Merge branch 'master' into fix_unnecessary_dylib

.github/workflows/build.yml

@@ -27,7 +27,7 @@ jobs:
   build:
     name: ${{ matrix.OS_NAME }} (${{ matrix.configuration }})
     runs-on: ${{ matrix.os }}
-    timeout-minutes: 35
+    timeout-minutes: 45
     strategy:
       matrix:
         os: [ubuntu-latest, macOS-latest, windows-latest]
@@ -110,7 +110,7 @@ jobs:
   build_macos:
     name: macOS Universal (${{ matrix.configuration }})
     runs-on: ubuntu-latest
-    timeout-minutes: 35
+    timeout-minutes: 45
     strategy:
       matrix:
         configuration: [ Debug, Release ]

.github/workflows/flatpak.yml

@@ -12,7 +12,7 @@ concurrency: flatpak-release
 
 jobs:
   release:
-    timeout-minutes: 35
+    timeout-minutes: ${{ fromJSON(vars.JOB_TIMEOUT) }}
     runs-on: ubuntu-latest
 
     env:

.github/workflows/nightly_pr_comment.yml

@@ -7,7 +7,7 @@ jobs:
   pr_comment:
     if: github.event.workflow_run.event == 'pull_request' && github.event.workflow_run.conclusion == 'success'
     runs-on: ubuntu-latest
-    timeout-minutes: 35
+    timeout-minutes: ${{ fromJSON(vars.JOB_TIMEOUT) }}
     steps:
       - uses: actions/github-script@v6
         with:

.github/workflows/release.yml

@@ -33,7 +33,7 @@ jobs:
         shell: bash
 
       - name: Create tag
-        uses: actions/github-script@v5
+        uses: actions/github-script@v6
         with:
           script: |
             github.rest.git.createRef({
@@ -46,7 +46,7 @@ jobs:
   release:
     name: Release ${{ matrix.OS_NAME }}
     runs-on: ${{ matrix.os }}
-    timeout-minutes: 35
+    timeout-minutes: ${{ fromJSON(vars.JOB_TIMEOUT) }}
     strategy:
       matrix:
         os: [ ubuntu-latest, windows-latest ]
@@ -144,7 +144,7 @@ jobs:
   macos_release:
     name: Release MacOS universal
     runs-on: ubuntu-latest
-    timeout-minutes: 35
+    timeout-minutes: ${{ fromJSON(vars.JOB_TIMEOUT) }}
     steps:
       - uses: actions/checkout@v3
 

Directory.Packages.props

@@ -3,11 +3,11 @@
     <ManagePackageVersionsCentrally>true</ManagePackageVersionsCentrally>
   </PropertyGroup>
   <ItemGroup>
-    <PackageVersion Include="Avalonia" Version="0.10.19" />
+    <PackageVersion Include="Avalonia" Version="0.10.21" />
-    <PackageVersion Include="Avalonia.Controls.DataGrid" Version="0.10.19" />
+    <PackageVersion Include="Avalonia.Controls.DataGrid" Version="0.10.21" />
-    <PackageVersion Include="Avalonia.Desktop" Version="0.10.19" />
+    <PackageVersion Include="Avalonia.Desktop" Version="0.10.21" />
-    <PackageVersion Include="Avalonia.Diagnostics" Version="0.10.19" />
+    <PackageVersion Include="Avalonia.Diagnostics" Version="0.10.21" />
-    <PackageVersion Include="Avalonia.Markup.Xaml.Loader" Version="0.10.19" />
+    <PackageVersion Include="Avalonia.Markup.Xaml.Loader" Version="0.10.21" />
     <PackageVersion Include="Avalonia.Svg" Version="0.10.18" />
     <PackageVersion Include="Avalonia.Svg.Skia" Version="0.10.18" />
     <PackageVersion Include="CommandLineParser" Version="2.9.1" />

distribution/linux/Ryujinx.sh

@@ -11,4 +11,10 @@ if [ -f "$SCRIPT_DIR/Ryujinx.Headless.SDL2" ]; then
     RYUJINX_BIN="Ryujinx.Headless.SDL2"
 fi
 
-env DOTNET_EnableAlternateStackCheck=1 "$SCRIPT_DIR/$RYUJINX_BIN" "$@"
+COMMAND="env DOTNET_EnableAlternateStackCheck=1"
+
+if command -v gamemoderun > /dev/null 2>&1; then
+    COMMAND="$COMMAND gamemoderun"
+fi
+
+$COMMAND "$SCRIPT_DIR/$RYUJINX_BIN" "$@"

distribution/macos/updater.sh

@@ -25,14 +25,27 @@ error_handler() {
     exit 1
 }
 
-# Wait for Ryujinx to exit
-# NOTE: in case no fds are open, lsof could be returning with a process still living.
-# We wait 1s and assume the process stopped after that
-lsof -p $APP_PID +r 1 &>/dev/null
-sleep 1
-
 trap 'error_handler ${LINENO}' ERR
 
+# Wait for Ryujinx to exit.
+# If the main process is still acitve, we wait for 1 second and check it again.
+# After the fifth time checking, this script exits with status 1.
+
+attempt=0
+while true; do
+    if lsof -p $APP_PID +r 1 &>/dev/null || ps -p "$APP_PID" &>/dev/null; then
+        if [ "$attempt" -eq 4 ]; then
+            exit 1
+        fi
+        sleep 1
+    else
+        break
+    fi
+    (( attempt++ ))
+done
+
+sleep 1
+
 # Now replace and reopen.
 rm -rf "$INSTALL_DIRECTORY"
 mv "$NEW_APP_DIRECTORY" "$INSTALL_DIRECTORY"

src/Ryujinx.Audio.Backends.SDL2/SDL2HardwareDeviceDriver.cs

@@ -45,7 +45,7 @@ namespace Ryujinx.Audio.Backends.SDL2
             }
             else
             {
-                _supportSurroundConfiguration = spec.channels == 6;
+                _supportSurroundConfiguration = spec.channels >= 6;
             }
         }
 

src/Ryujinx.Audio/Renderer/Server/AudioRenderSystem.cs

@@ -31,7 +31,6 @@ namespace Ryujinx.Audio.Renderer.Server
         private AudioRendererRenderingDevice _renderingDevice;
         private AudioRendererExecutionMode _executionMode;
         private IWritableEvent _systemEvent;
-        private ManualResetEvent _terminationEvent;
         private MemoryPoolState _dspMemoryPoolState;
         private VoiceContext _voiceContext;
         private MixContext _mixContext;
@@ -83,7 +82,6 @@ namespace Ryujinx.Audio.Renderer.Server
         public AudioRenderSystem(AudioRendererManager manager, IWritableEvent systemEvent)
         {
             _manager = manager;
-            _terminationEvent = new ManualResetEvent(false);
             _dspMemoryPoolState = MemoryPoolState.Create(MemoryPoolState.LocationType.Dsp);
             _voiceContext = new VoiceContext();
             _mixContext = new MixContext();
@@ -387,11 +385,6 @@ namespace Ryujinx.Audio.Renderer.Server
                 _isActive = false;
             }
 
-            if (_executionMode == AudioRendererExecutionMode.Auto)
-            {
-                _terminationEvent.WaitOne();
-            }
-
             Logger.Info?.Print(LogClass.AudioRenderer, $"Stopped renderer id {_sessionId}");
         }
 
@@ -668,8 +661,6 @@ namespace Ryujinx.Audio.Renderer.Server
             {
                 if (_isActive)
                 {
-                    _terminationEvent.Reset();
-
                     if (!_manager.Processor.HasRemainingCommands(_sessionId))
                     {
                         GenerateCommandList(out CommandList commands);
@@ -686,10 +677,6 @@ namespace Ryujinx.Audio.Renderer.Server
                         _isDspRunningBehind = true;
                     }
                 }
-                else
-                {
-                    _terminationEvent.Set();
-                }
             }
         }
 
@@ -857,7 +844,6 @@ namespace Ryujinx.Audio.Renderer.Server
                 }
 
                 _manager.Unregister(this);
-                _terminationEvent.Dispose();
                 _workBufferMemoryPin.Dispose();
 
                 if (MemoryManager is IRefCounted rc)

src/Ryujinx.Ava/AppHost.cs

@@ -270,7 +270,7 @@ namespace Ryujinx.Ava
 
                         string directory = AppDataManager.Mode switch
                         {
-                            AppDataManager.LaunchMode.Portable => Path.Combine(AppDataManager.BaseDirPath, "screenshots"),
+                            AppDataManager.LaunchMode.Portable or AppDataManager.LaunchMode.Custom => Path.Combine(AppDataManager.BaseDirPath, "screenshots"),
                             _                                  => Path.Combine(Environment.GetFolderPath(Environment.SpecialFolder.MyPictures), "Ryujinx")
                         };
 

src/Ryujinx.Ava/Assets/Locales/en_US.json

@@ -637,5 +637,7 @@
   "SettingsTabNetworkInterface": "Network Interface:",
   "NetworkInterfaceTooltip": "The network interface used for LAN features",
   "NetworkInterfaceDefault": "Default",
-  "PackagingShaders": "Packaging Shaders"
+  "PackagingShaders": "Packaging Shaders",
+  "AboutChangelogButton": "View Changelog on GitHub",
+  "AboutChangelogButtonTooltipMessage": "Click to open the changelog for this version in your default browser."
 }

src/Ryujinx.Ava/UI/Helpers/NotificationHelper.cs

@@ -3,21 +3,20 @@ using Avalonia.Controls;
 using Avalonia.Controls.Notifications;
 using Avalonia.Threading;
 using Ryujinx.Ava.Common.Locale;
+using Ryujinx.Common;
 using System;
 using System.Collections.Concurrent;
 using System.Threading;
-using System.Threading.Tasks;
 
 namespace Ryujinx.Ava.UI.Helpers
 {
     public static class NotificationHelper
     {
-        private const int MaxNotifications      = 4; 
+        private const int MaxNotifications      = 4;
         private const int NotificationDelayInMs = 5000;
 
         private static WindowNotificationManager _notificationManager;
 
-        private static readonly ManualResetEvent                 _templateAppliedEvent = new(false);
         private static readonly BlockingCollection<Notification> _notifications        = new();
 
         public static void SetNotificationManager(Window host)
@@ -29,25 +28,31 @@ namespace Ryujinx.Ava.UI.Helpers
                 Margin   = new Thickness(0, 0, 15, 40)
             };
 
+            var maybeAsyncWorkQueue = new Lazy<AsyncWorkQueue<Notification>>(
+                () => new AsyncWorkQueue<Notification>(notification =>
+                    {
+                        Dispatcher.UIThread.Post(() =>
+                        {
+                            _notificationManager.Show(notification);
+                        });
+                    },
+                    "UI.NotificationThread",
+                    _notifications),
+                LazyThreadSafetyMode.ExecutionAndPublication);
+
             _notificationManager.TemplateApplied += (sender, args) =>
             {
-                _templateAppliedEvent.Set();
+                // NOTE: Force creation of the AsyncWorkQueue.
+                _ = maybeAsyncWorkQueue.Value;
             };
 
-            Task.Run(async () =>
+            host.Closing += (sender, args) =>
             {
-                _templateAppliedEvent.WaitOne();
+                if (maybeAsyncWorkQueue.IsValueCreated)
-
-                foreach (var notification in _notifications.GetConsumingEnumerable())
                 {
-                    Dispatcher.UIThread.Post(() =>
+                    maybeAsyncWorkQueue.Value.Dispose();
-                    {
-                        _notificationManager.Show(notification);
-                    });
-
-                    await Task.Delay(NotificationDelayInMs / MaxNotifications);
                 }
-            });
+            };
         }
 
         public static void Show(string title, string text, NotificationType type, bool waitingExit = false, Action onClick = null, Action onClose = null)

src/Ryujinx.Ava/UI/ViewModels/MainWindowViewModel.cs

@@ -1529,6 +1529,8 @@ namespace Ryujinx.Ava.UI.ViewModels
                     double sessionTimePlayed = DateTime.UtcNow.Subtract(appMetadata.LastPlayed.Value).TotalSeconds;
                     appMetadata.TimePlayed += Math.Round(sessionTimePlayed, MidpointRounding.AwayFromZero);
                 }
+
+                appMetadata.LastPlayed = DateTime.UtcNow;
             });
         }
 

src/Ryujinx.Ava/UI/Windows/AboutWindow.axaml

@@ -72,6 +72,18 @@
                     LineHeight="12"
                     Text="{Binding Version}"
                     TextAlignment="Center" />
+                <Button
+                    Padding="5"
+                    HorizontalAlignment="Center"
+                    Background="Transparent"
+                    Click="Button_OnClick"
+                    Tag="https://github.com/Ryujinx/Ryujinx/wiki/Changelog#ryujinx-changelog">
+                    <TextBlock
+                        FontSize="10"
+                        Text="{locale:Locale AboutChangelogButton}"
+                        TextAlignment="Center"
+                        ToolTip.Tip="{locale:Locale AboutChangelogButtonTooltipMessage}" />
+                </Button>
             </StackPanel>
             <StackPanel
                 Grid.Row="2"

src/Ryujinx.Common/AsyncWorkQueue.cs

@@ -22,9 +22,11 @@ namespace Ryujinx.Common
             _cts = new CancellationTokenSource();
             _queue = collection;
             _workerAction = callback;
-            _workerThread = new Thread(DoWork) { Name = name };
+            _workerThread = new Thread(DoWork)
-
+            {
-            _workerThread.IsBackground = true;
+                Name = name,
+                IsBackground = true
+            };
             _workerThread.Start();
         }
 

src/Ryujinx.Graphics.GAL/Capabilities.cs

@@ -39,6 +39,7 @@ namespace Ryujinx.Graphics.GAL
         public readonly bool SupportsViewportMask;
         public readonly bool SupportsViewportSwizzle;
         public readonly bool SupportsIndirectParameters;
+        public readonly bool SupportsDepthClipControl;
 
         public readonly uint MaximumUniformBuffersPerStage;
         public readonly uint MaximumStorageBuffersPerStage;
@@ -85,6 +86,7 @@ namespace Ryujinx.Graphics.GAL
             bool supportsViewportMask,
             bool supportsViewportSwizzle,
             bool supportsIndirectParameters,
+            bool supportsDepthClipControl,
             uint maximumUniformBuffersPerStage,
             uint maximumStorageBuffersPerStage,
             uint maximumTexturesPerStage,
@@ -127,6 +129,7 @@ namespace Ryujinx.Graphics.GAL
             SupportsViewportMask = supportsViewportMask;
             SupportsViewportSwizzle = supportsViewportSwizzle;
             SupportsIndirectParameters = supportsIndirectParameters;
+            SupportsDepthClipControl = supportsDepthClipControl;
             MaximumUniformBuffersPerStage = maximumUniformBuffersPerStage;
             MaximumStorageBuffersPerStage = maximumStorageBuffersPerStage;
             MaximumTexturesPerStage = maximumTexturesPerStage;

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

@@ -22,7 +22,7 @@ namespace Ryujinx.Graphics.Gpu.Shader.DiskCache
         private const ushort FileFormatVersionMajor = 1;
         private const ushort FileFormatVersionMinor = 2;
         private const uint FileFormatVersionPacked = ((uint)FileFormatVersionMajor << 16) | FileFormatVersionMinor;
-        private const uint CodeGenVersion = 4646;
+        private const uint CodeGenVersion = 5027;
 
         private const string SharedTocFileName = "shared.toc";
         private const string SharedDataFileName = "shared.data";

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

@@ -165,6 +165,8 @@ namespace Ryujinx.Graphics.Gpu.Shader
 
         public bool QueryHostSupportsViewportMask() => _context.Capabilities.SupportsViewportMask;
 
+        public bool QueryHostSupportsDepthClipControl() => _context.Capabilities.SupportsDepthClipControl;
+
         /// <summary>
         /// Converts a packed Maxwell texture format to the shader translator texture format.
         /// </summary>

src/Ryujinx.Graphics.OpenGL/OpenGLRenderer.cs

@@ -163,6 +163,7 @@ namespace Ryujinx.Graphics.OpenGL
                 supportsViewportMask: HwCapabilities.SupportsViewportArray2,
                 supportsViewportSwizzle: HwCapabilities.SupportsViewportSwizzle,
                 supportsIndirectParameters: HwCapabilities.SupportsIndirectParameters,
+                supportsDepthClipControl: true,
                 maximumUniformBuffersPerStage: 13, // TODO: Avoid hardcoding those limits here and get from driver?
                 maximumStorageBuffersPerStage: 16,
                 maximumTexturesPerStage: 32,

src/Ryujinx.Graphics.Shader/CodeGen/Glsl/Declarations.cs

@@ -239,33 +239,10 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Glsl
                 context.AppendLine();
             }
 
