iGPU fixes

Snooper/Core/Containers/Buffers/Buffer.cs

@@ -38,7 +38,7 @@ public abstract class Buffer<T>(BufferTarget target, BufferUsageHint usageHint,
     public int Count { get; private set; }
     public int Capacity { get; private set; }
 
-    private bool _bInitialized;
+    protected bool IsAllocated;
     private readonly Dictionary<int, BufferAllocationMetadata> _allocations = new();
     private readonly SortedSet<FreeBlock> _freeBlocks = new(Comparer<FreeBlock>.Create((a, b) =>
     {
@@ -52,12 +52,12 @@ public abstract class Buffer<T>(BufferTarget target, BufferUsageHint usageHint,
 
     public override void Generate()
     {
-        if (_bInitialized)
+        if (IsAllocated)
             throw new InvalidOperationException("Buffer is already initialized.");
 
         GL.CreateBuffers(1, out uint handle);
         Handle = handle;
-        _bInitialized = false;
+        IsAllocated = false;
     }
 
     public void Bind()
@@ -78,11 +78,11 @@ private void ResizeIfNeeded(int newSize, double factor = 1.5, bool copy = false)
         var oldCapacity = Capacity;
         Capacity = (int) Math.Max(Capacity * factor, newSize);
 
-        if (_bInitialized)
+        if (IsAllocated)
         {
             Log.Warning("Resizing buffer {0} ({1}) from {2} to {3} (asked: {4}) (initialized!!!!!!)", Handle, PName, oldCapacity, Capacity, newSize);
 
-            _bInitialized = false;
+            IsAllocated = false;
             if (copy)
             {
                 var oldBuffer = Handle;
@@ -109,15 +109,15 @@ private void ResizeIfNeeded(int newSize, double factor = 1.5, bool copy = false)
 
     public void Reallocate(int size)
     {
-        _bInitialized = false;
+        IsAllocated = false;
         Allocate(size);
     }
 
     public void Allocate(uint size) => Allocate((int)size);
     public void Allocate(int size)
     {
         ArgumentOutOfRangeException.ThrowIfNegativeOrZero(size);
-        if (_bInitialized)
+        if (IsAllocated)
             throw new InvalidOperationException("Buffer is already initialized. Use Update method to modify data.");
 
         if (size > Capacity)
@@ -133,7 +133,7 @@ public void Allocate(int size)
         // _allocationIdCounter = 0;
         // _allocations.Clear();
         // _freeBlocks.Clear();
-        _bInitialized = true;
+        IsAllocated = true;
     }
 
     public BufferAllocation Add(T data) => AddInternal([data]);
@@ -143,7 +143,7 @@ private BufferAllocation AddInternal(T[] data)
         var length = data.Length;
         ArgumentOutOfRangeException.ThrowIfNegativeOrZero(length);
 
-        if (!_bInitialized)
+        if (!IsAllocated)
         {
             Allocate(length);
         }
@@ -176,7 +176,7 @@ private void UpsertInternal(int index, T[] data)
         ArgumentOutOfRangeException.ThrowIfNegativeOrZero(length);
         ArgumentOutOfRangeException.ThrowIfNegative(index);
 
-        if (!_bInitialized)
+        if (!IsAllocated)
         {
             Allocate(index + length);
         }
@@ -196,7 +196,7 @@ private void UpsertInternal(int index, T[] data)
     public void Update(int allocationId, T[] data) => UpdateInternal(allocationId, data);
     private void UpdateInternal(int allocationId, T[] data, bool batched = false)
     {
-        if (!_bInitialized)
+        if (!IsAllocated)
             throw new InvalidOperationException("Buffer is not initialized. Use Add method to initialize it.");
 
         if (!_allocations.TryGetValue(allocationId, out var metadata))
@@ -215,7 +215,7 @@ private void UpdateInternal(int allocationId, T[] data, bool batched = false)
     public void UpdateCustom<TCustom>(BufferAllocation allocation, TCustom data, int offset) where TCustom : unmanaged => UpdateCustomInternal(allocation.AllocationId, data, offset);
     private void UpdateCustomInternal<TCustom>(int allocationId, TCustom data, int offset) where TCustom : unmanaged
     {
-        if (!_bInitialized)
+        if (!IsAllocated)
             throw new InvalidOperationException("Buffer is not initialized. Use Add method to initialize it.");
 
         if (!_allocations.TryGetValue(allocationId, out var metadata))
@@ -304,7 +304,7 @@ public BufferAllocation CopyFrom(Buffer<T> sourceBuffer, BufferAllocation source
         var length = sourceAllocation.Length;
         ArgumentOutOfRangeException.ThrowIfNegativeOrZero(length);
 
-        if (!_bInitialized)
+        if (!IsAllocated)
         {
             Allocate(length);
         }
@@ -326,7 +326,7 @@ public BufferAllocation CopyFrom(Buffer<T> sourceBuffer, BufferAllocation source
 
     public void Clear()
     {
-        if (!_bInitialized)
+        if (!IsAllocated)
             throw new InvalidOperationException("Cannot clear a buffer that is not initialized.");
 
