl
new file mode 100644
index 00000000..381566d3
--- /dev/null
+++ b/Ryujinx.Graphics.Shader/CodeGen/Glsl/HelperFunctions/TexelFetchScale_cp.glsl
@@ -0,0 +1,7 @@
+ivec2 Helper_TexelFetchScale(ivec2 inputVec, int samplerIndex) {
+    float scale = cp_renderScale[samplerIndex];
+    if (scale == 1.0) {
+        return inputVec;
+    }
+    return ivec2(vec2(inputVec) * scale);
+}
\ No newline at end of file
diff --git a/Ryujinx.Graphics.Shader/CodeGen/Glsl/HelperFunctions/TexelFetchScale_fp.glsl b/Ryujinx.Graphics.Shader/CodeGen/Glsl/HelperFunctions/TexelFetchScale_fp.glsl
new file mode 100644
index 00000000..4efaa65a
--- /dev/null
+++ b/Ryujinx.Graphics.Shader/CodeGen/Glsl/HelperFunctions/TexelFetchScale_fp.glsl
@@ -0,0 +1,11 @@
+ivec2 Helper_TexelFetchScale(ivec2 inputVec, int samplerIndex) {
+    float scale = fp_renderScale[1 + samplerIndex];
+    if (scale == 1.0) {
+        return inputVec;
+    }
+    if (scale < 0.0) { // If less than 0, try interpolate between texels by using the screen position.
+        return ivec2(vec2(inputVec) * (-scale) + mod(gl_FragCoord.xy, -scale));
+    } else {
+        return ivec2(vec2(inputVec) * scale);
+    }
+}
\ No newline at end of file
diff --git a/Ryujinx.Graphics.Shader/CodeGen/Glsl/Instructions/InstGenMemory.cs b/Ryujinx.Graphics.Shader/CodeGen/Glsl/Instructions/InstGenMemory.cs
index d05c77df..b951798d 100644
--- a/Ryujinx.Graphics.Shader/CodeGen/Glsl/Instructions/InstGenMemory.cs
+++ b/Ryujinx.Graphics.Shader/CodeGen/Glsl/Instructions/InstGenMemory.cs
@@ -390,7 +390,34 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Glsl.Instructions
                 }
             }
 
-            Append(AssemblePVector(pCount));
+            string ApplyScaling(string vector)
+            {
+                if (intCoords)
+                {
+                    int index = context.FindTextureDescriptorIndex(texOp);
+
+                    if ((context.Config.Stage == ShaderStage.Fragment || context.Config.Stage == ShaderStage.Compute) &&
+                        (texOp.Flags & TextureFlags.Bindless) == 0 &&
+                        texOp.Type != SamplerType.Indexed &&
+                        pCount == 2)
+                    {
+                        return "Helper_TexelFetchScale(" + vector + ", " + index + ")";
+                    }
+                    else
+                    {
+                        // Resolution scaling cannot be applied to this texture right now.
+                        // Flag so that we know to blacklist scaling on related textures when binding them.
+
+                        TextureDescriptor descriptor = context.TextureDescriptors[index];
+                        descriptor.Flags |= TextureUsageFlags.ResScaleUnsupported;
+                        context.TextureDescriptors[index] = descriptor;
+                    }
+                }
+
+                return vector;
+            }
+
+            Append(ApplyScaling(AssemblePVector(pCount)));
 
             string AssembleDerivativesVector(int count)
             {
diff --git a/Ryujinx.Graphics.Shader/CodeGen/Glsl/OperandManager.cs b/Ryujinx.Graphics.Shader/CodeGen/Glsl/OperandManager.cs
index 4c3f802a..971284f4 100644
--- a/Ryujinx.Graphics.Shader/CodeGen/Glsl/OperandManager.cs
+++ b/Ryujinx.Graphics.Shader/CodeGen/Glsl/OperandManager.cs
@@ -185,8 +185,8 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Glsl
                     {
                         switch (value & ~3)
                         {
-                            case AttributeConsts.PositionX: return "gl_FragCoord.x";
-                            case AttributeConsts.PositionY: return "gl_FragCoord.y";
+                            case AttributeConsts.PositionX: return "(gl_FragCoord.x / fp_renderScale[0])";
+                            case AttributeConsts.PositionY: return "(gl_FragCoord.y / fp_renderScale[0])";
                             case AttributeConsts.PositionZ: return "gl_FragCoord.z";
                             case AttributeConsts.PositionW: return "gl_FragCoord.w";
                         }
diff --git a/Ryujinx.Graphics.Shader/Instructions/InstEmitMemory.cs b/Ryujinx.Graphics.Shader/Instructions/InstEmitMemory.cs
index 2418293d..24ba9a06 100644
--- a/Ryujinx.Graphics.Shader/Instructions/InstEmitMemory.cs
+++ b/Ryujinx.Graphics.Shader/Instructions/InstEmitMemory.cs
@@ -32,6 +32,8 @@ namespace Ryujinx.Graphics.Shader.Instructions
 