-            bool isFragment = context.Config.Stage == ShaderStage.Fragment;
+            if (context.Config.Stage == ShaderStage.Fragment && context.Config.GpuAccessor.QueryEarlyZForce())
-
-            if (isFragment || context.Config.Stage == ShaderStage.Compute || context.Config.Stage == ShaderStage.Vertex)
             {
-                if (isFragment && context.Config.GpuAccessor.QueryEarlyZForce())
+                context.AppendLine("layout(early_fragment_tests) in;");
-                {
+                context.AppendLine();
-                    context.AppendLine("layout(early_fragment_tests) in;");
-                    context.AppendLine();
-                }
-
-                if ((context.Config.UsedFeatures & (FeatureFlags.FragCoordXY | FeatureFlags.IntegerSampling)) != 0)
-                {
-                    string stage = OperandManager.GetShaderStagePrefix(context.Config.Stage);
-
-                    int scaleElements = context.Config.GetTextureDescriptors().Length + context.Config.GetImageDescriptors().Length;
-
-                    if (isFragment)
-                    {
-                        scaleElements++; // Also includes render target scale, for gl_FragCoord.
-                    }
-
-                    if (context.Config.UsedFeatures.HasFlag(FeatureFlags.IntegerSampling) && scaleElements != 0)
-                    {
-                        AppendHelperFunction(context, $"Ryujinx.Graphics.Shader/CodeGen/Glsl/HelperFunctions/TexelFetchScale_{stage}.glsl");
-                        context.AppendLine();
-                    }
-                }
             }
 
             if ((info.HelperFunctionsMask & HelperFunctionsMask.AtomicMinMaxS32Shared) != 0)

src/Ryujinx.Graphics.Shader/CodeGen/Glsl/HelperFunctions/TexelFetchScale_cp.glsl

@@ -1,19 +0,0 @@
-ivec2 Helper_TexelFetchScale(ivec2 inputVec, int samplerIndex)
-{
-    float scale = support_buffer.s_render_scale[1 + samplerIndex];
-    if (scale == 1.0)
-    {
-        return inputVec;
-    }
-    return ivec2(vec2(inputVec) * scale);
-}
-
-int Helper_TextureSizeUnscale(int size, int samplerIndex)
-{
-    float scale = support_buffer.s_render_scale[1 + samplerIndex];
-    if (scale == 1.0)
-    {
-        return size;
-    }
-    return int(float(size) / scale);
-}

src/Ryujinx.Graphics.Shader/CodeGen/Glsl/HelperFunctions/TexelFetchScale_fp.glsl

@@ -1,26 +0,0 @@
-ivec2 Helper_TexelFetchScale(ivec2 inputVec, int samplerIndex)
-{
-    float scale = support_buffer.s_render_scale[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, 0.0 - scale));
-    }
-    else
-    {
-        return ivec2(vec2(inputVec) * scale);
-    }
-}
-
-int Helper_TextureSizeUnscale(int size, int samplerIndex)
-{
-    float scale = abs(support_buffer.s_render_scale[1 + samplerIndex]);
-    if (scale == 1.0)
-    {
-        return size;
-    }
-    return int(float(size) / scale);
-}

src/Ryujinx.Graphics.Shader/CodeGen/Glsl/HelperFunctions/TexelFetchScale_vp.glsl

@@ -1,20 +0,0 @@
-ivec2 Helper_TexelFetchScale(ivec2 inputVec, int samplerIndex)
-{
-    float scale = abs(support_buffer.s_render_scale[1 + samplerIndex + support_buffer.s_frag_scale_count]);
-    if (scale == 1.0)
-    {
-        return inputVec;
-    }
-
-    return ivec2(vec2(inputVec) * scale);
-}
-
-int Helper_TextureSizeUnscale(int size, int samplerIndex)
-{
-    float scale = abs(support_buffer.s_render_scale[1 + samplerIndex + support_buffer.s_frag_scale_count]);
-    if (scale == 1.0)
-    {
-        return size;
-    }
-    return int(float(size) / scale);
-}

src/Ryujinx.Graphics.Shader/CodeGen/Glsl/Instructions/InstGenHelper.cs

@@ -101,6 +101,7 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Glsl.Instructions
             Add(Instruction.MemoryBarrier,            InstType.CallNullary,    "memoryBarrier");
             Add(Instruction.Minimum,                  InstType.CallBinary,     "min");
             Add(Instruction.MinimumU32,               InstType.CallBinary,     "min");
+            Add(Instruction.Modulo,                   InstType.CallBinary,     "mod");
             Add(Instruction.Multiply,                 InstType.OpBinaryCom,    "*",               1);
             Add(Instruction.MultiplyHighS32,          InstType.CallBinary,     HelperFunctionNames.MultiplyHighS32);
             Add(Instruction.MultiplyHighU32,          InstType.CallBinary,     HelperFunctionNames.MultiplyHighU32);

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

@@ -97,30 +97,6 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Glsl.Instructions
                 texCallBuilder.Append(str);
             }
 
-            string ApplyScaling(string vector)
-            {
-                if (context.Config.Stage.SupportsRenderScale() &&
-                    texOp.Inst == Instruction.ImageLoad &&
-                    !isBindless &&
-                    !isIndexed)
-                {
-                    // Image scales start after texture ones.
-                    int scaleIndex = context.Config.GetTextureDescriptors().Length + context.Config.FindImageDescriptorIndex(texOp);
-
-                    if (pCount == 3 && isArray)
-                    {
-                        // The array index is not scaled, just x and y.
-                        vector = $"ivec3(Helper_TexelFetchScale(({vector}).xy, {scaleIndex}), ({vector}).z)";
-                    }
-                    else if (pCount == 2 && !isArray)
-                    {
-                        vector = $"Helper_TexelFetchScale({vector}, {scaleIndex})";
-                    }
-                }
-
-                return vector;
-            }
-
             if (pCount > 1)
             {
                 string[] elems = new string[pCount];
@@ -130,7 +106,7 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Glsl.Instructions
                     elems[index] = Src(AggregateType.S32);
                 }
 
-                Append(ApplyScaling($"ivec{pCount}({string.Join(", ", elems)})"));
+                Append($"ivec{pCount}({string.Join(", ", elems)})");
             }
             else
             {
@@ -584,53 +560,7 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Glsl.Instructions
                 }
             }
 
-            string ApplyScaling(string vector)
+            Append(AssemblePVector(pCount));
-            {
-                if (intCoords)
-                {
-                    if (context.Config.Stage.SupportsRenderScale() &&
-                        !isBindless &&
-                        !isIndexed)
-                    {
-                        int index = context.Config.FindTextureDescriptorIndex(texOp);
-
-                        if (pCount == 3 && isArray)
-                        {
-                            // The array index is not scaled, just x and y.
-                            vector = "ivec3(Helper_TexelFetchScale((" + vector + ").xy, " + index + "), (" + vector + ").z)";
-                        }
-                        else if (pCount == 2 && !isArray)
-                        {
-                            vector = "Helper_TexelFetchScale(" + vector + ", " + index + ")";
-                        }
-                    }
-                }
-
-                return vector;
-            }
-
-            string ApplyBias(string vector)
-            {
-                int gatherBiasPrecision = context.Config.GpuAccessor.QueryHostGatherBiasPrecision();
-                if (isGather && gatherBiasPrecision != 0)
-                {
-                    // GPU requires texture gather to be slightly offset to match NVIDIA behaviour when point is exactly between two texels.
-                    // Offset by the gather precision divided by 2 to correct for rounding.
-
-                    if (pCount == 1)
-                    {
-                        vector = $"{vector} + (1.0 / (float(textureSize({samplerName}, 0)) * float({1 << (gatherBiasPrecision + 1)})))";
-                    }
-                    else
-                    {
-                        vector = $"{vector} + (1.0 / (vec{pCount}(textureSize({samplerName}, 0).{"xyz".Substring(0, pCount)}) * float({1 << (gatherBiasPrecision + 1)})))";
-                    }
-                }
-
-                return vector;
-            }
-
-            Append(ApplyBias(ApplyScaling(AssemblePVector(pCount))));
 
             string AssembleDerivativesVector(int count)
             {
@@ -750,7 +680,7 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Glsl.Instructions
             }
             else
             {
-                (TextureDescriptor descriptor, int descriptorIndex) = context.Config.FindTextureDescriptor(texOp);
+                TextureDescriptor descriptor = context.Config.FindTextureDescriptor(texOp);
                 bool hasLod = !descriptor.Type.HasFlag(SamplerType.Multisample) && descriptor.Type != SamplerType.TextureBuffer;
                 string texCall;
 
@@ -767,14 +697,6 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Glsl.Instructions
                     texCall = $"textureSize({samplerName}){GetMask(texOp.Index)}";
                 }
 
-                if (context.Config.Stage.SupportsRenderScale() &&
-                    (texOp.Index < 2 || (texOp.Type & SamplerType.Mask) == SamplerType.Texture3D) &&
-                    !isBindless &&
-                    !isIndexed)
-                {
-                    texCall = $"Helper_TextureSizeUnscale({texCall}, {descriptorIndex})";
-                }
-
                 return texCall;
             }
         }

src/Ryujinx.Graphics.Shader/CodeGen/Spirv/CodeGenContext.cs

@@ -36,6 +36,7 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Spirv
         public Dictionary<IoDefinition, Instruction> OutputsPerPatch { get; } = new Dictionary<IoDefinition, Instruction>();
 
         public Instruction CoordTemp { get; set; }
+        public StructuredFunction CurrentFunction { get; set; }
         private readonly Dictionary<AstOperand, Instruction> _locals = new Dictionary<AstOperand, Instruction>();
         private readonly Dictionary<int, Instruction[]> _localForArgs = new Dictionary<int, Instruction[]>();
         private readonly Dictionary<int, Instruction> _funcArgs = new Dictionary<int, Instruction>();

src/Ryujinx.Graphics.Shader/CodeGen/Spirv/Instructions.cs

@@ -4,7 +4,6 @@ using Ryujinx.Graphics.Shader.Translation;
 using System;
 using System.Collections.Generic;
 using System.Diagnostics;
-using System.Linq;
 using System.Numerics;
 using static Spv.Specification;
 
@@ -114,6 +113,7 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Spirv
             Add(Instruction.MemoryBarrier,            GenerateMemoryBarrier);
             Add(Instruction.Minimum,                  GenerateMinimum);
             Add(Instruction.MinimumU32,               GenerateMinimumU32);
+            Add(Instruction.Modulo,                   GenerateModulo);
             Add(Instruction.Multiply,                 GenerateMultiply);
             Add(Instruction.MultiplyHighS32,          GenerateMultiplyHighS32);
             Add(Instruction.MultiplyHighU32,          GenerateMultiplyHighU32);
@@ -744,8 +744,6 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Spirv
                 pCoords = Src(AggregateType.S32);
             }
 
-            pCoords = ScalingHelpers.ApplyScaling(context, texOp, pCoords, intCoords: true, isBindless, isIndexed, isArray, pCount);
-
             (var imageType, var imageVariable) = context.Images[new TextureMeta(texOp.CbufSlot, texOp.Handle, texOp.Format)];
 
             var image = context.Load(imageType, imageVariable);
@@ -1040,6 +1038,11 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Spirv
             return GenerateBinaryU32(context, operation, context.Delegates.GlslUMin);
         }
 
+        private static OperationResult GenerateModulo(CodeGenContext context, AstOperation operation)
+        {
+            return GenerateBinary(context, operation, context.Delegates.FMod, null);
+        }
+
         private static OperationResult GenerateMultiply(CodeGenContext context, AstOperation operation)
         {
             return GenerateBinary(context, operation, context.Delegates.FMul, context.Delegates.IMul);
@@ -1101,7 +1104,15 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Spirv
 
         private static OperationResult GenerateReturn(CodeGenContext context, AstOperation operation)
         {
-            context.Return();
+            if (operation.SourcesCount != 0)
+            {
+                context.ReturnValue(context.Get(context.CurrentFunction.ReturnType, operation.GetSource(0)));
+            }
+            else
+            {
+                context.Return();
+            }
+
             return OperationResult.Invalid;
         }
 
@@ -1439,35 +1450,7 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Spirv
                 }
             }
 
-            SpvInstruction ApplyBias(SpvInstruction vector, SpvInstruction image)
-            {
-                int gatherBiasPrecision = context.Config.GpuAccessor.QueryHostGatherBiasPrecision();
-                if (isGather && gatherBiasPrecision != 0)
-                {
-                    // GPU requires texture gather to be slightly offset to match NVIDIA behaviour when point is exactly between two texels.
-                    // Offset by the gather precision divided by 2 to correct for rounding.
-                    var sizeType = pCount == 1 ? context.TypeS32() : context.TypeVector(context.TypeS32(), pCount);
-                    var pVectorType = pCount == 1 ? context.TypeFP32() : context.TypeVector(context.TypeFP32(), pCount);
-
-                    var bias = context.Constant(context.TypeFP32(), (float)(1 << (gatherBiasPrecision + 1)));
-                    var biasVector = context.CompositeConstruct(pVectorType, Enumerable.Repeat(bias, pCount).ToArray());
-
-                    var one = context.Constant(context.TypeFP32(), 1f);
-                    var oneVector = context.CompositeConstruct(pVectorType, Enumerable.Repeat(one, pCount).ToArray());
-
-                    var divisor = context.FMul(
-                        pVectorType,
-                        context.ConvertSToF(pVectorType, context.ImageQuerySize(sizeType, image)),
-                        biasVector);
-
-                    vector = context.FAdd(pVectorType, vector, context.FDiv(pVectorType, oneVector, divisor));
-                }
-
-                return vector;
-            }
-
             SpvInstruction pCoords = AssemblePVector(pCount);
-            pCoords = ScalingHelpers.ApplyScaling(context, texOp, pCoords, intCoords, isBindless, isIndexed, isArray, pCount);
 
             SpvInstruction AssembleDerivativesVector(int count)
             {
@@ -1638,8 +1621,6 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Spirv
                 image = context.Image(imageType, image);
             }
 
-            pCoords = ApplyBias(pCoords, image);
-
             var operands = operandsList.ToArray();
 
             SpvInstruction result;
@@ -1755,11 +1736,6 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Spirv
                     result = context.CompositeExtract(context.TypeS32(), result, (SpvLiteralInteger)texOp.Index);
                 }
 
-                if (texOp.Index < 2 || (type & SamplerType.Mask) == SamplerType.Texture3D)
-                {
-                    result = ScalingHelpers.ApplyUnscaling(context, texOp.WithType(type), result, isBindless, isIndexed);
-                }
-
                 return new OperationResult(AggregateType.S32, result);
             }
         }
@@ -2269,7 +2245,7 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Spirv
             {
                 var result = emitF(context.TypeFP64(), context.GetFP64(src1), context.GetFP64(src2));
 
-                if (!context.Config.GpuAccessor.QueryHostReducedPrecision())
+                if (!context.Config.GpuAccessor.QueryHostReducedPrecision() || operation.ForcePrecise)
                 {
                     context.Decorate(result, Decoration.NoContraction);
                 }
@@ -2280,7 +2256,7 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Spirv
             {
                 var result = emitF(context.TypeFP32(), context.GetFP32(src1), context.GetFP32(src2));
 
-                if (!context.Config.GpuAccessor.QueryHostReducedPrecision())
+                if (!context.Config.GpuAccessor.QueryHostReducedPrecision() || operation.ForcePrecise)
                 {
                     context.Decorate(result, Decoration.NoContraction);
                 }
@@ -2340,7 +2316,7 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Spirv
             {
                 var result = emitF(context.TypeFP64(), context.GetFP64(src1), context.GetFP64(src2), context.GetFP64(src3));
 
-                if (!context.Config.GpuAccessor.QueryHostReducedPrecision())
+                if (!context.Config.GpuAccessor.QueryHostReducedPrecision() || operation.ForcePrecise)
                 {
                     context.Decorate(result, Decoration.NoContraction);
                 }
@@ -2351,7 +2327,7 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Spirv
             {
                 var result = emitF(context.TypeFP32(), context.GetFP32(src1), context.GetFP32(src2), context.GetFP32(src3));
 
