Remove shader dependency on SPV_KHR_shader_ballot and SPV_KHR_subgroup_vote extensions (#3943)

* Remove shader dependency on SPV_KHR_shader_ballot and SPV_KHR_subgroup_vote extensions

* Shader cache version bump

README.md

@@ -36,7 +36,7 @@
 
 ## Compatibility
 
-As of October 2022, Ryujinx has been tested on approximately 3,700 titles; over 3,500 boot past menus and into gameplay, with roughly 3,000 of those being considered playable.
+As of November 2022, Ryujinx has been tested on approximately 3,800 titles; over 3,600 boot past menus and into gameplay, with roughly 3,200 of those being considered playable.
 You can check out the compatibility list [here](https://github.com/Ryujinx/Ryujinx-Games-List/issues). Anyone is free to submit a new game test or update an existing game test entry; simply follow the new issue template and testing guidelines, or post as a reply to the applicable game issue. Use the search function to see if a game has been tested already!
 
 ## Usage
@@ -90,7 +90,7 @@ Ryujinx system files are stored in the `Ryujinx` folder. This folder is located
 
 - **GPU**
 
-  The GPU emulator emulates the Switch's Maxwell GPU using either the OpenGL (version 4.5 minimum) or Vulkan APIs through a custom build of OpenTK or Silk.NET respectively. There are currently four graphics enhancements available to the end user in Ryujinx: Disk Shader Caching, Resolution Scaling, Aspect Ratio Adjustment, and Anisotropic Filtering. These enhancements can be adjusted or toggled as desired in the GUI.
+  The GPU emulator emulates the Switch's Maxwell GPU using either the OpenGL (version 4.5 minimum), Vulkan, or Metal (via MoltenVK) APIs through a custom build of OpenTK or Silk.NET respectively. There are currently four graphics enhancements available to the end user in Ryujinx: Disk Shader Caching, Resolution Scaling, Aspect Ratio Adjustment, and Anisotropic Filtering. These enhancements can be adjusted or toggled as desired in the GUI.
 
 - **Input**
 

Ryujinx.Common/Ryujinx.Common.csproj

@@ -8,7 +8,7 @@
   <ItemGroup>
     <PackageReference Include="MsgPack.Cli" Version="1.0.1" />
     <PackageReference Include="System.Drawing.Common" Version="6.0.0" />
-    <PackageReference Include="System.Management" Version="6.0.0" />
+    <PackageReference Include="System.Management" Version="7.0.0" />
   </ItemGroup>
 
 </Project>

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

@@ -37,6 +37,7 @@ namespace Ryujinx.Graphics.Gpu.Engine.Threed
         private bool _vsUsesDrawParameters;
         private bool _vtgWritesRtLayer;
         private byte _vsClipDistancesWritten;
+        private uint _vbEnableMask;
 
         private bool _prevDrawIndexed;
         private bool _prevDrawIndirect;
@@ -76,6 +77,7 @@ namespace Ryujinx.Graphics.Gpu.Engine.Threed
                     nameof(ThreedClassState.VertexBufferState),
                     nameof(ThreedClassState.VertexBufferEndAddress)),
 
+                // Must be done after vertex buffer updates.
                 new StateUpdateCallbackEntry(UpdateVertexAttribState, nameof(ThreedClassState.VertexAttribState)),
 
                 new StateUpdateCallbackEntry(UpdateBlendState,
@@ -852,12 +854,23 @@ namespace Ryujinx.Graphics.Gpu.Engine.Threed
         /// </summary>
         private void UpdateVertexAttribState()
         {
+            uint vbEnableMask = _vbEnableMask;
+
             Span<VertexAttribDescriptor> vertexAttribs = stackalloc VertexAttribDescriptor[Constants.TotalVertexAttribs];
 
             for (int index = 0; index < Constants.TotalVertexAttribs; index++)
             {
                 var vertexAttrib = _state.State.VertexAttribState[index];
 
+                int bufferIndex = vertexAttrib.UnpackBufferIndex();
+
+                if ((vbEnableMask & (1u << bufferIndex)) == 0)
+                {
+                    // Using a vertex buffer that doesn't exist is invalid, so let's use a dummy attribute for those cases.
+                    vertexAttribs[index] = new VertexAttribDescriptor(0, 0, true, Format.R32G32B32A32Float);
+                    continue;
+                }
+
                 if (!FormatTable.TryGetAttribFormat(vertexAttrib.UnpackFormat(), out Format format))
                 {
                     Logger.Debug?.Print(LogClass.Gpu, $"Invalid attribute format 0x{vertexAttrib.UnpackFormat():X}.");
@@ -866,7 +879,7 @@ namespace Ryujinx.Graphics.Gpu.Engine.Threed
                 }
 