                 Operand src = Attribute(op.AttributeOffset + index * 4);
 
+                context.FlagAttributeRead(src.Value);
+
                 context.Copy(Register(rd), context.LoadAttribute(src, primVertex));
             }
         }
@@ -96,6 +98,8 @@ namespace Ryujinx.Graphics.Shader.Instructions
         {
             OpCodeIpa op = (OpCodeIpa)context.CurrOp;
 
+            context.FlagAttributeRead(op.AttributeOffset);
+
             Operand res = Attribute(op.AttributeOffset);
 
             if (op.AttributeOffset >= AttributeConsts.UserAttributeBase &&
diff --git a/Ryujinx.Graphics.Shader/Instructions/InstEmitTexture.cs b/Ryujinx.Graphics.Shader/Instructions/InstEmitTexture.cs
index 304906d0..ea153b11 100644
--- a/Ryujinx.Graphics.Shader/Instructions/InstEmitTexture.cs
+++ b/Ryujinx.Graphics.Shader/Instructions/InstEmitTexture.cs
@@ -283,11 +283,15 @@ namespace Ryujinx.Graphics.Shader.Instructions
 
         public static void Tld(EmitterContext context)
         {
+            context.UsedFeatures |= FeatureFlags.IntegerSampling;
+
             EmitTextureSample(context, TextureFlags.IntCoords);
         }
 
         public static void TldB(EmitterContext context)
         {
+            context.UsedFeatures |= FeatureFlags.IntegerSampling;
+
             EmitTextureSample(context, TextureFlags.IntCoords | TextureFlags.Bindless);
         }
 
@@ -428,6 +432,8 @@ namespace Ryujinx.Graphics.Shader.Instructions
                     return;
                 }
 
+                context.UsedFeatures |= FeatureFlags.IntegerSampling;
+
                 flags = ConvertTextureFlags(tldsOp.Target) | TextureFlags.IntCoords;
 
                 if (tldsOp.Target == TexelLoadTarget.Texture1DLodZero && context.Config.GpuAccessor.QueryIsTextureBuffer(tldsOp.Immediate))
diff --git a/Ryujinx.Graphics.Shader/Ryujinx.Graphics.Shader.csproj b/Ryujinx.Graphics.Shader/Ryujinx.Graphics.Shader.csproj
index d53bdce9..b2d8a2a7 100644
--- a/Ryujinx.Graphics.Shader/Ryujinx.Graphics.Shader.csproj
+++ b/Ryujinx.Graphics.Shader/Ryujinx.Graphics.Shader.csproj
@@ -8,6 +8,8 @@
     <EmbeddedResource Include="CodeGen\Glsl\HelperFunctions\ShuffleUp.glsl" />
     <EmbeddedResource Include="CodeGen\Glsl\HelperFunctions\ShuffleXor.glsl" />
     <EmbeddedResource Include="CodeGen\Glsl\HelperFunctions\SwizzleAdd.glsl" />
+    <EmbeddedResource Include="CodeGen\Glsl\HelperFunctions\TexelFetchScale_fp.glsl" />
+    <EmbeddedResource Include="CodeGen\Glsl\HelperFunctions\TexelFetchScale_cp.glsl" />
   </ItemGroup>
 