-                if (!context.Config.GpuAccessor.QueryHostReducedPrecision())
+                if (!context.Config.GpuAccessor.QueryHostReducedPrecision() || operation.ForcePrecise)
                 {
                     context.Decorate(result, Decoration.NoContraction);
                 }

src/Ryujinx.Graphics.Shader/CodeGen/Spirv/ScalingHelpers.cs

@@ -1,227 +0,0 @@
-using Ryujinx.Graphics.Shader.IntermediateRepresentation;
-using Ryujinx.Graphics.Shader.StructuredIr;
-using Ryujinx.Graphics.Shader.Translation;
-using static Spv.Specification;
-
-namespace Ryujinx.Graphics.Shader.CodeGen.Spirv
-{
-    using SpvInstruction = Spv.Generator.Instruction;
-
-    static class ScalingHelpers
-    {
-        public static SpvInstruction ApplyScaling(
-            CodeGenContext context,
-            AstTextureOperation texOp,
-            SpvInstruction vector,
-            bool intCoords,
-            bool isBindless,
-            bool isIndexed,
-            bool isArray,
-            int pCount)
-        {
-            if (intCoords)
-            {
-                if (context.Config.Stage.SupportsRenderScale() &&
-                    !isBindless &&
-                    !isIndexed)
-                {
-                    int index = texOp.Inst == Instruction.ImageLoad
-                        ? context.Config.GetTextureDescriptors().Length + context.Config.FindImageDescriptorIndex(texOp)
-                        : context.Config.FindTextureDescriptorIndex(texOp);
-
-                    if (pCount == 3 && isArray)
-                    {
-                        return ApplyScaling2DArray(context, vector, index);
-                    }
-                    else if (pCount == 2 && !isArray)
-                    {
-                        return ApplyScaling2D(context, vector, index);
-                    }
-                }
-            }
-
-            return vector;
-        }
-
-        private static SpvInstruction ApplyScaling2DArray(CodeGenContext context, SpvInstruction vector, int index)
-        {
-            // The array index is not scaled, just x and y.
-            var vectorXY = context.VectorShuffle(context.TypeVector(context.TypeS32(), 2), vector, vector, 0, 1);
-            var vectorZ = context.CompositeExtract(context.TypeS32(), vector, 2);
-            var vectorXYScaled = ApplyScaling2D(context, vectorXY, index);
-            var vectorScaled = context.CompositeConstruct(context.TypeVector(context.TypeS32(), 3), vectorXYScaled, vectorZ);
-
-            return vectorScaled;
-        }
-
-        private static SpvInstruction ApplyScaling2D(CodeGenContext context, SpvInstruction vector, int index)
-        {
-            var pointerType = context.TypePointer(StorageClass.Uniform, context.TypeFP32());
-            var fieldIndex = context.Constant(context.TypeU32(), 4);
-            var scaleIndex = context.Constant(context.TypeU32(), index);
-
-            if (context.Config.Stage == ShaderStage.Vertex)
-            {
-                var scaleCountPointerType = context.TypePointer(StorageClass.Uniform, context.TypeS32());
-                var scaleCountElemPointer = context.AccessChain(scaleCountPointerType, context.ConstantBuffers[0], context.Constant(context.TypeU32(), 3));
-                var scaleCount = context.Load(context.TypeS32(), scaleCountElemPointer);
-
-                scaleIndex = context.IAdd(context.TypeU32(), scaleIndex, scaleCount);
-            }
-
-            scaleIndex = context.IAdd(context.TypeU32(), scaleIndex, context.Constant(context.TypeU32(), 1));
-
-            var scaleElemPointer = context.AccessChain(pointerType, context.ConstantBuffers[0], fieldIndex, scaleIndex);
-            var scale = context.Load(context.TypeFP32(), scaleElemPointer);
-
-            var ivector2Type = context.TypeVector(context.TypeS32(), 2);
-            var localVector = context.CoordTemp;
-
-            var passthrough = context.FOrdEqual(context.TypeBool(), scale, context.Constant(context.TypeFP32(), 1f));
-
-            var mergeLabel = context.Label();
-
-            if (context.Config.Stage == ShaderStage.Fragment)
-            {
-                var scaledInterpolatedLabel = context.Label();
-                var scaledNoInterpolationLabel = context.Label();
-
-                var needsInterpolation = context.FOrdLessThan(context.TypeBool(), scale, context.Constant(context.TypeFP32(), 0f));
-
-                context.SelectionMerge(mergeLabel, SelectionControlMask.MaskNone);
-                context.BranchConditional(needsInterpolation, scaledInterpolatedLabel, scaledNoInterpolationLabel);
-
-                // scale < 0.0
-                context.AddLabel(scaledInterpolatedLabel);
-
-                ApplyScalingInterpolated(context, localVector, vector, scale);
-                context.Branch(mergeLabel);
-
-                // scale >= 0.0
-                context.AddLabel(scaledNoInterpolationLabel);
-
-                ApplyScalingNoInterpolation(context, localVector, vector, scale);
-                context.Branch(mergeLabel);
-
-                context.AddLabel(mergeLabel);
-
-                var passthroughLabel = context.Label();
-                var finalMergeLabel = context.Label();
-
-                context.SelectionMerge(finalMergeLabel, SelectionControlMask.MaskNone);
-                context.BranchConditional(passthrough, passthroughLabel, finalMergeLabel);
-
-                context.AddLabel(passthroughLabel);
-
-                context.Store(localVector, vector);
-                context.Branch(finalMergeLabel);
-
-                context.AddLabel(finalMergeLabel);
-
-                return context.Load(ivector2Type, localVector);
-            }
-            else
-            {
-                var passthroughLabel = context.Label();
-                var scaledLabel = context.Label();
-
-                context.SelectionMerge(mergeLabel, SelectionControlMask.MaskNone);
-                context.BranchConditional(passthrough, passthroughLabel, scaledLabel);
-
-                // scale == 1.0
-                context.AddLabel(passthroughLabel);
-
-                context.Store(localVector, vector);
-                context.Branch(mergeLabel);
-
-                // scale != 1.0
-                context.AddLabel(scaledLabel);
-
-                ApplyScalingNoInterpolation(context, localVector, vector, scale);
-                context.Branch(mergeLabel);
-
-                context.AddLabel(mergeLabel);
-
-                return context.Load(ivector2Type, localVector);
-            }
-        }
-
-        private static void ApplyScalingInterpolated(CodeGenContext context, SpvInstruction output, SpvInstruction vector, SpvInstruction scale)
-        {
-            var vector2Type = context.TypeVector(context.TypeFP32(), 2);
-
-            var scaleNegated = context.FNegate(context.TypeFP32(), scale);
-            var scaleVector = context.CompositeConstruct(vector2Type, scaleNegated, scaleNegated);
-
-            var vectorFloat = context.ConvertSToF(vector2Type, vector);
-            var vectorScaled = context.VectorTimesScalar(vector2Type, vectorFloat, scaleNegated);
-
-            var fragCoordPointer = context.Inputs[new IoDefinition(StorageKind.Input, IoVariable.FragmentCoord)];
-            var fragCoord = context.Load(context.TypeVector(context.TypeFP32(), 4), fragCoordPointer);
-            var fragCoordXY = context.VectorShuffle(vector2Type, fragCoord, fragCoord, 0, 1);
-
-            var scaleMod = context.FMod(vector2Type, fragCoordXY, scaleVector);
-            var vectorInterpolated = context.FAdd(vector2Type, vectorScaled, scaleMod);
-
-            context.Store(output, context.ConvertFToS(context.TypeVector(context.TypeS32(), 2), vectorInterpolated));
-        }
-
-        private static void ApplyScalingNoInterpolation(CodeGenContext context, SpvInstruction output, SpvInstruction vector, SpvInstruction scale)
-        {
-            if (context.Config.Stage == ShaderStage.Vertex)
-            {
-                scale = context.GlslFAbs(context.TypeFP32(), scale);
-            }
-
-            var vector2Type = context.TypeVector(context.TypeFP32(), 2);
-
-            var vectorFloat = context.ConvertSToF(vector2Type, vector);
-            var vectorScaled = context.VectorTimesScalar(vector2Type, vectorFloat, scale);
-
-            context.Store(output, context.ConvertFToS(context.TypeVector(context.TypeS32(), 2), vectorScaled));
-        }
-
-        public static SpvInstruction ApplyUnscaling(
-            CodeGenContext context,
-            AstTextureOperation texOp,
-            SpvInstruction size,
-            bool isBindless,
-            bool isIndexed)
-        {
-            if (context.Config.Stage.SupportsRenderScale() &&
-                !isBindless &&
-                !isIndexed)
-            {
-                int index = context.Config.FindTextureDescriptorIndex(texOp);
-
-                var pointerType = context.TypePointer(StorageClass.Uniform, context.TypeFP32());
-                var fieldIndex = context.Constant(context.TypeU32(), 4);
-                var scaleIndex = context.Constant(context.TypeU32(), index);
-
-                if (context.Config.Stage == ShaderStage.Vertex)
-                {
-                    var scaleCountPointerType = context.TypePointer(StorageClass.Uniform, context.TypeS32());
-                    var scaleCountElemPointer = context.AccessChain(scaleCountPointerType, context.ConstantBuffers[0], context.Constant(context.TypeU32(), 3));
-                    var scaleCount = context.Load(context.TypeS32(), scaleCountElemPointer);
-
-                    scaleIndex = context.IAdd(context.TypeU32(), scaleIndex, scaleCount);
-                }
-
-                scaleIndex = context.IAdd(context.TypeU32(), scaleIndex, context.Constant(context.TypeU32(), 1));
-
-                var scaleElemPointer = context.AccessChain(pointerType, context.ConstantBuffers[0], fieldIndex, scaleIndex);
-                var scale = context.GlslFAbs(context.TypeFP32(), context.Load(context.TypeFP32(), scaleElemPointer));
-
-                var passthrough = context.FOrdEqual(context.TypeBool(), scale, context.Constant(context.TypeFP32(), 1f));
-
-                var sizeFloat = context.ConvertSToF(context.TypeFP32(), size);
-                var sizeUnscaled = context.FDiv(context.TypeFP32(), sizeFloat, scale);
-                var sizeUnscaledInt = context.ConvertFToS(context.TypeS32(), sizeUnscaled);
-
-                return context.Select(context.TypeS32(), passthrough, size, sizeUnscaledInt);
-            }
-
-            return size;
-        }
-    }
-}

src/Ryujinx.Graphics.Shader/CodeGen/Spirv/SpirvDelegates.cs

@@ -67,6 +67,7 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Spirv
         public readonly FuncBinaryInstruction GlslSMax;
         public readonly FuncBinaryInstruction GlslFMin;
         public readonly FuncBinaryInstruction GlslSMin;
+        public readonly FuncBinaryInstruction FMod;
         public readonly FuncBinaryInstruction FMul;
         public readonly FuncBinaryInstruction IMul;
         public readonly FuncBinaryInstruction FSub;
@@ -174,6 +175,7 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Spirv
             GlslSMax = context.GlslSMax;
             GlslFMin = context.GlslFMin;
             GlslSMin = context.GlslSMin;
+            FMod = context.FMod;
             FMul = context.FMul;
             IMul = context.IMul;
             FSub = context.FSub;

src/Ryujinx.Graphics.Shader/CodeGen/Spirv/SpirvGenerator.cs

@@ -144,10 +144,9 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Spirv
 
         private static void Generate(CodeGenContext context, StructuredProgramInfo info, int funcIndex)
         {
-            var function = info.Functions[funcIndex];
+            (var function, var spvFunc) = context.GetFunction(funcIndex);
-
-            (_, var spvFunc) = context.GetFunction(funcIndex);
 
+            context.CurrentFunction = function;
             context.AddFunction(spvFunc);
             context.StartFunction();
 

src/Ryujinx.Graphics.Shader/IGpuAccessor.cs

@@ -367,6 +367,15 @@ namespace Ryujinx.Graphics.Shader
             return true;
         }
 
+        /// <summary>
+        /// Queries whether the host supports depth clip control.
+        /// </summary>
+        /// <returns>True if the GPU and driver supports depth clip control, false otherwise</returns>
+        bool QueryHostSupportsDepthClipControl()
+        {
+            return true;
+        }
+
         /// <summary>
         /// Queries the point size from the GPU state, used when it is not explicitly set on the shader.
         /// </summary>

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

@@ -187,27 +187,6 @@ namespace Ryujinx.Graphics.Shader.Instructions
             context.Config.GpuAccessor.Log("Shader instruction Longjmp is not implemented.");
         }
 
-        public static void P2rR(EmitterContext context)
-        {
-            InstP2rR op = context.GetOp<InstP2rR>();
-
-            context.Config.GpuAccessor.Log("Shader instruction P2rR is not implemented.");
-        }
-
-        public static void P2rI(EmitterContext context)
-        {
-            InstP2rI op = context.GetOp<InstP2rI>();
-
-            context.Config.GpuAccessor.Log("Shader instruction P2rI is not implemented.");
-        }
-
-        public static void P2rC(EmitterContext context)
-        {
-            InstP2rC op = context.GetOp<InstP2rC>();
-
-            context.Config.GpuAccessor.Log("Shader instruction P2rC is not implemented.");
-        }
-
         public static void Pexit(EmitterContext context)
         {
             InstPexit op = context.GetOp<InstPexit>();

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

@@ -209,21 +209,15 @@ namespace Ryujinx.Graphics.Shader.Instructions
                 return context.ICompareNotEqual(context.BitwiseAnd(value, Const(1 << bit)), Const(0));
             }
 
-            if (ccpr)
+            int count = ccpr ? RegisterConsts.FlagsCount : RegisterConsts.PredsCount;
+            RegisterType type = ccpr ? RegisterType.Flag : RegisterType.Predicate;
+            int shift = (int)byteSel * 8;
+            
+            for (int bit = 0; bit < count; bit++)
             {
-                // TODO: Support Register to condition code flags copy.
+                Operand flag = Register(bit, type);
-                context.Config.GpuAccessor.Log("R2P.CC not implemented.");
+                Operand res = context.ConditionalSelect(Test(mask, bit), Test(value, bit + shift), flag);
-            }
+                context.Copy(flag, res);
-            else
-            {
-                int shift = (int)byteSel * 8;
-
-                for (int bit = 0; bit < RegisterConsts.PredsCount; bit++)
-                {
-                    Operand pred = Register(bit, RegisterType.Predicate);
-                    Operand res = context.ConditionalSelect(Test(mask, bit), Test(value, bit + shift), pred);
-                    context.Copy(pred, res);
-                }
             }
         }
 

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

@@ -1,7 +1,6 @@
 using Ryujinx.Graphics.Shader.Decoders;
 using Ryujinx.Graphics.Shader.IntermediateRepresentation;
 using Ryujinx.Graphics.Shader.Translation;
-
 using static Ryujinx.Graphics.Shader.Instructions.InstEmitAluHelper;
 using static Ryujinx.Graphics.Shader.Instructions.InstEmitHelper;
 using static Ryujinx.Graphics.Shader.IntermediateRepresentation.OperandHelper;
@@ -50,5 +49,68 @@ namespace Ryujinx.Graphics.Shader.Instructions
             context.Copy(Register(op.DestPred, RegisterType.Predicate), p0Res);
             context.Copy(Register(op.DestPredInv, RegisterType.Predicate), p1Res);
         }
+
+        public static void P2rC(EmitterContext context)
+        {
+            InstP2rC op = context.GetOp<InstP2rC>();
+
+            Operand srcA = GetSrcReg(context, op.SrcA);
+            Operand dest = GetSrcReg(context, op.Dest);
+            Operand mask = GetSrcCbuf(context, op.CbufSlot, op.CbufOffset);
+
+            EmitP2r(context, srcA, dest, mask, op.ByteSel, op.Ccpr);
+        }
+
+        public static void P2rI(EmitterContext context)
+        {
+            InstP2rI op = context.GetOp<InstP2rI>();
+
+            Operand srcA = GetSrcReg(context, op.SrcA);
+            Operand dest = GetSrcReg(context, op.Dest);
+            Operand mask = GetSrcImm(context, op.Imm20);
+
+            EmitP2r(context, srcA, dest, mask, op.ByteSel, op.Ccpr);
+        }
+
+        public static void P2rR(EmitterContext context)
+        {
+            InstP2rR op = context.GetOp<InstP2rR>();
+
+            Operand srcA = GetSrcReg(context, op.SrcA);
+            Operand dest = GetSrcReg(context, op.Dest);
+            Operand mask = GetSrcReg(context, op.SrcB);
+
+            EmitP2r(context, srcA, dest, mask, op.ByteSel, op.Ccpr);
+        }
+
+        private static void EmitP2r(
+            EmitterContext context,
+            Operand srcA,
+            Operand dest,
+            Operand mask,
+            ByteSel byteSel,
+            bool ccpr)
+        {
+            int count = ccpr ? RegisterConsts.FlagsCount : RegisterConsts.PredsCount;
+            int shift = (int)byteSel * 8;
+            mask = context.BitwiseAnd(mask, Const(0xff));
+
+            Operand insert = Const(0);
+            for (int i = 0; i < count; i++)
+            {
+                Operand condition = ccpr
+                    ? Register(i, RegisterType.Flag)
+                    : Register(i, RegisterType.Predicate);
+
+                Operand bit = context.ConditionalSelect(condition, Const(1 << (i + shift)), Const(0));
+                insert = context.BitwiseOr(insert, bit);
+            }
+
+            Operand maskShifted = context.ShiftLeft(mask, Const(shift));
+            Operand masked = context.BitwiseAnd(srcA, context.BitwiseNot(maskShifted));
+            Operand res = context.BitwiseOr(masked, context.BitwiseAnd(insert, maskShifted));
+
+            context.Copy(dest, res);
+        }
     }
 }

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

@@ -97,6 +97,7 @@ namespace Ryujinx.Graphics.Shader.IntermediateRepresentation
         MemoryBarrier,
         Minimum,
         MinimumU32,
+        Modulo,
         Multiply,
         MultiplyHighS32,
         MultiplyHighU32,

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

@@ -8,6 +8,8 @@ namespace Ryujinx.Graphics.Shader.IntermediateRepresentation
         public Instruction Inst { get; private set; }
         public StorageKind StorageKind { get; }
 