                 vertexAttribs[index] = new VertexAttribDescriptor(
-                    vertexAttrib.UnpackBufferIndex(),
+                    bufferIndex,
                     vertexAttrib.UnpackOffset(),
                     vertexAttrib.UnpackIsConstant(),
                     format);
@@ -954,6 +967,7 @@ namespace Ryujinx.Graphics.Gpu.Engine.Threed
 
             bool drawIndexed = _drawState.DrawIndexed;
             bool drawIndirect = _drawState.DrawIndirect;
+            uint vbEnableMask = 0;
 
             for (int index = 0; index < Constants.TotalVertexBuffers; index++)
             {
@@ -971,6 +985,11 @@ namespace Ryujinx.Graphics.Gpu.Engine.Threed
 
                 ulong address = vertexBuffer.Address.Pack();
 
+                if (_channel.MemoryManager.IsMapped(address))
+                {
+                    vbEnableMask |= 1u << index;
+                }
+
                 int stride = vertexBuffer.UnpackStride();
 
                 bool instanced = _state.State.VertexBufferInstanced[index];
@@ -1017,6 +1036,12 @@ namespace Ryujinx.Graphics.Gpu.Engine.Threed
                 _pipeline.VertexBuffers[index] = new BufferPipelineDescriptor(_channel.MemoryManager.IsMapped(address), stride, divisor);
                 _channel.BufferManager.SetVertexBuffer(index, address, size, stride, divisor);
             }
+
+            if (_vbEnableMask != vbEnableMask)
+            {
+                _vbEnableMask = vbEnableMask;
+                UpdateVertexAttribState();
+            }
         }
 
         /// <summary>

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 = 3897;
+        private const uint CodeGenVersion = 3943;
 
         private const string SharedTocFileName = "shared.toc";
         private const string SharedDataFileName = "shared.data";

Ryujinx.Graphics.OpenGL/VertexArray.cs

@@ -10,13 +10,9 @@ namespace Ryujinx.Graphics.OpenGL
     {
         public int Handle { get; private set; }
 
-        private bool _needsAttribsUpdate;
-
         private readonly VertexAttribDescriptor[] _vertexAttribs;
         private readonly VertexBufferDescriptor[] _vertexBuffers;
 
-        private int _vertexAttribsCount;
-        private int _vertexBuffersCount;
         private int _minVertexCount;
 
         private uint _vertexAttribsInUse;
@@ -76,9 +72,7 @@ namespace Ryujinx.Graphics.OpenGL
                 _vertexBuffers[bindingIndex] = vb;
             }
 
-            _vertexBuffersCount = bindingIndex;
             _minVertexCount = minVertexCount;
-            _needsAttribsUpdate = true;
         }
 
         public void SetVertexAttributes(ReadOnlySpan<VertexAttribDescriptor> vertexAttribs)
@@ -131,8 +125,6 @@ namespace Ryujinx.Graphics.OpenGL
                 _vertexAttribs[index] = attrib;
             }
 
-            _vertexAttribsCount = index;
-
             for (; index < Constants.MaxVertexAttribs; index++)
             {
                 DisableVertexAttrib(index);
@@ -160,13 +152,11 @@ namespace Ryujinx.Graphics.OpenGL
         public void PreDraw(int vertexCount)
         {
             LimitVertexBuffers(vertexCount);
-            Validate();
         }
 
         public void PreDrawVbUnbounded()
         {
             UnlimitVertexBuffers();
-            Validate();
         }
 
         public void LimitVertexBuffers(int vertexCount)
@@ -252,36 +242,6 @@ namespace Ryujinx.Graphics.OpenGL
             _vertexBuffersLimited = 0;
         }
 
-        public void Validate()
-        {
-            for (int attribIndex = 0; attribIndex < _vertexAttribsCount; attribIndex++)
-            {
-                VertexAttribDescriptor attrib = _vertexAttribs[attribIndex];
-
-                if (!attrib.IsZero)
-                {
-                    if ((uint)attrib.BufferIndex >= _vertexBuffersCount)
-                    {
-                        DisableVertexAttrib(attribIndex);
-                        continue;
-                    }
-
-                    if (_vertexBuffers[attrib.BufferIndex].Buffer.Handle == BufferHandle.Null)
-                    {
-                        DisableVertexAttrib(attribIndex);
-                        continue;
-                    }
-
-                    if (_needsAttribsUpdate)
-                    {
-                        EnableVertexAttrib(attribIndex);
-                    }
-                }
-            }
-
-            _needsAttribsUpdate = false;
-        }
-
         [MethodImpl(MethodImplOptions.AggressiveInlining)]
         private void EnableVertexAttrib(int index)
         {