         ClearStorage(0, TotalElements * Stride);

Snooper/Core/Containers/Buffers/ShaderStorageBuffer.cs

@@ -11,7 +11,11 @@ public sealed class ShaderStorageBuffer<T>(BufferUsageHint usageHint = BufferUsa
 
     public void Bind(uint index)
     {
-        GL.BindBufferBase(BufferRangeTarget.ShaderStorageBuffer, index, Handle);
+        // iGPUs don't like binding unallocated buffers, and unlike some other buffers, we never unbind SSBOs
+        // silently skipping the bind leaves whatever another system bound at this index in place
+        // that's too risky, so we unbind the slot instead so an unallocated buffer reads as out of range
+        // TODO: we should properly unbind SSBOs, and skip if unallocated
+        GL.BindBufferBase(BufferRangeTarget.ShaderStorageBuffer, index, IsAllocated ? Handle : 0);
     }
 
     public void QueueUpdate(BufferAllocation allocation, T data) => _batcher.Add(allocation, data);

Snooper/Core/Containers/Programs/EmbeddedShader.cs

@@ -1,6 +1,7 @@
 using System.Reflection;
 using OpenTK.Graphics.OpenGL4;
 using Snooper.Core.Containers.Buffers;
+using Snooper.Core.Hardware;
 
 namespace Snooper.Core.Containers.Programs;
 
@@ -25,7 +26,7 @@ protected override uint CompileShader(ShaderType type, string file)
             content = string.Join("\n", Defines.Select(d => $"#define {d}")) + "\n" + content;
         }
 
-        content = string.Join('\n', "#version 460 core", "", Bindings.GlslDefines, "", content);
+        content = string.Join('\n', "#version 460 core", "", Bindings.GlslDefines, DeviceInfo.GlslDefines, content);
 
         return base.CompileShader(type, content);
     }

Snooper/Core/Containers/Resources/GeometryPool.cs

@@ -1,8 +1,7 @@
-using System.Numerics;
-using CUE4Parse.UE4.Objects.Core.Misc;
+using CUE4Parse.UE4.Objects.Core.Misc;
 using OpenTK.Graphics.OpenGL4;
 using Snooper.Core.Containers.Buffers;
-using Snooper.Rendering.Components.Camera;
+using Snooper.Core.Hardware;
 using Snooper.Rendering.Components.Descriptors;
 
 namespace Snooper.Core.Containers.Resources;
@@ -18,6 +17,51 @@ public class GeometryHandle(uint firstIndex, uint baseVertex, BufferAllocation m
     public int OverrideLod { get; internal set; } = overrideLod;
 }
 
+public readonly struct VertexArrayLayout
+{
+    private readonly uint _vao;
+    private readonly uint _vbo;
+    private readonly int _stride;
+
+    public VertexArrayLayout(uint vao, uint vbo, int stride)
+    {
+        _vao = vao;
+        _vbo = vbo;
+        _stride = stride;
+
+        if (!DeviceInfo.IsIntel)
+        {
+            GL.VertexArrayVertexBuffer(vao, 0, vbo, 0, stride);
+        }
+    }
+
+    public VertexArrayLayout Float(uint location, int size, VertexAttribType type = VertexAttribType.Float, bool normalized = false, uint offset = 0)
+    {
+        GL.VertexArrayAttribFormat(_vao, location, size, type, normalized, DeviceInfo.IsIntel ? 0 : offset);
+        return Enable(location, offset);
+    }
+
+    public VertexArrayLayout Integer(uint location, int size, VertexAttribIType type = VertexAttribIType.UnsignedInt, uint offset = 0)
+    {
+        GL.VertexArrayAttribIFormat(_vao, location, size, type, DeviceInfo.IsIntel ? 0 : offset);
+        return Enable(location, offset);
+    }
+
+    private VertexArrayLayout Enable(uint location, uint offset)
+    {
+        var binding = 0u;
+        if (DeviceInfo.IsIntel)
+        {
+            binding = location;
+            GL.VertexArrayVertexBuffer(_vao, binding, _vbo, (nint)offset, _stride);
+        }
+
+        GL.VertexArrayAttribBinding(_vao, location, binding);
+        GL.EnableVertexArrayAttrib(_vao, location);
+        return this;
+    }
+}
+
 public class GeometryPool<TVertex> : IMemoryDetailsProvider, IDisposable where TVertex : unmanaged
 {
     private readonly VertexArray _vao = new();
@@ -27,7 +71,7 @@ public class GeometryPool<TVertex> : IMemoryDetailsProvider, IDisposable where T
     private readonly CullingResources _culling = new();
 
     private readonly Dictionary<FGuid, GeometryHandle> _cache = new();
-    private Action<uint>? _vertexLayoutSetter;
+    private Action<VertexArrayLayout>? _vertexLayoutSetter;
 
     public void Generate()
     {
@@ -41,18 +85,16 @@ public void Generate()
         _vbo.OnHandleChanged += (_, _) => BindBuffersToVao();
     }
 