+        public bool ForcePrecise { get; set; }
+
         private Operand[] _dests;
 
         public Operand Dest

src/Ryujinx.Graphics.Shader/Ryujinx.Graphics.Shader.csproj

@@ -4,10 +4,6 @@
     <TargetFramework>net7.0</TargetFramework>
   </PropertyGroup>
 
-  <ItemGroup>
-    <None Remove="CodeGen\Glsl\HelperFunctions\TexelFetchScale_vp.glsl" />
-  </ItemGroup>
-
   <ItemGroup>
     <ProjectReference Include="..\Ryujinx.Common\Ryujinx.Common.csproj" />
     <ProjectReference Include="..\Spv.Generator\Spv.Generator.csproj" />
@@ -25,9 +21,6 @@
     <EmbeddedResource Include="CodeGen\Glsl\HelperFunctions\StoreSharedSmallInt.glsl" />
     <EmbeddedResource Include="CodeGen\Glsl\HelperFunctions\StoreStorageSmallInt.glsl" />
     <EmbeddedResource Include="CodeGen\Glsl\HelperFunctions\SwizzleAdd.glsl" />
-    <EmbeddedResource Include="CodeGen\Glsl\HelperFunctions\TexelFetchScale_vp.glsl" />
-    <EmbeddedResource Include="CodeGen\Glsl\HelperFunctions\TexelFetchScale_fp.glsl" />
-    <EmbeddedResource Include="CodeGen\Glsl\HelperFunctions\TexelFetchScale_cp.glsl" />
   </ItemGroup>
 
 </Project>

src/Ryujinx.Graphics.Shader/StructuredIr/AstOperation.cs

@@ -10,6 +10,7 @@ namespace Ryujinx.Graphics.Shader.StructuredIr
     {
         public Instruction Inst { get; }
         public StorageKind StorageKind { get; }
+        public bool ForcePrecise { get; }
 
         public int Index { get; }
 
@@ -17,10 +18,11 @@ namespace Ryujinx.Graphics.Shader.StructuredIr
 
         public int SourcesCount => _sources.Length;
 
-        public AstOperation(Instruction inst, StorageKind storageKind, IAstNode[] sources, int sourcesCount)
+        public AstOperation(Instruction inst, StorageKind storageKind, bool forcePrecise, IAstNode[] sources, int sourcesCount)
         {
             Inst = inst;
             StorageKind = storageKind;
+            ForcePrecise = forcePrecise;
             _sources = sources;
 
             for (int index = 0; index < sources.Length; index++)
@@ -38,12 +40,18 @@ namespace Ryujinx.Graphics.Shader.StructuredIr
             Index = 0;
         }
 
-        public AstOperation(Instruction inst, StorageKind storageKind, int index, IAstNode[] sources, int sourcesCount) : this(inst, storageKind, sources, sourcesCount)
+        public AstOperation(
+            Instruction inst,
+            StorageKind storageKind,
+            bool forcePrecise,
+            int index,
+            IAstNode[] sources,
+            int sourcesCount) : this(inst, storageKind, forcePrecise, sources, sourcesCount)
         {
             Index = index;
         }
 
-        public AstOperation(Instruction inst, params IAstNode[] sources) : this(inst, StorageKind.None, sources, sources.Length)
+        public AstOperation(Instruction inst, params IAstNode[] sources) : this(inst, StorageKind.None, false, sources, sources.Length)
         {
         }
 

src/Ryujinx.Graphics.Shader/StructuredIr/AstTextureOperation.cs

@@ -19,7 +19,7 @@ namespace Ryujinx.Graphics.Shader.StructuredIr
             int cbufSlot,
             int handle,
             int index,
-            params IAstNode[] sources) : base(inst, StorageKind.None, index, sources, sources.Length)
+            params IAstNode[] sources) : base(inst, StorageKind.None, false, index, sources, sources.Length)
         {
             Type = type;
             Format = format;
@@ -27,10 +27,5 @@ namespace Ryujinx.Graphics.Shader.StructuredIr
             CbufSlot = cbufSlot;
             Handle = handle;
         }
-
-        public AstTextureOperation WithType(SamplerType type)
-        {
-            return new AstTextureOperation(Inst, type, Format, Flags, CbufSlot, Handle, Index);
-        }
     }
 }

src/Ryujinx.Graphics.Shader/StructuredIr/InstructionInfo.cs

@@ -104,6 +104,7 @@ namespace Ryujinx.Graphics.Shader.StructuredIr
             Add(Instruction.MaximumU32,               AggregateType.U32,    AggregateType.U32,     AggregateType.U32);
             Add(Instruction.Minimum,                  AggregateType.Scalar, AggregateType.Scalar,  AggregateType.Scalar);
             Add(Instruction.MinimumU32,               AggregateType.U32,    AggregateType.U32,     AggregateType.U32);
+            Add(Instruction.Modulo,                   AggregateType.Scalar, AggregateType.Scalar,  AggregateType.Scalar);
             Add(Instruction.Multiply,                 AggregateType.Scalar, AggregateType.Scalar,  AggregateType.Scalar);
             Add(Instruction.MultiplyHighS32,          AggregateType.S32,    AggregateType.S32,     AggregateType.S32);
             Add(Instruction.MultiplyHighU32,          AggregateType.U32,    AggregateType.U32,     AggregateType.U32);

src/Ryujinx.Graphics.Shader/StructuredIr/StructuredProgram.cs

@@ -8,11 +8,11 @@ namespace Ryujinx.Graphics.Shader.StructuredIr
 {
     static class StructuredProgram
     {
-        public static StructuredProgramInfo MakeStructuredProgram(Function[] functions, ShaderConfig config)
+        public static StructuredProgramInfo MakeStructuredProgram(IReadOnlyList<Function> functions, ShaderConfig config)
         {
             StructuredProgramContext context = new StructuredProgramContext(config);
 
-            for (int funcIndex = 0; funcIndex < functions.Length; funcIndex++)
+            for (int funcIndex = 0; funcIndex < functions.Count; funcIndex++)
             {
                 Function function = functions[funcIndex];
 
@@ -156,7 +156,13 @@ namespace Ryujinx.Graphics.Shader.StructuredIr
                 }
                 else
                 {
-                    source = new AstOperation(inst, operation.StorageKind, operation.Index, sources, operation.SourcesCount);
+                    source = new AstOperation(
+                        inst,
+                        operation.StorageKind,
+                        operation.ForcePrecise,
+                        operation.Index,
+                        sources,
+                        operation.SourcesCount);
                 }
 
                 AggregateType destElemType = destType;
@@ -179,7 +185,7 @@ namespace Ryujinx.Graphics.Shader.StructuredIr
 
                     dest.VarType = destElemType;
 
-                    context.AddNode(new AstAssignment(dest, new AstOperation(Instruction.VectorExtract, StorageKind.None, new[] { destVec, index }, 2)));
+                    context.AddNode(new AstAssignment(dest, new AstOperation(Instruction.VectorExtract, StorageKind.None, false, new[] { destVec, index }, 2)));
                 }
             }
             else if (operation.Dest != null)
@@ -227,7 +233,13 @@ namespace Ryujinx.Graphics.Shader.StructuredIr
                 }
                 else if (!isCopy)
                 {
-                    source = new AstOperation(inst, operation.StorageKind, operation.Index, sources, operation.SourcesCount);
+                    source = new AstOperation(
+                        inst,
+                        operation.StorageKind,
+                        operation.ForcePrecise,
+                        operation.Index,
+                        sources,
+                        operation.SourcesCount);
                 }
                 else
                 {
@@ -248,7 +260,13 @@ namespace Ryujinx.Graphics.Shader.StructuredIr
             }
             else
             {
-                context.AddNode(new AstOperation(inst, operation.StorageKind, operation.Index, sources, operation.SourcesCount));
+                context.AddNode(new AstOperation(
+                    inst,
+                    operation.StorageKind,
+                    operation.ForcePrecise,
+                    operation.Index,
+                    sources,
+                    operation.SourcesCount));
             }
 
             // Those instructions needs to be emulated by using helper functions,

src/Ryujinx.Graphics.Shader/StructuredIr/StructuredProgramContext.cs

@@ -319,7 +319,7 @@ namespace Ryujinx.Graphics.Shader.StructuredIr
                     new AstOperand(OperandType.Constant, elemIndex)
                 };
 
-                return new AstOperation(Instruction.Load, StorageKind.ConstantBuffer, sources, sources.Length);
+                return new AstOperation(Instruction.Load, StorageKind.ConstantBuffer, false, sources, sources.Length);
             }
 
             return GetOperand(operand);

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

@@ -49,13 +49,17 @@ namespace Ryujinx.Graphics.Shader.Translation
         private readonly List<Operation> _operations;
         private readonly Dictionary<ulong, BlockLabel> _labels;
 
-        public EmitterContext(DecodedProgram program, ShaderConfig config, bool isNonMain)
+        public EmitterContext()
+        {
+            _operations = new List<Operation>();
+            _labels = new Dictionary<ulong, BlockLabel>();
+        }
+
+        public EmitterContext(DecodedProgram program, ShaderConfig config, bool isNonMain) : this()
         {
             Program = program;
             Config = config;
             IsNonMain = isNonMain;
-            _operations = new List<Operation>();
-            _labels = new Dictionary<ulong, BlockLabel>();
 
             EmitStart();
         }
@@ -242,7 +246,7 @@ namespace Ryujinx.Graphics.Shader.Translation
                 this.Store(StorageKind.Output, IoVariable.Position, null, Const(1), this.FPFusedMultiplyAdd(y, yScale, negativeOne));
             }
 
-            if (Config.Options.TargetApi == TargetApi.Vulkan && Config.GpuAccessor.QueryTransformDepthMinusOneToOne())
+            if (Config.GpuAccessor.QueryTransformDepthMinusOneToOne() && !Config.GpuAccessor.QueryHostSupportsDepthClipControl())
             {
                 Operand z = this.Load(StorageKind.Output, IoVariable.Position, null, Const(2));
                 Operand w = this.Load(StorageKind.Output, IoVariable.Position, null, Const(3));
@@ -279,7 +283,7 @@ namespace Ryujinx.Graphics.Shader.Translation
                 oldYLocal = null;
             }
 
-            if (Config.Options.TargetApi == TargetApi.Vulkan && Config.GpuAccessor.QueryTransformDepthMinusOneToOne())
+            if (Config.GpuAccessor.QueryTransformDepthMinusOneToOne() && !Config.GpuAccessor.QueryHostSupportsDepthClipControl())
             {
                 oldZLocal = Local();
                 this.Copy(oldZLocal, this.Load(StorageKind.Output, IoVariable.Position, null, Const(2)));

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

@@ -307,6 +307,11 @@ namespace Ryujinx.Graphics.Shader.Translation
             return context.Add(fpType | Instruction.Minimum, Local(), a, b);
         }
 
+        public static Operand FPModulo(this EmitterContext context, Operand a, Operand b, Instruction fpType = Instruction.FP32)
+        {
+            return context.Add(fpType | Instruction.Modulo, Local(), a, b);
+        }
+
         public static Operand FPMultiply(this EmitterContext context, Operand a, Operand b, Instruction fpType = Instruction.FP32)
         {
             return context.Add(fpType | Instruction.Multiply, Local(), a, b);
@@ -656,7 +661,6 @@ namespace Ryujinx.Graphics.Shader.Translation
 
         public static void Return(this EmitterContext context, Operand returnValue)
         {
-            context.PrepareForReturn();
             context.Add(Instruction.Return, null, returnValue);
         }
 

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

@@ -0,0 +1,134 @@
+using Ryujinx.Graphics.Shader.IntermediateRepresentation;
+using System;
+using System.Collections.Generic;
+
+using static Ryujinx.Graphics.Shader.IntermediateRepresentation.OperandHelper;
+
+namespace Ryujinx.Graphics.Shader.Translation
+{
+    class HelperFunctionManager
+    {
+        private readonly List<Function> _functionList;
+        private readonly Dictionary<HelperFunctionName, int> _functionIds;
+        private readonly ShaderStage _stage;
+
+        public HelperFunctionManager(List<Function> functionList, ShaderStage stage)
+        {
+            _functionList = functionList;
+            _functionIds = new Dictionary<HelperFunctionName, int>();
+            _stage = stage;
+        }
+
+        public int GetOrCreateFunctionId(HelperFunctionName functionName)
+        {
+            if (_functionIds.TryGetValue(functionName, out int functionId))
+            {
+                return functionId;
+            }
+
+            Function function = GenerateFunction(functionName);
+            functionId = _functionList.Count;
+            _functionList.Add(function);
+            _functionIds.Add(functionName, functionId);
+
+            return functionId;
+        }
+
+        private Function GenerateFunction(HelperFunctionName functionName)
+        {
+            return functionName switch
+            {
+                HelperFunctionName.TexelFetchScale => GenerateTexelFetchScaleFunction(),
+                HelperFunctionName.TextureSizeUnscale => GenerateTextureSizeUnscaleFunction(),
+                _ => throw new ArgumentException($"Invalid function name {functionName}")
+            };
+        }
+
+        private Function GenerateTexelFetchScaleFunction()
+        {
+            EmitterContext context = new EmitterContext();
+
+            Operand input = Argument(0);
+            Operand samplerIndex = Argument(1);
+            Operand index = GetScaleIndex(context, samplerIndex);
+
+            Operand scale = context.Load(StorageKind.ConstantBuffer, 0, Const((int)SupportBufferField.RenderScale), index);
+
+            Operand scaleIsOne = context.FPCompareEqual(scale, ConstF(1f));
+            Operand lblScaleNotOne = Label();
+
+            context.BranchIfFalse(lblScaleNotOne, scaleIsOne);
+            context.Return(input);
+            context.MarkLabel(lblScaleNotOne);
+
+            int inArgumentsCount;
+
+            if (_stage == ShaderStage.Fragment)
+            {
+                Operand scaleIsLessThanZero = context.FPCompareLess(scale, ConstF(0f));
+                Operand lblScaleGreaterOrEqualZero = Label();
+
+                context.BranchIfFalse(lblScaleGreaterOrEqualZero, scaleIsLessThanZero);
+
+                Operand negScale = context.FPNegate(scale);
+                Operand inputScaled = context.FPMultiply(context.IConvertS32ToFP32(input), negScale);
+                Operand fragCoordX = context.Load(StorageKind.Input, IoVariable.FragmentCoord, null, Const(0));
+                Operand fragCoordY = context.Load(StorageKind.Input, IoVariable.FragmentCoord, null, Const(1));
+                Operand fragCoord = context.ConditionalSelect(Argument(2), fragCoordY, fragCoordX);
+                Operand inputBias = context.FPModulo(fragCoord, negScale);
+                Operand inputWithBias = context.FPAdd(inputScaled, inputBias);
+
+                context.Return(context.FP32ConvertToS32(inputWithBias));
+                context.MarkLabel(lblScaleGreaterOrEqualZero);
+
+                inArgumentsCount = 3;
+            }
+            else
+            {
+                inArgumentsCount = 2;
+            }
+
+            Operand inputScaled2 = context.FPMultiply(context.IConvertS32ToFP32(input), scale);
+
+            context.Return(context.FP32ConvertToS32(inputScaled2));
+
+            return new Function(ControlFlowGraph.Create(context.GetOperations()).Blocks, "TexelFetchScale", true, inArgumentsCount, 0);
+        }
+
+        private Function GenerateTextureSizeUnscaleFunction()
+        {
+            EmitterContext context = new EmitterContext();
+
+            Operand input = Argument(0);
+            Operand samplerIndex = Argument(1);
+            Operand index = GetScaleIndex(context, samplerIndex);
+
+            Operand scale = context.FPAbsolute(context.Load(StorageKind.ConstantBuffer, 0, Const((int)SupportBufferField.RenderScale), index));
+
+            Operand scaleIsOne = context.FPCompareEqual(scale, ConstF(1f));
+            Operand lblScaleNotOne = Label();
+
+            context.BranchIfFalse(lblScaleNotOne, scaleIsOne);
+            context.Return(input);
+            context.MarkLabel(lblScaleNotOne);
+
+            Operand inputUnscaled = context.FPDivide(context.IConvertS32ToFP32(input), scale);
+
+            context.Return(context.FP32ConvertToS32(inputUnscaled));
+
+            return new Function(ControlFlowGraph.Create(context.GetOperations()).Blocks, "TextureSizeUnscale", true, 2, 0);
+        }
+
+        private Operand GetScaleIndex(EmitterContext context, Operand index)
+        {
+            switch (_stage)
+            {
+                case ShaderStage.Vertex:
+                    Operand fragScaleCount = context.Load(StorageKind.ConstantBuffer, 0, Const((int)SupportBufferField.FragmentRenderScaleCount));
+                    return context.IAdd(Const(1), context.IAdd(index, fragScaleCount));
+                default:
+                    return context.IAdd(Const(1), index);
+            }
+        }
+    }
+}