   <ItemGroup>
diff --git a/Ryujinx.Graphics.Shader/TextureDescriptor.cs b/Ryujinx.Graphics.Shader/TextureDescriptor.cs
index fae9b58c..a9900fb8 100644
--- a/Ryujinx.Graphics.Shader/TextureDescriptor.cs
+++ b/Ryujinx.Graphics.Shader/TextureDescriptor.cs
@@ -13,6 +13,8 @@ namespace Ryujinx.Graphics.Shader
         public int CbufSlot   { get; }
         public int CbufOffset { get; }
 
+        public TextureUsageFlags Flags { get; set; }
+
         public TextureDescriptor(string name, SamplerType type, int handleIndex)
         {
             Name        = name;
@@ -23,6 +25,8 @@ namespace Ryujinx.Graphics.Shader
 
             CbufSlot   = 0;
             CbufOffset = 0;
+
+            Flags = TextureUsageFlags.None;
         }
 
         public TextureDescriptor(string name, SamplerType type, int cbufSlot, int cbufOffset)
@@ -35,6 +39,8 @@ namespace Ryujinx.Graphics.Shader
 
             CbufSlot   = cbufSlot;
             CbufOffset = cbufOffset;
+
+            Flags = TextureUsageFlags.None;
         }
     }
 }
\ No newline at end of file
diff --git a/Ryujinx.Graphics.Shader/TextureUsageFlags.cs b/Ryujinx.Graphics.Shader/TextureUsageFlags.cs
new file mode 100644
index 00000000..d9fa1a50
--- /dev/null
+++ b/Ryujinx.Graphics.Shader/TextureUsageFlags.cs
@@ -0,0 +1,16 @@
+using System;
+
+namespace Ryujinx.Graphics.Shader
+{
+    /// <summary>
+    /// Flags that indicate how a texture will be used in a shader.
+    /// </summary>
+    [Flags]
+    public enum TextureUsageFlags
+    {
+        None = 0,
+
+        // Integer sampled textures must be noted for resolution scaling.
+        ResScaleUnsupported = 1 << 0
+    }
+}
diff --git a/Ryujinx.Graphics.Shader/Translation/EmitterContext.cs b/Ryujinx.Graphics.Shader/Translation/EmitterContext.cs
index 70476dcd..39532a64 100644
--- a/Ryujinx.Graphics.Shader/Translation/EmitterContext.cs
+++ b/Ryujinx.Graphics.Shader/Translation/EmitterContext.cs
@@ -11,6 +11,8 @@ namespace Ryujinx.Graphics.Shader.Translation
         public Block  CurrBlock { get; set; }
         public OpCode CurrOp    { get; set; }
 
+        public FeatureFlags UsedFeatures { get; set; }
+
         public ShaderConfig Config { get; }
 
         private List<Operation> _operations;
@@ -40,6 +42,20 @@ namespace Ryujinx.Graphics.Shader.Translation
             _operations.Add(operation);
         }
 
+        public void FlagAttributeRead(int attribute)
+        {
+            if (Config.Stage == ShaderStage.Fragment)
+            {
+                switch (attribute)
+                {
+                    case AttributeConsts.PositionX:
+                    case AttributeConsts.PositionY:
+                        UsedFeatures |= FeatureFlags.FragCoordXY;
+                        break;
+                }
+            }
+        }
+
         public void MarkLabel(Operand label)
         {
             Add(Instruction.MarkLabel, label);
diff --git a/Ryujinx.Graphics.Shader/Translation/FeatureFlags.cs b/Ryujinx.Graphics.Shader/Translation/FeatureFlags.cs
new file mode 100644
index 00000000..9c65038a
--- /dev/null
+++ b/Ryujinx.Graphics.Shader/Translation/FeatureFlags.cs
@@ -0,0 +1,18 @@
+using System;
+
+namespace Ryujinx.Graphics.Shader.Translation
+{
+    /// <summary>
+    /// Features used by the shader that are important for the code generator to know in advance.
+    /// These typically change the declarations in the shader header.
+    /// </summary>
+    [Flags]
+    public enum FeatureFlags
+    {
+        None = 0,
+
+        // Affected by resolution scaling.
+        FragCoordXY     = 1 << 1,
+        IntegerSampling = 1 << 0
+    }
+}
diff --git a/Ryujinx.Graphics.Shader/Translation/ShaderConfig.cs b/Ryujinx.Graphics.Shader/Translation/ShaderConfig.cs
index 8b38afb9..aabd9ca6 100644
--- a/Ryujinx.Graphics.Shader/Translation/ShaderConfig.cs
+++ b/Ryujinx.Graphics.Shader/Translation/ShaderConfig.cs
@@ -22,6 +22,8 @@ namespace Ryujinx.Graphics.Shader.Translation
 