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

@@ -234,7 +234,7 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Spirv
             var source = operation.GetSource(0);
 
             var uvec4Type = context.TypeVector(context.TypeU32(), 4);
-            var execution = context.Constant(context.TypeU32(), 3); // Subgroup
+            var execution = context.Constant(context.TypeU32(), Scope.Subgroup);
 
             var maskVector = context.GroupNonUniformBallot(uvec4Type, execution, context.Get(AggregateType.Bool, source));
             var mask = context.CompositeExtract(context.TypeU32(), maskVector, (SpvLiteralInteger)0);
@@ -1233,7 +1233,7 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Spirv
             var maxThreadId = context.BitwiseOr(context.TypeU32(), minThreadId, clampNotSegMask);
             var srcThreadId = context.BitwiseOr(context.TypeU32(), indexNotSegMask, minThreadId);
             var valid = context.ULessThanEqual(context.TypeBool(), srcThreadId, maxThreadId);
-            var value = context.SubgroupReadInvocationKHR(context.TypeFP32(), x, srcThreadId);
+            var value = context.GroupNonUniformShuffle(context.TypeFP32(), context.Constant(context.TypeU32(), (int)Scope.Subgroup), x, srcThreadId);
             var result = context.Select(context.TypeFP32(), valid, value, x);
 
             var validLocal = (AstOperand)operation.GetSource(3);
@@ -1263,7 +1263,7 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Spirv
             var maxThreadId = context.BitwiseOr(context.TypeU32(), minThreadId, clampNotSegMask);
             var srcThreadId = context.IAdd(context.TypeU32(), threadId, index);
             var valid = context.ULessThanEqual(context.TypeBool(), srcThreadId, maxThreadId);
-            var value = context.SubgroupReadInvocationKHR(context.TypeFP32(), x, srcThreadId);
+            var value = context.GroupNonUniformShuffle(context.TypeFP32(), context.Constant(context.TypeU32(), (int)Scope.Subgroup), x, srcThreadId);
             var result = context.Select(context.TypeFP32(), valid, value, x);
 
             var validLocal = (AstOperand)operation.GetSource(3);
@@ -1289,7 +1289,7 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Spirv
             var minThreadId = context.BitwiseAnd(context.TypeU32(), threadId, segMask);
             var srcThreadId = context.ISub(context.TypeU32(), threadId, index);
             var valid = context.SGreaterThanEqual(context.TypeBool(), srcThreadId, minThreadId);
-            var value = context.SubgroupReadInvocationKHR(context.TypeFP32(), x, srcThreadId);
+            var value = context.GroupNonUniformShuffle(context.TypeFP32(), context.Constant(context.TypeU32(), (int)Scope.Subgroup), x, srcThreadId);
             var result = context.Select(context.TypeFP32(), valid, value, x);
 
             var validLocal = (AstOperand)operation.GetSource(3);
@@ -1319,7 +1319,7 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Spirv
             var maxThreadId = context.BitwiseOr(context.TypeU32(), minThreadId, clampNotSegMask);
             var srcThreadId = context.BitwiseXor(context.TypeU32(), threadId, index);
             var valid = context.ULessThanEqual(context.TypeBool(), srcThreadId, maxThreadId);
-            var value = context.SubgroupReadInvocationKHR(context.TypeFP32(), x, srcThreadId);
+            var value = context.GroupNonUniformShuffle(context.TypeFP32(), context.Constant(context.TypeU32(), (int)Scope.Subgroup), x, srcThreadId);
             var result = context.Select(context.TypeFP32(), valid, value, x);
 
             var validLocal = (AstOperand)operation.GetSource(3);
@@ -1861,19 +1861,22 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Spirv
 
         private static OperationResult GenerateVoteAll(CodeGenContext context, AstOperation operation)
         {
-            var result = context.SubgroupAllKHR(context.TypeBool(), context.Get(AggregateType.Bool, operation.GetSource(0)));
+            var execution = context.Constant(context.TypeU32(), Scope.Subgroup);
+            var result = context.GroupNonUniformAll(context.TypeBool(), execution, context.Get(AggregateType.Bool, operation.GetSource(0)));
             return new OperationResult(AggregateType.Bool, result);
         }
 
         private static OperationResult GenerateVoteAllEqual(CodeGenContext context, AstOperation operation)
         {
-            var result = context.SubgroupAllEqualKHR(context.TypeBool(), context.Get(AggregateType.Bool, operation.GetSource(0)));
+            var execution = context.Constant(context.TypeU32(), Scope.Subgroup);
+            var result = context.GroupNonUniformAllEqual(context.TypeBool(), execution, context.Get(AggregateType.Bool, operation.GetSource(0)));
             return new OperationResult(AggregateType.Bool, result);
         }
 
         private static OperationResult GenerateVoteAny(CodeGenContext context, AstOperation operation)
         {
-            var result = context.SubgroupAnyKHR(context.TypeBool(), context.Get(AggregateType.Bool, operation.GetSource(0)));
+            var execution = context.Constant(context.TypeU32(), Scope.Subgroup);
+            var result = context.GroupNonUniformAny(context.TypeBool(), execution, context.Get(AggregateType.Bool, operation.GetSource(0)));
             return new OperationResult(AggregateType.Bool, result);
         }
 