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

@@ -0,0 +1,11 @@
+using Ryujinx.Graphics.Shader.IntermediateRepresentation;
+using System.Collections.Generic;
+
+namespace Ryujinx.Graphics.Shader.Translation
+{
+    enum HelperFunctionName
+    {
+        TexelFetchScale,
+        TextureSizeUnscale
+    }
+}

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

@@ -11,9 +11,10 @@ namespace Ryujinx.Graphics.Shader.Translation
 {
     static class Rewriter
     {
-        public static void RunPass(BasicBlock[] blocks, ShaderConfig config)
+        public static void RunPass(HelperFunctionManager hfm, BasicBlock[] blocks, ShaderConfig config)
         {
             bool isVertexShader = config.Stage == ShaderStage.Vertex;
+            bool isImpreciseFragmentShader = config.Stage == ShaderStage.Fragment && config.GpuAccessor.QueryHostReducedPrecision();
             bool hasConstantBufferDrawParameters = config.GpuAccessor.QueryHasConstantBufferDrawParameters();
             bool hasVectorIndexingBug = config.GpuAccessor.QueryHostHasVectorIndexingBug();
             bool supportsSnormBufferTextureFormat = config.GpuAccessor.QueryHostSupportsSnormBufferTextureFormat();
@@ -45,6 +46,11 @@ namespace Ryujinx.Graphics.Shader.Translation
                         }
                     }
 
+                    if (isImpreciseFragmentShader)
+                    {
+                        EnableForcePreciseIfNeeded(operation);
+                    }
+
                     if (hasVectorIndexingBug)
                     {
                         InsertVectorComponentSelect(node, config);
@@ -54,9 +60,14 @@ namespace Ryujinx.Graphics.Shader.Translation
 
                     if (operation is TextureOperation texOp)
                     {
+                        node = InsertTexelFetchScale(hfm, node, config);
+                        node = InsertTextureSizeUnscale(hfm, node, config);
+
                         if (texOp.Inst == Instruction.TextureSample)
                         {
-                            node = RewriteTextureSample(node, config);
+                            node = InsertCoordNormalization(node, config);
+                            node = InsertCoordGatherBias(node, config);
+                            node = InsertConstOffsets(node, config);
 
                             if (texOp.Type == SamplerType.TextureBuffer && !supportsSnormBufferTextureFormat)
                             {
@@ -76,6 +87,25 @@ namespace Ryujinx.Graphics.Shader.Translation
             }
         }
 
+        private static void EnableForcePreciseIfNeeded(Operation operation)
+        {
+            // There are some cases where a small bias is added to values to prevent division by zero.
+            // When operating with reduced precision, it is possible for this bias to get rounded to 0
+            // and cause a division by zero.
+            // To prevent that, we force those operations to be precise even if the host wants
+            // imprecise operations for performance.
+
+            if (operation.Inst == (Instruction.FP32 | Instruction.Divide) &&
+                operation.GetSource(0).Type == OperandType.Constant &&
+                operation.GetSource(0).AsFloat() == 1f &&
+                operation.GetSource(1).AsgOp is Operation addOp &&
+                addOp.Inst == (Instruction.FP32 | Instruction.Add) &&
+                addOp.GetSource(1).Type == OperandType.Constant)
+            {
+                addOp.ForcePrecise = true;
+            }
+        }
+
         private static void InsertVectorComponentSelect(LinkedListNode<INode> node, ShaderConfig config)
         {
             Operation operation = (Operation)node.Value;
@@ -344,10 +374,279 @@ namespace Ryujinx.Graphics.Shader.Translation
             return node;
         }
 
-        private static LinkedListNode<INode> RewriteTextureSample(LinkedListNode<INode> node, ShaderConfig config)
+        private static LinkedListNode<INode> InsertTexelFetchScale(HelperFunctionManager hfm, LinkedListNode<INode> node, ShaderConfig config)
         {
             TextureOperation texOp = (TextureOperation)node.Value;
 
+            bool isBindless = (texOp.Flags & TextureFlags.Bindless)  != 0;
+            bool intCoords  = (texOp.Flags & TextureFlags.IntCoords) != 0;
+
+            bool isArray   = (texOp.Type & SamplerType.Array)   != 0;
+            bool isIndexed = (texOp.Type & SamplerType.Indexed) != 0;
+
+            int coordsCount = texOp.Type.GetDimensions();
+
+            int coordsIndex = isBindless || isIndexed ? 1 : 0;
+
+            bool isImage = IsImageInstructionWithScale(texOp.Inst);
+
+            if ((texOp.Inst == Instruction.TextureSample || isImage) &&
+                (intCoords || isImage) &&
+                !isBindless &&
+                !isIndexed &&
+                config.Stage.SupportsRenderScale() &&
+                TypeSupportsScale(texOp.Type))
+            {
+                int functionId = hfm.GetOrCreateFunctionId(HelperFunctionName.TexelFetchScale);
+                int samplerIndex = isImage
+                    ? config.GetTextureDescriptors().Length + config.FindImageDescriptorIndex(texOp)
+                    : config.FindTextureDescriptorIndex(texOp);
+
+                for (int index = 0; index < coordsCount; index++)
+                {
+                    Operand scaledCoord = Local();
+                    Operand[] callArgs;
+
+                    if (config.Stage == ShaderStage.Fragment)
+                    {
+                        callArgs = new Operand[] { Const(functionId), texOp.GetSource(coordsIndex + index), Const(samplerIndex), Const(index) };
+                    }
+                    else
+                    {
+                        callArgs = new Operand[] { Const(functionId), texOp.GetSource(coordsIndex + index), Const(samplerIndex) };
+                    }
+
+                    node.List.AddBefore(node, new Operation(Instruction.Call, 0, scaledCoord, callArgs));
+
+                    texOp.SetSource(coordsIndex + index, scaledCoord);
+                }
+            }
+
+            return node;
+        }
+
+        private static LinkedListNode<INode> InsertTextureSizeUnscale(HelperFunctionManager hfm, LinkedListNode<INode> node, ShaderConfig config)
+        {
+            TextureOperation texOp = (TextureOperation)node.Value;
+
+            bool isBindless = (texOp.Flags & TextureFlags.Bindless)  != 0;
+            bool intCoords  = (texOp.Flags & TextureFlags.IntCoords) != 0;
+
+            bool isArray   = (texOp.Type & SamplerType.Array)   != 0;
+            bool isIndexed = (texOp.Type & SamplerType.Indexed) != 0;
+
+            if (texOp.Inst == Instruction.TextureSize &&
+                texOp.Index < 2 &&
+                !isBindless &&
+                !isIndexed &&
+                config.Stage.SupportsRenderScale() &&
+                TypeSupportsScale(texOp.Type))
+            {
+                int functionId = hfm.GetOrCreateFunctionId(HelperFunctionName.TextureSizeUnscale);
+                int samplerIndex = config.FindTextureDescriptorIndex(texOp, ignoreType: true);
+
+                for (int index = texOp.DestsCount - 1; index >= 0; index--)
+                {
+                    Operand dest = texOp.GetDest(index);
+
+                    Operand unscaledSize = Local();
+
+                    // Replace all uses with the unscaled size value.
+                    // This must be done before the call is added, since it also is a use of the original size.
+                    foreach (INode useOp in dest.UseOps)
+                    {
+                        for (int srcIndex = 0; srcIndex < useOp.SourcesCount; srcIndex++)
+                        {
+                            if (useOp.GetSource(srcIndex) == dest)
+                            {
+                                useOp.SetSource(srcIndex, unscaledSize);
+                            }
+                        }
+                    }
+
+                    Operand[] callArgs = new Operand[] { Const(functionId), dest, Const(samplerIndex) };
+
+                    node.List.AddAfter(node, new Operation(Instruction.Call, 0, unscaledSize, callArgs));
+                }
+            }
+
+            return node;
+        }
+
+        private static bool IsImageInstructionWithScale(Instruction inst)
+        {
+            // Currently, we don't support scaling images that are modified,
+            // so we only need to care about the load instruction.
+            return inst == Instruction.ImageLoad;
+        }
+
+        private static bool TypeSupportsScale(SamplerType type)
+        {
+            return (type & SamplerType.Mask) == SamplerType.Texture2D;
+        }
+
+        private static LinkedListNode<INode> InsertCoordNormalization(LinkedListNode<INode> node, ShaderConfig config)
+        {
+            // Emulate non-normalized coordinates by normalizing the coordinates on the shader.
+            // Without normalization, the coordinates are expected to the in the [0, W or H] range,
+            // and otherwise, it is expected to be in the [0, 1] range.
+            // We normalize by dividing the coords by the texture size.
+
+            TextureOperation texOp = (TextureOperation)node.Value;
+
+            bool isBindless = (texOp.Flags & TextureFlags.Bindless)  != 0;
+            bool intCoords  = (texOp.Flags & TextureFlags.IntCoords) != 0;
+
+            bool isCoordNormalized = isBindless || config.GpuAccessor.QueryTextureCoordNormalized(texOp.Handle, texOp.CbufSlot);
+
+            if (isCoordNormalized || intCoords)
+            {
+                return node;
+            }
+
+            bool isArray   = (texOp.Type & SamplerType.Array)   != 0;
+            bool isIndexed = (texOp.Type & SamplerType.Indexed) != 0;
+
+            int coordsCount = texOp.Type.GetDimensions();
+            int coordsIndex = isBindless || isIndexed ? 1 : 0;
+
+            config.SetUsedFeature(FeatureFlags.IntegerSampling);
+
+            int normCoordsCount = (texOp.Type & SamplerType.Mask) == SamplerType.TextureCube ? 2 : coordsCount;
+
+            for (int index = 0; index < normCoordsCount; index++)
+            {
+                Operand coordSize = Local();
+
+                Operand[] texSizeSources;
+
+                if (isBindless || isIndexed)
+                {
+                    texSizeSources = new Operand[] { texOp.GetSource(0), Const(0) };
+                }
+                else
+                {
+                    texSizeSources = new Operand[] { Const(0) };
+                }
+
+                node.List.AddBefore(node, new TextureOperation(
+                    Instruction.TextureSize,
+                    texOp.Type,
+                    texOp.Format,
+                    texOp.Flags,
+                    texOp.CbufSlot,
+                    texOp.Handle,
+                    index,
+                    new[] { coordSize },
+                    texSizeSources));
+
+                config.SetUsedTexture(Instruction.TextureSize, texOp.Type, texOp.Format, texOp.Flags, texOp.CbufSlot, texOp.Handle);
+
+                Operand source = texOp.GetSource(coordsIndex + index);
+
+                Operand coordNormalized = Local();
+
+                node.List.AddBefore(node, new Operation(Instruction.FP32 | Instruction.Divide, coordNormalized, source, GenerateI2f(node, coordSize)));
+
+                texOp.SetSource(coordsIndex + index, coordNormalized);
+            }
+
+            return node;
+        }
+
+        private static LinkedListNode<INode> InsertCoordGatherBias(LinkedListNode<INode> node, ShaderConfig config)
+        {
+            // The gather behavior when the coordinate sits right in the middle of two texels is not well defined.
+            // To ensure the correct texel is sampled, we add a small bias value to the coordinate.
+            // This value is calculated as the minimum value required to change the texel it will sample from,
+            // and is 0 if the host does not require the bias.
+
+            TextureOperation texOp = (TextureOperation)node.Value;
+
+            bool isBindless = (texOp.Flags & TextureFlags.Bindless) != 0;
+            bool isGather   = (texOp.Flags & TextureFlags.Gather)   != 0;
+
+            int gatherBiasPrecision = config.GpuAccessor.QueryHostGatherBiasPrecision();
+
+            if (!isGather || gatherBiasPrecision == 0)
+            {
+                return node;
+            }
+
+            bool intCoords = (texOp.Flags & TextureFlags.IntCoords) != 0;
+
+            bool isArray   = (texOp.Type & SamplerType.Array)   != 0;
+            bool isIndexed = (texOp.Type & SamplerType.Indexed) != 0;
+
+            int coordsCount = texOp.Type.GetDimensions();
+            int coordsIndex = isBindless || isIndexed ? 1 : 0;
+
+            config.SetUsedFeature(FeatureFlags.IntegerSampling);
+
+            int normCoordsCount = (texOp.Type & SamplerType.Mask) == SamplerType.TextureCube ? 2 : coordsCount;
+
+            for (int index = 0; index < normCoordsCount; index++)
+            {
+                Operand coordSize = Local();
+                Operand scaledSize = Local();
+                Operand bias = Local();
+
+                Operand[] texSizeSources;
+
+                if (isBindless || isIndexed)
+                {
+                    texSizeSources = new Operand[] { texOp.GetSource(0), Const(0) };
+                }
+                else
+                {
+                    texSizeSources = new Operand[] { Const(0) };
+                }
+
+                node.List.AddBefore(node, new TextureOperation(
+                    Instruction.TextureSize,
+                    texOp.Type,
+                    texOp.Format,
+                    texOp.Flags,
+                    texOp.CbufSlot,
+                    texOp.Handle,
+                    index,
+                    new[] { coordSize },
+                    texSizeSources));
+
+                config.SetUsedTexture(Instruction.TextureSize, texOp.Type, texOp.Format, texOp.Flags, texOp.CbufSlot, texOp.Handle);
+
+                node.List.AddBefore(node, new Operation(
+                    Instruction.FP32 | Instruction.Multiply,
+                    scaledSize,
+                    GenerateI2f(node, coordSize),
+                    ConstF((float)(1 << (gatherBiasPrecision + 1)))));
+                node.List.AddBefore(node, new Operation(Instruction.FP32 | Instruction.Divide, bias, ConstF(1f), scaledSize));
+
+                Operand source = texOp.GetSource(coordsIndex + index);
+
+                Operand coordBiased = Local();
+
+                node.List.AddBefore(node, new Operation(Instruction.FP32 | Instruction.Add, coordBiased, source, bias));
+
+                texOp.SetSource(coordsIndex + index, coordBiased);
+            }
+
+            return node;
+        }
+
+        private static LinkedListNode<INode> InsertConstOffsets(LinkedListNode<INode> node, ShaderConfig config)
+        {
+            // Non-constant texture offsets are not allowed (according to the spec),
+            // however some GPUs does support that.
+            // For GPUs where it is not supported, we can replace the instruction with the following:
+            // For texture*Offset, we replace it by texture*, and add the offset to the P coords.
+            // The offset can be calculated as offset / textureSize(lod), where lod = textureQueryLod(coords).
+            // For texelFetchOffset, we replace it by texelFetch and add the offset to the P coords directly.
+            // For textureGatherOffset, we split the operation into up to 4 operations, one for each component
+            // that is accessed, where each textureGather operation has a different offset for each pixel.
+
+            TextureOperation texOp = (TextureOperation)node.Value;
+
             bool hasOffset  = (texOp.Flags & TextureFlags.Offset)  != 0;
             bool hasOffsets = (texOp.Flags & TextureFlags.Offsets) != 0;
 
@@ -355,9 +654,7 @@ namespace Ryujinx.Graphics.Shader.Translation
 
             bool isBindless = (texOp.Flags & TextureFlags.Bindless) != 0;
 
-            bool isCoordNormalized = isBindless || config.GpuAccessor.QueryTextureCoordNormalized(texOp.Handle, texOp.CbufSlot);
+            if (!hasInvalidOffset)
-
-            if (!hasInvalidOffset && isCoordNormalized)
             {
                 return node;
             }
@@ -454,7 +751,7 @@ namespace Ryujinx.Graphics.Shader.Translation
 
             hasInvalidOffset &= !areAllOffsetsConstant;
 
-            if (!hasInvalidOffset && isCoordNormalized)
+            if (!hasInvalidOffset)
             {
                 return node;
             }
@@ -473,63 +770,6 @@ namespace Ryujinx.Graphics.Shader.Translation
 
             int componentIndex = texOp.Index;
 
-            Operand Float(Operand value)
-            {
-                Operand res = Local();
-
-                node.List.AddBefore(node, new Operation(Instruction.ConvertS32ToFP32, res, value));
-
-                return res;
-            }
-
-            // Emulate non-normalized coordinates by normalizing the coordinates on the shader.
-            // Without normalization, the coordinates are expected to the in the [0, W or H] range,
-            // and otherwise, it is expected to be in the [0, 1] range.
-            // We normalize by dividing the coords by the texture size.
-            if (!isCoordNormalized && !intCoords)
-            {
-                config.SetUsedFeature(FeatureFlags.IntegerSampling);
-
-                int normCoordsCount = (texOp.Type & SamplerType.Mask) == SamplerType.TextureCube ? 2 : coordsCount;
-
-                for (int index = 0; index < normCoordsCount; index++)
-                {
-                    Operand coordSize = Local();
-
-                    Operand[] texSizeSources;
-
-                    if (isBindless || isIndexed)
-                    {
-                        texSizeSources = new Operand[] { sources[0], Const(0) };
-                    }
-                    else
-                    {
-                        texSizeSources = new Operand[] { Const(0) };
-                    }
-
-                    node.List.AddBefore(node, new TextureOperation(
-                        Instruction.TextureSize,
-                        texOp.Type,
-                        texOp.Format,
-                        texOp.Flags,
-                        texOp.CbufSlot,
-                        texOp.Handle,
-                        index,
-                        new[] { coordSize },
-                        texSizeSources));
-
-                    config.SetUsedTexture(Instruction.TextureSize, texOp.Type, texOp.Format, texOp.Flags, texOp.CbufSlot, texOp.Handle);
-
-                    Operand source = sources[coordsIndex + index];
-
-                    Operand coordNormalized = Local();
-
-                    node.List.AddBefore(node, new Operation(Instruction.FP32 | Instruction.Divide, coordNormalized, source, Float(coordSize)));
-
-                    sources[coordsIndex + index] = coordNormalized;
-                }
-            }
-
             Operand[] dests = new Operand[texOp.DestsCount];
 
             for (int i = 0; i < texOp.DestsCount; i++)
@@ -541,15 +781,7 @@ namespace Ryujinx.Graphics.Shader.Translation
 