         public TranslationFlags Flags { get; }
 
+        public FeatureFlags UsedFeatures { get; set; }
+
         public ShaderConfig(IGpuAccessor gpuAccessor, TranslationFlags flags)
         {
             Stage             = ShaderStage.Compute;
@@ -34,6 +36,7 @@ namespace Ryujinx.Graphics.Shader.Translation
             OmapDepth         = false;
             GpuAccessor       = gpuAccessor;
             Flags             = flags;
+            UsedFeatures      = FeatureFlags.None;
         }
 
         public ShaderConfig(ShaderHeader header, IGpuAccessor gpuAccessor, TranslationFlags flags)
@@ -48,6 +51,7 @@ namespace Ryujinx.Graphics.Shader.Translation
             OmapDepth         = header.OmapDepth;
             GpuAccessor       = gpuAccessor;
             Flags             = flags;
+            UsedFeatures      = FeatureFlags.None;
         }
 
         public int GetDepthRegister()
diff --git a/Ryujinx.Graphics.Shader/Translation/Translator.cs b/Ryujinx.Graphics.Shader/Translation/Translator.cs
index 1a69c511..1d7aacdd 100644
--- a/Ryujinx.Graphics.Shader/Translation/Translator.cs
+++ b/Ryujinx.Graphics.Shader/Translation/Translator.cs
@@ -16,15 +16,19 @@ namespace Ryujinx.Graphics.Shader.Translation
 
         public static ShaderProgram Translate(ulong address, IGpuAccessor gpuAccessor, TranslationFlags flags)
         {
-            Operation[] ops = DecodeShader(address, gpuAccessor, flags, out ShaderConfig config, out int size);
+            Operation[] ops = DecodeShader(address, gpuAccessor, flags, out ShaderConfig config, out int size, out FeatureFlags featureFlags);
+
+            config.UsedFeatures = featureFlags;
 
             return Translate(ops, config, size);
         }
 
         public static ShaderProgram Translate(ulong addressA, ulong addressB, IGpuAccessor gpuAccessor, TranslationFlags flags)
         {
-            Operation[] opsA = DecodeShader(addressA, gpuAccessor, flags | TranslationFlags.VertexA, out _, out int sizeA);
-            Operation[] opsB = DecodeShader(addressB, gpuAccessor, flags, out ShaderConfig config, out int sizeB);
+            Operation[] opsA = DecodeShader(addressA, gpuAccessor, flags | TranslationFlags.VertexA, out _, out int sizeA, out FeatureFlags featureFlagsA);
+            Operation[] opsB = DecodeShader(addressB, gpuAccessor, flags, out ShaderConfig config, out int sizeB, out FeatureFlags featureFlagsB);
+
+            config.UsedFeatures = featureFlagsA | featureFlagsB;
 
             return Translate(Combine(opsA, opsB), config, sizeB, sizeA);
         }
@@ -67,7 +71,8 @@ namespace Ryujinx.Graphics.Shader.Translation
             IGpuAccessor     gpuAccessor,
             TranslationFlags flags,
             out ShaderConfig config,
-            out int          size)
+            out int          size,
+            out FeatureFlags featureFlags)
         {
             Block[] cfg;
 
@@ -90,6 +95,8 @@ namespace Ryujinx.Graphics.Shader.Translation
 
                 size = 0;
 
+                featureFlags = FeatureFlags.None;
+
                 return Array.Empty<Operation>();
             }
 
@@ -192,6 +199,8 @@ namespace Ryujinx.Graphics.Shader.Translation
 
             size = (int)maxEndAddress + (((flags & TranslationFlags.Compute) != 0) ? 0 : HeaderSize);
 
+            featureFlags = context.UsedFeatures;
+
             return context.GetOperations();
         }