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

@@ -50,12 +50,12 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Spirv
             CodeGenContext context = new CodeGenContext(info, config, instPool, integerPool);
 
             context.AddCapability(Capability.GroupNonUniformBallot);
+            context.AddCapability(Capability.GroupNonUniformShuffle);
+            context.AddCapability(Capability.GroupNonUniformVote);
             context.AddCapability(Capability.ImageBuffer);
             context.AddCapability(Capability.ImageGatherExtended);
             context.AddCapability(Capability.ImageQuery);
             context.AddCapability(Capability.SampledBuffer);
-            context.AddCapability(Capability.SubgroupBallotKHR);
-            context.AddCapability(Capability.SubgroupVoteKHR);
 
             if (config.TransformFeedbackEnabled && config.LastInVertexPipeline)
             {
@@ -94,9 +94,6 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Spirv
                 context.AddCapability(Capability.DrawParameters);
             }
 
-            context.AddExtension("SPV_KHR_shader_ballot");
-            context.AddExtension("SPV_KHR_subgroup_vote");
-
             Declarations.DeclareAll(context, info);
 
             if ((info.HelperFunctionsMask & NeedsInvocationIdMask) != 0)

Ryujinx.Graphics.Vulkan/VulkanInitialization.cs

@@ -37,7 +37,6 @@ namespace Ryujinx.Graphics.Vulkan
         public static string[] RequiredExtensions { get; } = new string[]
         {
             KhrSwapchain.ExtensionName,
-            "VK_EXT_shader_subgroup_vote",
             ExtTransformFeedback.ExtensionName
         };
 

Ryujinx.Memory/Tracking/ConcurrentBitmap.cs

@@ -41,7 +41,7 @@ namespace Ryujinx.Memory.Tracking
         {
             for (int i = 0; i < Masks.Length; i++)
             {
-                if (Volatile.Read(ref Masks[i]) != 0)
+                if (Interlocked.Read(ref Masks[i]) != 0)
                 {
                     return true;
                 }
@@ -62,7 +62,7 @@ namespace Ryujinx.Memory.Tracking
 
             long wordMask = 1L << wordBit;
 
-            return (Volatile.Read(ref Masks[wordIndex]) & wordMask) != 0;
+            return (Interlocked.Read(ref Masks[wordIndex]) & wordMask) != 0;
         }
 
         /// <summary>
@@ -86,7 +86,7 @@ namespace Ryujinx.Memory.Tracking
             int endBit = end & IntMask;
             long endMask = (long)(ulong.MaxValue >> (IntMask - endBit));
 
-            long startValue = Volatile.Read(ref Masks[startIndex]);
+            long startValue = Interlocked.Read(ref Masks[startIndex]);
 
             if (startIndex == endIndex)
             {
@@ -100,13 +100,13 @@ namespace Ryujinx.Memory.Tracking
 
             for (int i = startIndex + 1; i < endIndex; i++)
             {
-                if (Volatile.Read(ref Masks[i]) != 0)
+                if (Interlocked.Read(ref Masks[i]) != 0)
                 {
                     return true;
                 }
             }
 
-            long endValue = Volatile.Read(ref Masks[endIndex]);
+            long endValue = Interlocked.Read(ref Masks[endIndex]);
 
             if ((endValue & endMask) != 0)
             {
@@ -128,23 +128,14 @@ namespace Ryujinx.Memory.Tracking
 
             long wordMask = 1L << wordBit;
 
-            long existing;
-            long newValue;
-
-            do
+            if (value)
             {
-                existing = Volatile.Read(ref Masks[wordIndex]);
-
-                if (value)
-                {
-                    newValue = existing | wordMask;
-                }
-                else
-                {
-                    newValue = existing & ~wordMask;
-                }
+                Interlocked.Or(ref Masks[wordIndex], wordMask);
+            }
+            else
+            {
+                Interlocked.And(ref Masks[wordIndex], ~wordMask);
             }
-            while (Interlocked.CompareExchange(ref Masks[wordIndex], newValue, existing) != existing);
         }
 
         /// <summary>
@@ -154,7 +145,7 @@ namespace Ryujinx.Memory.Tracking
         {
             for (int i = 0; i < Masks.Length; i++)
             {
-                Volatile.Write(ref Masks[i], 0);
+                Interlocked.Exchange(ref Masks[i], 0);
             }
         }
     }