             LinkedListNode<INode> oldNode = node;
 
-            // Technically, non-constant texture offsets are not allowed (according to the spec),
+            if (isGather && !isShadow)
-            // however some GPUs does support that.
-            // For GPUs where it is not supported, we can replace the instruction with the following:
-            // For texture*Offset, we replace it by texture*, and add the offset to the P coords.
-            // The offset can be calculated as offset / textureSize(lod), where lod = textureQueryLod(coords).
-            // For texelFetchOffset, we replace it by texelFetch and add the offset to the P coords directly.
-            // For textureGatherOffset, we split the operation into up to 4 operations, one for each component
-            // that is accessed, where each textureGather operation has a different offset for each pixel.
-            if (hasInvalidOffset && isGather && !isShadow)
             {
                 config.SetUsedFeature(FeatureFlags.IntegerSampling);
 
@@ -557,7 +789,7 @@ namespace Ryujinx.Graphics.Shader.Translation
 
                 sources.CopyTo(newSources, 0);
 
-                Operand[] texSizes = InsertTextureSize(node, texOp, lodSources, bindlessHandle, coordsCount);
+                Operand[] texSizes = InsertTextureLod(node, texOp, lodSources, bindlessHandle, coordsCount);
 
                 int destIndex = 0;
 
@@ -576,7 +808,11 @@ namespace Ryujinx.Graphics.Shader.Translation
 
                         Operand intOffset = offsets[index + (hasOffsets ? compIndex * coordsCount : 0)];
 
-                        node.List.AddBefore(node, new Operation(Instruction.FP32 | Instruction.Divide, offset, Float(intOffset), Float(texSizes[index])));
+                        node.List.AddBefore(node, new Operation(
+                            Instruction.FP32 | Instruction.Divide,
+                            offset,
+                            GenerateI2f(node, intOffset),
+                            GenerateI2f(node, texSizes[index])));
 
                         Operand source = sources[coordsIndex + index];
 
@@ -603,45 +839,46 @@ namespace Ryujinx.Graphics.Shader.Translation
             }
             else
             {
-                if (hasInvalidOffset)
+                if (intCoords)
                 {
-                    if (intCoords)
+                    for (int index = 0; index < coordsCount; index++)
                     {
-                        for (int index = 0; index < coordsCount; index++)
+                        Operand source = sources[coordsIndex + index];
-                        {
-                            Operand source = sources[coordsIndex + index];
 
-                            Operand coordPlusOffset = Local();
+                        Operand coordPlusOffset = Local();
 
-                            node.List.AddBefore(node, new Operation(Instruction.Add, coordPlusOffset, source, offsets[index]));
+                        node.List.AddBefore(node, new Operation(Instruction.Add, coordPlusOffset, source, offsets[index]));
 
-                            sources[coordsIndex + index] = coordPlusOffset;
+                        sources[coordsIndex + index] = coordPlusOffset;
-                        }
                     }
-                    else
+                }
+                else
+                {
+                    config.SetUsedFeature(FeatureFlags.IntegerSampling);
+
+                    Operand[] texSizes = InsertTextureLod(node, texOp, lodSources, bindlessHandle, coordsCount);
+
+                    for (int index = 0; index < coordsCount; index++)
                     {
-                        config.SetUsedFeature(FeatureFlags.IntegerSampling);
+                        config.SetUsedTexture(Instruction.TextureSize, texOp.Type, texOp.Format, texOp.Flags, texOp.CbufSlot, texOp.Handle);
 
-                        Operand[] texSizes = InsertTextureSize(node, texOp, lodSources, bindlessHandle, coordsCount);
+                        Operand offset = Local();
 
-                        for (int index = 0; index < coordsCount; index++)
+                        Operand intOffset = offsets[index];
-                        {
-                            config.SetUsedTexture(Instruction.TextureSize, texOp.Type, texOp.Format, texOp.Flags, texOp.CbufSlot, texOp.Handle);
 
-                            Operand offset = Local();
+                        node.List.AddBefore(node, new Operation(
+                            Instruction.FP32 | Instruction.Divide,
+                            offset,
+                            GenerateI2f(node, intOffset),
+                            GenerateI2f(node, texSizes[index])));
 
-                            Operand intOffset = offsets[index];
+                        Operand source = sources[coordsIndex + index];
 
-                            node.List.AddBefore(node, new Operation(Instruction.FP32 | Instruction.Divide, offset, Float(intOffset), Float(texSizes[index])));
+                        Operand coordPlusOffset = Local();
 
-                            Operand source = sources[coordsIndex + index];
+                        node.List.AddBefore(node, new Operation(Instruction.FP32 | Instruction.Add, coordPlusOffset, source, offset));
 
-                            Operand coordPlusOffset = Local();
+                        sources[coordsIndex + index] = coordPlusOffset;
-
-                            node.List.AddBefore(node, new Operation(Instruction.FP32 | Instruction.Add, coordPlusOffset, source, offset));
-
-                            sources[coordsIndex + index] = coordPlusOffset;
-                        }
                     }
                 }
 
@@ -669,22 +906,13 @@ namespace Ryujinx.Graphics.Shader.Translation
             return node;
         }
 
-        private static Operand[] InsertTextureSize(
+        private static Operand[] InsertTextureLod(
             LinkedListNode<INode> node,
             TextureOperation texOp,
             Operand[] lodSources,
             Operand bindlessHandle,
             int coordsCount)
         {
-            Operand Int(Operand value)
-            {
-                Operand res = Local();
-
-                node.List.AddBefore(node, new Operation(Instruction.ConvertFP32ToS32, res, value));
-
-                return res;
-            }
-
             Operand[] texSizes = new Operand[coordsCount];
 
             Operand lod = Local();
@@ -708,11 +936,11 @@ namespace Ryujinx.Graphics.Shader.Translation
 
                 if (bindlessHandle != null)
                 {
-                    texSizeSources = new Operand[] { bindlessHandle, Int(lod) };
+                    texSizeSources = new Operand[] { bindlessHandle, GenerateF2i(node, lod) };
                 }
                 else
                 {
-                    texSizeSources = new Operand[] { Int(lod) };
+                    texSizeSources = new Operand[] { GenerateF2i(node, lod) };
                 }
 
                 node.List.AddBefore(node, new TextureOperation(
@@ -796,6 +1024,24 @@ namespace Ryujinx.Graphics.Shader.Translation
             return node;
         }
 
+        private static Operand GenerateI2f(LinkedListNode<INode> node, Operand value)
+        {
+            Operand res = Local();
+
+            node.List.AddBefore(node, new Operation(Instruction.ConvertS32ToFP32, res, value));
+
+            return res;
+        }
+
+        private static Operand GenerateF2i(LinkedListNode<INode> node, Operand value)
+        {
+            Operand res = Local();
+
+            node.List.AddBefore(node, new Operation(Instruction.ConvertFP32ToS32, res, value));
+
+            return res;
+        }
+
         private static bool ReplaceConstantBufferWithDrawParameters(LinkedListNode<INode> node, Operation operation)
         {
             Operand GenerateLoad(IoVariable ioVariable)

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

@@ -860,7 +860,7 @@ namespace Ryujinx.Graphics.Shader.Translation
             return descriptors;
         }
 
-        public (TextureDescriptor, int) FindTextureDescriptor(AstTextureOperation texOp)
+        public TextureDescriptor FindTextureDescriptor(AstTextureOperation texOp)
         {
             TextureDescriptor[] descriptors = GetTextureDescriptors();
 
@@ -872,11 +872,11 @@ namespace Ryujinx.Graphics.Shader.Translation
                     descriptor.HandleIndex == texOp.Handle &&
                     descriptor.Format == texOp.Format)
                 {
-                    return (descriptor, i);
+                    return descriptor;
                 }
             }
 
-            return (default, -1);
+            return default;
         }
 
         private static int FindDescriptorIndex(TextureDescriptor[] array, AstTextureOperation texOp)
@@ -897,12 +897,30 @@ namespace Ryujinx.Graphics.Shader.Translation
             return -1;
         }
 
-        public int FindTextureDescriptorIndex(AstTextureOperation texOp)
+        private static int FindDescriptorIndex(TextureDescriptor[] array, TextureOperation texOp, bool ignoreType = false)
         {
-            return FindDescriptorIndex(GetTextureDescriptors(), texOp);
+            for (int i = 0; i < array.Length; i++)
+            {
+                var descriptor = array[i];
+
+                if ((descriptor.Type == texOp.Type || ignoreType) &&
+                    descriptor.CbufSlot == texOp.CbufSlot &&
+                    descriptor.HandleIndex == texOp.Handle &&
+                    descriptor.Format == texOp.Format)
+                {
+                    return i;
+                }
+            }
+
+            return -1;
         }
 
-        public int FindImageDescriptorIndex(AstTextureOperation texOp)
+        public int FindTextureDescriptorIndex(TextureOperation texOp, bool ignoreType = false)
+        {
+            return FindDescriptorIndex(GetTextureDescriptors(), texOp, ignoreType);
+        }
+
+        public int FindImageDescriptorIndex(TextureOperation texOp)
         {
             return FindDescriptorIndex(GetImageDescriptors(), texOp);
         }

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

@@ -1,11 +1,11 @@
 using Ryujinx.Graphics.Shader.IntermediateRepresentation;
-using static Ryujinx.Graphics.Shader.IntermediateRepresentation.OperandHelper;
+using System.Collections.Generic;
 
 namespace Ryujinx.Graphics.Shader.Translation
 {
     static class ShaderIdentifier
     {
-        public static ShaderIdentification Identify(Function[] functions, ShaderConfig config)
+        public static ShaderIdentification Identify(IReadOnlyList<Function> functions, ShaderConfig config)
         {
             if (config.Stage == ShaderStage.Geometry &&
                 config.GpuAccessor.QueryPrimitiveTopology() == InputTopology.Triangles &&
@@ -20,12 +20,12 @@ namespace Ryujinx.Graphics.Shader.Translation
             return ShaderIdentification.None;
         }
 
-        private static bool IsLayerPassthroughGeometryShader(Function[] functions, out int layerInputAttr)
+        private static bool IsLayerPassthroughGeometryShader(IReadOnlyList<Function> functions, out int layerInputAttr)
         {
             bool writesLayer = false;
             layerInputAttr = 0;
 
-            if (functions.Length != 1)
+            if (functions.Count != 1)
             {
                 return false;
             }

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

@@ -5,6 +5,7 @@ using Ryujinx.Graphics.Shader.IntermediateRepresentation;
 using Ryujinx.Graphics.Shader.StructuredIr;
 using Ryujinx.Graphics.Shader.Translation.Optimizations;
 using System;
+using System.Collections.Generic;
 using System.Linq;
 using static Ryujinx.Graphics.Shader.IntermediateRepresentation.OperandHelper;
 
@@ -44,7 +45,14 @@ namespace Ryujinx.Graphics.Shader.Translation
                 }
             }
 
-            Function[] funcs = new Function[functions.Length];
+            List<Function> funcs = new List<Function>(functions.Length);
+
+            for (int i = 0; i < functions.Length; i++)
+            {
+                funcs.Add(null);
+            }
+
+            HelperFunctionManager hfm = new HelperFunctionManager(funcs, config.Stage);
 
             for (int i = 0; i < functions.Length; i++)
             {
@@ -71,7 +79,7 @@ namespace Ryujinx.Graphics.Shader.Translation
                     Ssa.Rename(cfg.Blocks);
 
                     Optimizer.RunPass(cfg.Blocks, config);
-                    Rewriter.RunPass(cfg.Blocks, config);
+                    Rewriter.RunPass(hfm, cfg.Blocks, config);
                 }
 
                 funcs[i] = new Function(cfg.Blocks, $"fun{i}", false, inArgumentsCount, outArgumentsCount);

src/Ryujinx.Graphics.Vulkan/FormatCapabilities.cs

@@ -65,6 +65,13 @@ namespace Ryujinx.Graphics.Vulkan
             return (formatFeatureFlags & flags) == flags;
         }
 
+        public bool BufferFormatSupports(FormatFeatureFlags flags, VkFormat format)
+        {
+            _api.GetPhysicalDeviceFormatProperties(_physicalDevice, format, out var fp);
+
+            return (fp.BufferFeatures & flags) == flags;
+        }
+
         public bool OptimalFormatSupports(FormatFeatureFlags flags, GAL.Format format)
         {
             var formatFeatureFlags = _optimalTable[(int)format];

src/Ryujinx.Graphics.Vulkan/FormatTable.cs

@@ -168,5 +168,223 @@ namespace Ryujinx.Graphics.Vulkan
         {
             return _table[(int)format];
         }
+
+        public static int GetAttributeFormatSize(VkFormat format)
+        {
+            switch (format)
+            {
+                case VkFormat.R8Unorm:
+                case VkFormat.R8SNorm:
+                case VkFormat.R8Uint:
+                case VkFormat.R8Sint:
+                case VkFormat.R8Uscaled:
+                case VkFormat.R8Sscaled:
+                    return 1;
+
+                case VkFormat.R8G8Unorm:
+                case VkFormat.R8G8SNorm:
+                case VkFormat.R8G8Uint:
+                case VkFormat.R8G8Sint:
+                case VkFormat.R8G8Uscaled:
+                case VkFormat.R8G8Sscaled:
+                case VkFormat.R16Sfloat:
+                case VkFormat.R16Unorm:
+                case VkFormat.R16SNorm:
+                case VkFormat.R16Uint:
+                case VkFormat.R16Sint:
+                case VkFormat.R16Uscaled:
+                case VkFormat.R16Sscaled:
+                    return 2;
+
+                case VkFormat.R8G8B8Unorm:
+                case VkFormat.R8G8B8SNorm:
+                case VkFormat.R8G8B8Uint:
+                case VkFormat.R8G8B8Sint:
+                case VkFormat.R8G8B8Uscaled:
+                case VkFormat.R8G8B8Sscaled:
+                    return 3;
+
+                case VkFormat.R8G8B8A8Unorm:
+                case VkFormat.R8G8B8A8SNorm:
+                case VkFormat.R8G8B8A8Uint:
+                case VkFormat.R8G8B8A8Sint:
+                case VkFormat.R8G8B8A8Srgb:
+                case VkFormat.R8G8B8A8Uscaled:
+                case VkFormat.R8G8B8A8Sscaled:
+                case VkFormat.B8G8R8A8Unorm:
+                case VkFormat.B8G8R8A8Srgb:
+                case VkFormat.R16G16Sfloat:
+                case VkFormat.R16G16Unorm:
+                case VkFormat.R16G16SNorm:
+                case VkFormat.R16G16Uint:
+                case VkFormat.R16G16Sint:
+                case VkFormat.R16G16Uscaled:
+                case VkFormat.R16G16Sscaled:
+                case VkFormat.R32Sfloat:
+                case VkFormat.R32Uint:
+                case VkFormat.R32Sint:
+                case VkFormat.A2B10G10R10UnormPack32:
+                case VkFormat.A2B10G10R10UintPack32:
+                case VkFormat.B10G11R11UfloatPack32:
+                case VkFormat.E5B9G9R9UfloatPack32:
+                case VkFormat.A2B10G10R10SNormPack32:
+                case VkFormat.A2B10G10R10SintPack32:
+                case VkFormat.A2B10G10R10UscaledPack32:
+                case VkFormat.A2B10G10R10SscaledPack32:
+                    return 4;
+
+                case VkFormat.R16G16B16Sfloat:
+                case VkFormat.R16G16B16Unorm:
+                case VkFormat.R16G16B16SNorm:
+                case VkFormat.R16G16B16Uint:
+                case VkFormat.R16G16B16Sint:
+                case VkFormat.R16G16B16Uscaled:
+                case VkFormat.R16G16B16Sscaled:
+                    return 6;
+
+                case VkFormat.R16G16B16A16Sfloat:
+                case VkFormat.R16G16B16A16Unorm:
+                case VkFormat.R16G16B16A16SNorm:
+                case VkFormat.R16G16B16A16Uint:
+                case VkFormat.R16G16B16A16Sint:
+                case VkFormat.R16G16B16A16Uscaled:
+                case VkFormat.R16G16B16A16Sscaled:
+                case VkFormat.R32G32Sfloat:
+                case VkFormat.R32G32Uint:
+                case VkFormat.R32G32Sint:
+                    return 8;
+
+                case VkFormat.R32G32B32Sfloat:
+                case VkFormat.R32G32B32Uint:
+                case VkFormat.R32G32B32Sint:
+                    return 12;
+
+                case VkFormat.R32G32B32A32Sfloat:
+                case VkFormat.R32G32B32A32Uint:
+                case VkFormat.R32G32B32A32Sint:
+                    return 16;
+            }
+
+            return 1;
+        }
+
+        public static VkFormat DropLastComponent(VkFormat format)
+        {
+            switch (format)
+            {
+                case VkFormat.R8G8Unorm:
+                    return VkFormat.R8Unorm;
+                case VkFormat.R8G8SNorm:
+                    return VkFormat.R8SNorm;
+                case VkFormat.R8G8Uint:
+                    return VkFormat.R8Uint;
+                case VkFormat.R8G8Sint:
+                    return VkFormat.R8Sint;
+                case VkFormat.R8G8Uscaled:
+                    return VkFormat.R8Uscaled;
+                case VkFormat.R8G8Sscaled:
+                    return VkFormat.R8Sscaled;
+
+                case VkFormat.R8G8B8Unorm:
+                    return VkFormat.R8G8Unorm;
+                case VkFormat.R8G8B8SNorm:
+                    return VkFormat.R8G8SNorm;
+                case VkFormat.R8G8B8Uint:
+                    return VkFormat.R8G8Uint;
+                case VkFormat.R8G8B8Sint:
+                    return VkFormat.R8G8Sint;
+                case VkFormat.R8G8B8Uscaled:
+                    return VkFormat.R8G8Uscaled;
+                case VkFormat.R8G8B8Sscaled:
+                    return VkFormat.R8G8Sscaled;
+
+                case VkFormat.R8G8B8A8Unorm:
+                    return VkFormat.R8G8B8Unorm;
+                case VkFormat.R8G8B8A8SNorm:
+                    return VkFormat.R8G8B8SNorm;
+                case VkFormat.R8G8B8A8Uint:
+                    return VkFormat.R8G8B8Uint;
+                case VkFormat.R8G8B8A8Sint:
+                    return VkFormat.R8G8B8Sint;
+                case VkFormat.R8G8B8A8Srgb:
+                    return VkFormat.R8G8B8Srgb;
+                case VkFormat.R8G8B8A8Uscaled:
+                    return VkFormat.R8G8B8Uscaled;
+                case VkFormat.R8G8B8A8Sscaled:
+                    return VkFormat.R8G8B8Sscaled;
+                case VkFormat.B8G8R8A8Unorm:
+                    return VkFormat.B8G8R8Unorm;
+                case VkFormat.B8G8R8A8Srgb:
+                    return VkFormat.B8G8R8Srgb;
+
+                case VkFormat.R16G16Sfloat:
+                    return VkFormat.R16Sfloat;
+                case VkFormat.R16G16Unorm:
+                    return VkFormat.R16Unorm;
+                case VkFormat.R16G16SNorm:
+                    return VkFormat.R16SNorm;
+                case VkFormat.R16G16Uint:
+                    return VkFormat.R16Uint;
+                case VkFormat.R16G16Sint:
+                    return VkFormat.R16Sint;
+                case VkFormat.R16G16Uscaled:
+                    return VkFormat.R16Uscaled;
+                case VkFormat.R16G16Sscaled:
+                    return VkFormat.R16Sscaled;
+
+                case VkFormat.R16G16B16Sfloat:
+                    return VkFormat.R16G16Sfloat;
+                case VkFormat.R16G16B16Unorm:
+                    return VkFormat.R16G16Unorm;
+                case VkFormat.R16G16B16SNorm:
+                    return VkFormat.R16G16SNorm;
+                case VkFormat.R16G16B16Uint:
+                    return VkFormat.R16G16Uint;
+                case VkFormat.R16G16B16Sint:
+                    return VkFormat.R16G16Sint;
+                case VkFormat.R16G16B16Uscaled:
+                    return VkFormat.R16G16Uscaled;
+                case VkFormat.R16G16B16Sscaled:
+                    return VkFormat.R16G16Sscaled;
+
+                case VkFormat.R16G16B16A16Sfloat:
+                    return VkFormat.R16G16B16Sfloat;
+                case VkFormat.R16G16B16A16Unorm:
+                    return VkFormat.R16G16B16Unorm;
+                case VkFormat.R16G16B16A16SNorm:
+                    return VkFormat.R16G16B16SNorm;
+                case VkFormat.R16G16B16A16Uint:
+                    return VkFormat.R16G16B16Uint;
+                case VkFormat.R16G16B16A16Sint:
+                    return VkFormat.R16G16B16Sint;
+                case VkFormat.R16G16B16A16Uscaled:
+                    return VkFormat.R16G16B16Uscaled;
+                case VkFormat.R16G16B16A16Sscaled:
+                    return VkFormat.R16G16B16Sscaled;
+
+                case VkFormat.R32G32Sfloat:
+                    return VkFormat.R32Sfloat;
+                case VkFormat.R32G32Uint:
+                    return VkFormat.R32Uint;
+                case VkFormat.R32G32Sint:
+                    return VkFormat.R32Sint;
+
+                case VkFormat.R32G32B32Sfloat:
+                    return VkFormat.R32G32Sfloat;
+                case VkFormat.R32G32B32Uint:
+                    return VkFormat.R32G32Uint;
+                case VkFormat.R32G32B32Sint:
+                    return VkFormat.R32G32Sint;
+
+                case VkFormat.R32G32B32A32Sfloat:
+                    return VkFormat.R32G32B32Sfloat;
+                case VkFormat.R32G32B32A32Uint:
+                    return VkFormat.R32G32B32Uint;
+                case VkFormat.R32G32B32A32Sint:
+                    return VkFormat.R32G32B32Sint;
+            }
+
+            return format;
+        }
     }
 }

src/Ryujinx.Graphics.Vulkan/FramebufferParams.cs

@@ -21,6 +21,7 @@ namespace Ryujinx.Graphics.Vulkan
         public uint[] AttachmentSamples { get; }
         public VkFormat[] AttachmentFormats { get; }
         public int[] AttachmentIndices { get; }
+        public uint AttachmentIntegerFormatMask { get; }
 
         public int AttachmentsCount { get; }
         public int MaxColorAttachmentIndex => AttachmentIndices.Length > 0 ? AttachmentIndices[AttachmentIndices.Length - 1] : -1;
@@ -74,6 +75,7 @@ namespace Ryujinx.Graphics.Vulkan
 
             int index = 0;
             int bindIndex = 0;
+            uint attachmentIntegerFormatMask = 0;
 
             foreach (ITexture color in colors)
             {
@@ -89,6 +91,11 @@ namespace Ryujinx.Graphics.Vulkan
                     AttachmentFormats[index] = texture.VkFormat;
                     AttachmentIndices[index] = bindIndex;
 
+                    if (texture.Info.Format.IsInteger())
+                    {
+                        attachmentIntegerFormatMask |= 1u << bindIndex;
+                    }
+
                     width = Math.Min(width, (uint)texture.Width);
                     height = Math.Min(height, (uint)texture.Height);
                     layers = Math.Min(layers, (uint)texture.Layers);
@@ -102,6 +109,8 @@ namespace Ryujinx.Graphics.Vulkan
                 bindIndex++;
             }
 
+            AttachmentIntegerFormatMask = attachmentIntegerFormatMask;
+
             if (depthStencil is TextureView dsTexture && dsTexture.Valid)
             {
                 _attachments[count - 1] = dsTexture.GetImageViewForAttachment();

src/Ryujinx.Graphics.Vulkan/HardwareCapabilities.cs

@@ -28,6 +28,7 @@ namespace Ryujinx.Graphics.Vulkan
         public readonly bool SupportsSubgroupSizeControl;
         public readonly bool SupportsShaderInt8;
         public readonly bool SupportsShaderStencilExport;
+        public readonly bool SupportsShaderStorageImageMultisample;
         public readonly bool SupportsConditionalRendering;
         public readonly bool SupportsExtendedDynamicState;
         public readonly bool SupportsMultiView;
@@ -42,6 +43,7 @@ namespace Ryujinx.Graphics.Vulkan
         public readonly bool SupportsGeometryShader;
         public readonly bool SupportsViewportArray2;
         public readonly bool SupportsHostImportedMemory;
+        public readonly bool SupportsDepthClipControl;
         public readonly uint MinSubgroupSize;
         public readonly uint MaxSubgroupSize;
         public readonly ShaderStageFlags RequiredSubgroupSizeStages;
@@ -63,6 +65,7 @@ namespace Ryujinx.Graphics.Vulkan
             bool supportsSubgroupSizeControl,
             bool supportsShaderInt8,
             bool supportsShaderStencilExport,
+            bool supportsShaderStorageImageMultisample,
             bool supportsConditionalRendering,
             bool supportsExtendedDynamicState,
             bool supportsMultiView,
@@ -77,6 +80,7 @@ namespace Ryujinx.Graphics.Vulkan
             bool supportsGeometryShader,
             bool supportsViewportArray2,
             bool supportsHostImportedMemory,
+            bool supportsDepthClipControl,
             uint minSubgroupSize,
             uint maxSubgroupSize,
             ShaderStageFlags requiredSubgroupSizeStages,
@@ -97,6 +101,7 @@ namespace Ryujinx.Graphics.Vulkan
             SupportsSubgroupSizeControl = supportsSubgroupSizeControl;
             SupportsShaderInt8 = supportsShaderInt8;
             SupportsShaderStencilExport = supportsShaderStencilExport;
+            SupportsShaderStorageImageMultisample = supportsShaderStorageImageMultisample;
             SupportsConditionalRendering = supportsConditionalRendering;
             SupportsExtendedDynamicState = supportsExtendedDynamicState;
             SupportsMultiView = supportsMultiView;
@@ -111,6 +116,7 @@ namespace Ryujinx.Graphics.Vulkan
             SupportsGeometryShader = supportsGeometryShader;
             SupportsViewportArray2 = supportsViewportArray2;
             SupportsHostImportedMemory = supportsHostImportedMemory;
+            SupportsDepthClipControl = supportsDepthClipControl;
             MinSubgroupSize = minSubgroupSize;
             MaxSubgroupSize = maxSubgroupSize;
             RequiredSubgroupSizeStages = requiredSubgroupSizeStages;

src/Ryujinx.Graphics.Vulkan/PipelineBase.cs

@@ -80,6 +80,7 @@ namespace Ryujinx.Graphics.Vulkan
 
         private PipelineColorBlendAttachmentState[] _storedBlend;
 
+        private ulong _drawCountSinceBarrier;
         public ulong DrawCount { get; private set; }
         public bool RenderPassActive { get; private set; }
 
@@ -133,6 +134,18 @@ namespace Ryujinx.Graphics.Vulkan
 
         public unsafe void Barrier()
         {
+            if (_drawCountSinceBarrier != DrawCount)
+            {
+                _drawCountSinceBarrier = DrawCount;
+
+                // Barriers apparently have no effect inside a render pass on MoltenVK.
+                // As a workaround, end the render pass.
+                if (Gd.IsMoltenVk)
+                {
+                    EndRenderPass();
+                }
+            }
+
             MemoryBarrier memoryBarrier = new MemoryBarrier()
             {
                 SType = StructureType.MemoryBarrier,
@@ -551,7 +564,6 @@ namespace Ryujinx.Graphics.Vulkan
                         (uint)maxDrawCount,
                         (uint)stride);
                 }
-
             }
             else
             {
@@ -813,8 +825,12 @@ namespace Ryujinx.Graphics.Vulkan
 
         public void SetDepthMode(DepthMode mode)
         {
-            // Currently this is emulated on the shader, because Vulkan had no support for changing the depth mode.
+            bool oldMode = _newState.DepthMode;
-            // In the future, we may want to use the VK_EXT_depth_clip_control extension to change it here.
+            _newState.DepthMode = mode == DepthMode.MinusOneToOne;
+            if (_newState.DepthMode != oldMode)
+            {
+                SignalStateChange();
+            }
         }
 
         public void SetDepthTest(DepthTestDescriptor depthTest)
@@ -1471,6 +1487,7 @@ namespace Ryujinx.Graphics.Vulkan
         {
             var dstAttachmentFormats = _newState.Internal.AttachmentFormats.AsSpan();
             FramebufferParams.AttachmentFormats.CopyTo(dstAttachmentFormats);
+            _newState.Internal.AttachmentIntegerFormatMask = FramebufferParams.AttachmentIntegerFormatMask;
 
             for (int i = FramebufferParams.AttachmentFormats.Length; i < dstAttachmentFormats.Length; i++)
             {

src/Ryujinx.Graphics.Vulkan/PipelineConverter.cs

@@ -294,6 +294,7 @@ namespace Ryujinx.Graphics.Vulkan
 
             int attachmentCount = 0;
             int maxColorAttachmentIndex = -1;
+            uint attachmentIntegerFormatMask = 0;
 
             for (int i = 0; i < Constants.MaxRenderTargets; i++)
             {
@@ -301,6 +302,11 @@ namespace Ryujinx.Graphics.Vulkan
                 {
                     pipeline.Internal.AttachmentFormats[attachmentCount++] = gd.FormatCapabilities.ConvertToVkFormat(state.AttachmentFormats[i]);
                     maxColorAttachmentIndex = i;
+
+                    if (state.AttachmentFormats[i].IsInteger())
+                    {
+                        attachmentIntegerFormatMask |= 1u << i;
+                    }
                 }
             }
 
@@ -311,6 +317,7 @@ namespace Ryujinx.Graphics.Vulkan
 
             pipeline.ColorBlendAttachmentStateCount = (uint)(maxColorAttachmentIndex + 1);
             pipeline.VertexAttributeDescriptionsCount = (uint)Math.Min(Constants.MaxVertexAttributes, state.VertexAttribCount);
+            pipeline.Internal.AttachmentIntegerFormatMask = attachmentIntegerFormatMask;
 
             return pipeline;
         }

src/Ryujinx.Graphics.Vulkan/PipelineState.cs

@@ -1,5 +1,7 @@
-using Silk.NET.Vulkan;
+using Ryujinx.Common.Memory;
+using Silk.NET.Vulkan;
 using System;
+using System.Numerics;
 
 namespace Ryujinx.Graphics.Vulkan
 {
@@ -303,11 +305,19 @@ namespace Ryujinx.Graphics.Vulkan
             set => Internal.Id9 = (Internal.Id9 & 0xFFFFFFFFFFFFFFCF) | ((ulong)value << 4);
         }
 
+        public bool DepthMode
+        {
+            get => ((Internal.Id9 >> 6) & 0x1) != 0UL;
+            set => Internal.Id9 = (Internal.Id9 & 0xFFFFFFFFFFFFFFBF) | ((value ? 1UL : 0UL) << 6);
+        }
+
         public NativeArray<PipelineShaderStageCreateInfo> Stages;
         public NativeArray<PipelineShaderStageRequiredSubgroupSizeCreateInfoEXT> StageRequiredSubgroupSizes;
         public PipelineLayout PipelineLayout;
         public SpecData SpecializationData;
 
+        private Array32<VertexInputAttributeDescription> _vertexAttributeDescriptions2;
+
         public void Initialize()
         {
             Stages = new NativeArray<PipelineShaderStageCreateInfo>(Constants.MaxShaderStages);
@@ -328,6 +338,7 @@ namespace Ryujinx.Graphics.Vulkan
 
             LineWidth = 1f;
             SamplesCount = 1;
+            DepthMode = true;
         }
 
         public unsafe Auto<DisposablePipeline> CreateComputePipeline(
@@ -400,7 +411,15 @@ namespace Ryujinx.Graphics.Vulkan
 
             Pipeline pipelineHandle = default;
 
+            bool isMoltenVk = gd.IsMoltenVk;
+
+            if (isMoltenVk)
+            {
+                UpdateVertexAttributeDescriptions(gd);
+            }
+
             fixed (VertexInputAttributeDescription* pVertexAttributeDescriptions = &Internal.VertexAttributeDescriptions[0])
+            fixed (VertexInputAttributeDescription* pVertexAttributeDescriptions2 = &_vertexAttributeDescriptions2[0])
             fixed (VertexInputBindingDescription* pVertexBindingDescriptions = &Internal.VertexBindingDescriptions[0])
             fixed (Viewport* pViewports = &Internal.Viewports[0])
             fixed (Rect2D* pScissors = &Internal.Scissors[0])
@@ -410,7 +429,7 @@ namespace Ryujinx.Graphics.Vulkan
                 {
                     SType = StructureType.PipelineVertexInputStateCreateInfo,
                     VertexAttributeDescriptionCount = VertexAttributeDescriptionsCount,
-                    PVertexAttributeDescriptions = pVertexAttributeDescriptions,
+                    PVertexAttributeDescriptions = isMoltenVk ? pVertexAttributeDescriptions2 : pVertexAttributeDescriptions,
                     VertexBindingDescriptionCount = VertexBindingDescriptionsCount,
                     PVertexBindingDescriptions = pVertexBindingDescriptions
                 };
@@ -471,6 +490,17 @@ namespace Ryujinx.Graphics.Vulkan
                     PScissors = pScissors
                 };
 
+                if (gd.Capabilities.SupportsDepthClipControl)
+                {
+                    var viewportDepthClipControlState = new PipelineViewportDepthClipControlCreateInfoEXT()
+                    {
+                        SType = StructureType.PipelineViewportDepthClipControlCreateInfoExt,
+                        NegativeOneToOne = DepthMode
+                    };
+
+                    viewportState.PNext = &viewportDepthClipControlState;
+                }
+
                 var multisampleState = new PipelineMultisampleStateCreateInfo
                 {
                     SType = StructureType.PipelineMultisampleStateCreateInfo,
@@ -513,6 +543,27 @@ namespace Ryujinx.Graphics.Vulkan
                     MaxDepthBounds = MaxDepthBounds
                 };
 
+                uint blendEnables = 0;
+
+                if (gd.IsMoltenVk && Internal.AttachmentIntegerFormatMask != 0)
+                {
+                    // Blend can't be enabled for integer formats, so let's make sure it is disabled.
+                    uint attachmentIntegerFormatMask = Internal.AttachmentIntegerFormatMask;
+
+                    while (attachmentIntegerFormatMask != 0)
+                    {
+                        int i = BitOperations.TrailingZeroCount(attachmentIntegerFormatMask);
+
+                        if (Internal.ColorBlendAttachmentState[i].BlendEnable)
+                        {
+                            blendEnables |= 1u << i;
+                        }
+
+                        Internal.ColorBlendAttachmentState[i].BlendEnable = false;
+                        attachmentIntegerFormatMask &= ~(1u << i);
+                    }
+                }
+
                 var colorBlendState = new PipelineColorBlendStateCreateInfo()
                 {
                     SType = StructureType.PipelineColorBlendStateCreateInfo,
@@ -590,6 +641,15 @@ namespace Ryujinx.Graphics.Vulkan
                 };
 
                 gd.Api.CreateGraphicsPipelines(device, cache, 1, &pipelineCreateInfo, null, &pipelineHandle).ThrowOnError();
+
+                // Restore previous blend enable values if we changed it.
+                while (blendEnables != 0)
+                {
+                    int i = BitOperations.TrailingZeroCount(blendEnables);
+
+                    Internal.ColorBlendAttachmentState[i].BlendEnable = true;
+                    blendEnables &= ~(1u << i);
+                }
             }
 
             pipeline = new Auto<DisposablePipeline>(new DisposablePipeline(gd.Api, device, pipelineHandle));
@@ -612,6 +672,62 @@ namespace Ryujinx.Graphics.Vulkan
             }
         }
 
+        private void UpdateVertexAttributeDescriptions(VulkanRenderer gd)
+        {
+            // Vertex attributes exceeding the stride are invalid.
+            // In metal, they cause glitches with the vertex shader fetching incorrect values.
+            // To work around this, we reduce the format to something that doesn't exceed the stride if possible.
+            // The assumption is that the exceeding components are not actually accessed on the shader.
+
+            for (int index = 0; index < VertexAttributeDescriptionsCount; index++)
+            {
+                var attribute = Internal.VertexAttributeDescriptions[index];
+                int vbIndex = GetVertexBufferIndex(attribute.Binding);
+
+                if (vbIndex >= 0)
+                {
+                    ref var vb = ref Internal.VertexBindingDescriptions[vbIndex];
+
+                    Format format = attribute.Format;
+
+                    while (vb.Stride != 0 && attribute.Offset + FormatTable.GetAttributeFormatSize(format) > vb.Stride)
+                    {
+                        Format newFormat = FormatTable.DropLastComponent(format);
+
+                        if (newFormat == format)
+                        {
+                            // That case means we failed to find a format that fits within the stride,
+                            // so just restore the original format and give up.
+                            format = attribute.Format;
+                            break;
+                        }
+
+                        format = newFormat;
+                    }
+
+                    if (attribute.Format != format && gd.FormatCapabilities.BufferFormatSupports(FormatFeatureFlags.VertexBufferBit, format))
+                    {
+                        attribute.Format = format;
+                    }
+                }
+
+                _vertexAttributeDescriptions2[index] = attribute;
+            }
+        }
+
+        private int GetVertexBufferIndex(uint binding)
+        {
+            for (int index = 0; index < VertexBindingDescriptionsCount; index++)
+            {
+                if (Internal.VertexBindingDescriptions[index].Binding == binding)
+                {
+                    return index;
+                }
+            }
+
+            return -1;
+        }
+
         public void Dispose()
         {
             Stages.Dispose();

src/Ryujinx.Graphics.Vulkan/PipelineUid.cs

@@ -35,6 +35,7 @@ namespace Ryujinx.Graphics.Vulkan
         public Array16<Rect2D> Scissors;
         public Array8<PipelineColorBlendAttachmentState> ColorBlendAttachmentState;
         public Array9<Format> AttachmentFormats;
+        public uint AttachmentIntegerFormatMask;
 
         public override bool Equals(object obj)
         {

src/Ryujinx.Graphics.Vulkan/TextureStorage.cs

@@ -80,7 +80,7 @@ namespace Ryujinx.Graphics.Vulkan
 
             var sampleCountFlags = ConvertToSampleCountFlags(gd.Capabilities.SupportedSampleCounts, (uint)info.Samples);
 
-            var usage = GetImageUsageFromFormat(info.Format);
+            var usage = GetImageUsage(info.Format, info.Target, gd.Capabilities.SupportsShaderStorageImageMultisample);
 
             var flags = ImageCreateFlags.CreateMutableFormatBit;
 
@@ -293,7 +293,7 @@ namespace Ryujinx.Graphics.Vulkan
             }
         }
 
-        public static ImageUsageFlags GetImageUsageFromFormat(GAL.Format format)
+        public static ImageUsageFlags GetImageUsage(GAL.Format format, Target target, bool supportsMsStorage)
         {
             var usage = DefaultUsageFlags;
 
@@ -306,7 +306,7 @@ namespace Ryujinx.Graphics.Vulkan
                 usage |= ImageUsageFlags.ColorAttachmentBit;
             }
 
-            if (format.IsImageCompatible())
+            if (format.IsImageCompatible() && (supportsMsStorage || !target.IsMultisample()))
             {
                 usage |= ImageUsageFlags.StorageBit;
             }

src/Ryujinx.Graphics.Vulkan/TextureView.cs

@@ -54,7 +54,7 @@ namespace Ryujinx.Graphics.Vulkan
             gd.Textures.Add(this);
 
             var format = _gd.FormatCapabilities.ConvertToVkFormat(info.Format);
-            var usage = TextureStorage.GetImageUsageFromFormat(info.Format);
+            var usage = TextureStorage.GetImageUsage(info.Format, info.Target, gd.Capabilities.SupportsShaderStorageImageMultisample);
             var levels = (uint)info.Levels;
             var layers = (uint)info.GetLayers();
 

src/Ryujinx.Graphics.Vulkan/VulkanInitialization.cs

@@ -41,6 +41,7 @@ namespace Ryujinx.Graphics.Vulkan
             "VK_EXT_subgroup_size_control",
             "VK_NV_geometry_shader_passthrough",
             "VK_NV_viewport_array2",
+            "VK_EXT_depth_clip_control",
             "VK_KHR_portability_subset" // As per spec, we should enable this if present.
         };
 
@@ -345,6 +346,17 @@ namespace Ryujinx.Graphics.Vulkan
                 features2.PNext = &supportedFeaturesRobustness2;
             }
 
+            PhysicalDeviceDepthClipControlFeaturesEXT supportedFeaturesDepthClipControl = new PhysicalDeviceDepthClipControlFeaturesEXT()
+            {
+                SType = StructureType.PhysicalDeviceDepthClipControlFeaturesExt,
+                PNext = features2.PNext
+            };
+
+            if (physicalDevice.IsDeviceExtensionPresent("VK_EXT_depth_clip_control"))
+            {
+                features2.PNext = &supportedFeaturesDepthClipControl;
+            }
+
             api.GetPhysicalDeviceFeatures2(physicalDevice.PhysicalDevice, &features2);
 
             var supportedFeatures = features2.Features;
@@ -507,6 +519,21 @@ namespace Ryujinx.Graphics.Vulkan
                 pExtendedFeatures = &featuresCustomBorderColor;
             }
 
+            PhysicalDeviceDepthClipControlFeaturesEXT featuresDepthClipControl;
+
+            if (physicalDevice.IsDeviceExtensionPresent("VK_EXT_depth_clip_control") &&
+                supportedFeaturesDepthClipControl.DepthClipControl)
+            {
+                featuresDepthClipControl = new PhysicalDeviceDepthClipControlFeaturesEXT()
+                {
+                    SType = StructureType.PhysicalDeviceDepthClipControlFeaturesExt,
+                    PNext = pExtendedFeatures,
+                    DepthClipControl = true
+                };
+
+                pExtendedFeatures = &featuresDepthClipControl;
+            }
+
             var enabledExtensions = _requiredExtensions.Union(_desirableExtensions.Intersect(physicalDevice.DeviceExtensions)).ToArray();
 
             IntPtr* ppEnabledExtensions = stackalloc IntPtr[enabledExtensions.Length];

src/Ryujinx.Graphics.Vulkan/VulkanRenderer.cs

@@ -216,6 +216,11 @@ namespace Ryujinx.Graphics.Vulkan
                 SType = StructureType.PhysicalDeviceCustomBorderColorFeaturesExt
             };
 
+            PhysicalDeviceDepthClipControlFeaturesEXT featuresDepthClipControl = new PhysicalDeviceDepthClipControlFeaturesEXT()
+            {
+                SType = StructureType.PhysicalDeviceDepthClipControlFeaturesExt
+            };
+
             PhysicalDevicePortabilitySubsetFeaturesKHR featuresPortabilitySubset = new PhysicalDevicePortabilitySubsetFeaturesKHR()
             {
                 SType = StructureType.PhysicalDevicePortabilitySubsetFeaturesKhr
@@ -244,6 +249,14 @@ namespace Ryujinx.Graphics.Vulkan
                 features2.PNext = &featuresCustomBorderColor;
             }
 
+            bool supportsDepthClipControl = _physicalDevice.IsDeviceExtensionPresent("VK_EXT_depth_clip_control");
+
+            if (supportsDepthClipControl)
+            {
+                featuresDepthClipControl.PNext = features2.PNext;
+                features2.PNext = &featuresDepthClipControl;
+            }
+
             bool usePortability = _physicalDevice.IsDeviceExtensionPresent("VK_KHR_portability_subset");
 
             if (usePortability)
@@ -295,6 +308,7 @@ namespace Ryujinx.Graphics.Vulkan
                 supportsSubgroupSizeControl,
                 featuresShaderInt8.ShaderInt8,
                 _physicalDevice.IsDeviceExtensionPresent("VK_EXT_shader_stencil_export"),
+                features2.Features.ShaderStorageImageMultisample,
                 _physicalDevice.IsDeviceExtensionPresent(ExtConditionalRendering.ExtensionName),
                 _physicalDevice.IsDeviceExtensionPresent(ExtExtendedDynamicState.ExtensionName),
                 features2.Features.MultiViewport,
@@ -309,6 +323,7 @@ namespace Ryujinx.Graphics.Vulkan
                 _physicalDevice.PhysicalDeviceFeatures.GeometryShader,
                 _physicalDevice.IsDeviceExtensionPresent("VK_NV_viewport_array2"),
                 _physicalDevice.IsDeviceExtensionPresent(ExtExternalMemoryHost.ExtensionName),
+                supportsDepthClipControl && featuresDepthClipControl.DepthClipControl,
                 propertiesSubgroupSizeControl.MinSubgroupSize,
                 propertiesSubgroupSizeControl.MaxSubgroupSize,
                 propertiesSubgroupSizeControl.RequiredSubgroupSizeStages,
@@ -584,6 +599,7 @@ namespace Ryujinx.Graphics.Vulkan
                 supportsViewportMask: Capabilities.SupportsViewportArray2,
                 supportsViewportSwizzle: false,
                 supportsIndirectParameters: true,
+                supportsDepthClipControl: Capabilities.SupportsDepthClipControl,
                 maximumUniformBuffersPerStage: Constants.MaxUniformBuffersPerStage,
                 maximumStorageBuffersPerStage: Constants.MaxStorageBuffersPerStage,
                 maximumTexturesPerStage: Constants.MaxTexturesPerStage,

src/Ryujinx.HLE/HOS/ModLoader.cs

@@ -167,12 +167,12 @@ namespace Ryujinx.HLE.HOS
 
                 if (StrEquals(RomfsDir, modDir.Name))
                 {
-                    mods.RomfsDirs.Add(mod = new Mod<DirectoryInfo>($"<{titleId} RomFs>", modDir));
+                    mods.RomfsDirs.Add(mod = new Mod<DirectoryInfo>(dir.Name, modDir));
                     types.Append('R');
                 }
                 else if (StrEquals(ExefsDir, modDir.Name))
                 {
-                    mods.ExefsDirs.Add(mod = new Mod<DirectoryInfo>($"<{titleId} ExeFs>", modDir));
+                    mods.ExefsDirs.Add(mod = new Mod<DirectoryInfo>(dir.Name, modDir));
                     types.Append('E');
                 }
                 else if (StrEquals(CheatDir, modDir.Name))

src/Ryujinx/Ui/MainWindow.cs

@@ -1024,6 +1024,8 @@ namespace Ryujinx.Ui
                         double sessionTimePlayed = DateTime.UtcNow.Subtract(appMetadata.LastPlayed.Value).TotalSeconds;
                         appMetadata.TimePlayed += Math.Round(sessionTimePlayed, MidpointRounding.AwayFromZero);
                     }
+
+                    appMetadata.LastPlayed = DateTime.UtcNow;
                 });
             }
         }

src/Ryujinx/Ui/RendererWidgetBase.cs

@@ -381,7 +381,7 @@ namespace Ryujinx.Ui
                         string filename    = $"ryujinx_capture_{currentTime.Year}-{currentTime.Month:D2}-{currentTime.Day:D2}_{currentTime.Hour:D2}-{currentTime.Minute:D2}-{currentTime.Second:D2}.png";
                         string directory   = AppDataManager.Mode switch
                         {
-                            AppDataManager.LaunchMode.Portable => System.IO.Path.Combine(AppDataManager.BaseDirPath, "screenshots"),
+                            AppDataManager.LaunchMode.Portable or AppDataManager.LaunchMode.Custom => System.IO.Path.Combine(AppDataManager.BaseDirPath, "screenshots"),
                             _ => System.IO.Path.Combine(Environment.GetFolderPath(Environment.SpecialFolder.MyPictures), "Ryujinx")
                         };
 

src/Ryujinx/Ui/Windows/AboutWindow.Designer.cs

@@ -48,6 +48,8 @@ namespace Ryujinx.Ui.Windows
         private Label          _patreonNamesLabel;
         private ScrolledWindow _patreonNamesScrolled;
         private TextView       _patreonNamesText;
+        private EventBox       _changelogEventBox;
+        private Label          _changelogLinkLabel;
 
         private void InitializeComponent()
         {
@@ -148,6 +150,23 @@ namespace Ryujinx.Ui.Windows
                 Margin  = 5
             };
 
+            //
+            // _changelogEventBox
+            //
+            _changelogEventBox = new EventBox();
+            _changelogEventBox.ButtonPressEvent += ChangelogButton_Pressed;
+
+            //
+            // _changelogLinkLabel
+            //
+            _changelogLinkLabel = new Label("View Changelog on GitHub")
+            {
+                TooltipText = "Click to open the changelog for this version in your default browser.",
+                Justify = Justification.Center,
+                Attributes  = new AttrList()
+            };
+            _changelogLinkLabel.Attributes.Insert(new Pango.AttrUnderline(Underline.Single));
+
             //
             // _disclaimerLabel
             //
@@ -464,8 +483,11 @@ namespace Ryujinx.Ui.Windows
             _socialBox.Add(_discordEventBox);
             _socialBox.Add(_twitterEventBox);
 
+            _changelogEventBox.Add(_changelogLinkLabel);
+
             _leftBox.Add(_logoBox);
             _leftBox.Add(_versionLabel);
+            _leftBox.Add(_changelogEventBox);
             _leftBox.Add(_disclaimerLabel);
             _leftBox.Add(_amiiboApiLink);
             _leftBox.Add(_socialBox);

src/Ryujinx/Ui/Windows/AboutWindow.cs

@@ -76,5 +76,10 @@ namespace Ryujinx.Ui.Windows
         {
             OpenHelper.OpenUrl("https://github.com/Ryujinx/Ryujinx/graphs/contributors?type=a");
         }
+
+        private void ChangelogButton_Pressed(object sender, ButtonPressEventArgs args)
+        {
+            OpenHelper.OpenUrl("https://github.com/Ryujinx/Ryujinx/wiki/Changelog#ryujinx-changelog");
+        }
     }
 }