-    public void SetVertexLayout(Action<uint> setter)
+    public void SetVertexLayout(Action<VertexArrayLayout> setter)
     {
         _vertexLayoutSetter = setter;
         BindBuffersToVao();
     }
 
     private void BindBuffersToVao()
     {
-        GL.VertexArrayVertexBuffer(_vao, 0, _vbo, 0, _vbo.Stride);
         GL.VertexArrayElementBuffer(_vao, _ebo);
-
-        _vertexLayoutSetter?.Invoke(_vao);
+        _vertexLayoutSetter?.Invoke(new VertexArrayLayout(_vao, _vbo, _vbo.Stride));
     }
 
     public void Allocate(AllocationCounts counts)

Snooper/Core/Containers/Resources/IndirectResources.cs

@@ -41,7 +41,7 @@ public void Generate()
         _materialData.Generate();
     }
 
-    public void SetVertexLayout(Action<uint> setter) => _geometry.SetVertexLayout(setter);
+    public void SetVertexLayout(Action<VertexArrayLayout> setter) => _geometry.SetVertexLayout(setter);
 
     public void Allocate(AllocationCounts counts)
     {

Snooper/Core/Hardware/DeviceInfo.cs

@@ -10,12 +10,18 @@ public class DeviceInfo
     public ExtensionSupport ExtensionSupport { get; } = new();
     public GpuMemoryInfo Memory { get; } = new();
 
+    public static bool IsIntel { get; private set; }
+    public static string GlslDefines { get; private set; } = string.Empty;
+
     public void Load()
     {
         Name = GL.GetString(StringName.Renderer);
         Vendor = GL.GetString(StringName.Vendor);
         MaxShaderStorageBufferBindings = GL.GetInteger(GetPName.MaxShaderStorageBufferBindings);
         ExtensionSupport.Load();
         Memory.Load(ExtensionSupport);
+
+        IsIntel = Vendor.Contains("Intel", StringComparison.OrdinalIgnoreCase);
+        GlslDefines = IsIntel ? "#define BINDLESS_RAW_HANDLES\n" : string.Empty;
     }
 }

Snooper/Core/Managers/ActorManager.cs

@@ -191,6 +191,10 @@ private void DequeueSystems(int limit = 0)
 
             Systems.Add(system.Order, system);
             system.Load();
+
+            if (system is IResizable resizable)
+                resizable.Resize(_width, _height); // resize right away for the screen-sized resources to get allocated (ClusteredLightSystem)
+
             count++;
         }
     }
@@ -232,8 +236,13 @@ private void TrackBackgroundWork()
         }
     }
 
+    private int _width;
+    private int _height;
     public virtual void Resize(int newWidth, int newHeight)
     {
+        _width = newWidth;
+        _height = newHeight;
+
         foreach (var system in Systems.Values.OfType<IResizable>())
             system.Resize(newWidth, newHeight);
     }

Snooper/Rendering/Components/Mesh/MeshComponent.cs

@@ -23,7 +23,7 @@ namespace Snooper.Rendering.Components.Mesh;
 /// <summary>
 /// Packed vertex layout — 20 bytes total<br/>
 ///     loc 0: uvec2  — pos.x|pos.y (half2), pos.z|0 (half2)         [offset  0, 8 bytes]<br/>
-///     loc 1: uint   — normal  xyzw RGB10A2 SNorm                   [offset  8, 4 bytes]<br/>
+///     loc 1: uint   — normal  xyz RGB10A2 SNorm, w = basis sign    [offset  8, 4 bytes]<br/>
 ///     loc 2: uint   — tangent xyz RGB10A2 SNorm, w = texLayer(0-3) [offset 12, 4 bytes]<br/>
 ///     loc 3: uint   — uv.x|uv.y (half2)                            [offset 16, 4 bytes]<br/>
 /// </summary>
@@ -43,10 +43,11 @@ private static uint PackHalf2(float x, float y)
         return hx | (hy << 16);
     }
 
-    private static uint PackRgb10A2Snorm(Vector4 v) => PackRgb10A2Snorm(v.X, v.Y, v.Z, Snorm10(v.W));
+    private static uint PackRgb10A2Snorm(Vector4 v) => PackRgb10A2Snorm(v.X, v.Y, v.Z, Snorm2(v.W));
     private static uint PackRgb10A2Snorm(Vector3 v, uint texLayer) => PackRgb10A2Snorm(v.X, v.Y, v.Z, texLayer & 0x3u);
     private static uint PackRgb10A2Snorm(float x, float y, float z, uint w) => Snorm10(x) | (Snorm10(y) << 10) | (Snorm10(z) << 20) | (w << 30);
     private static uint Snorm10(float f) => (uint)(int)MathF.Round(Math.Clamp(f, -1f, 1f) * 511f) & 0x3FFu;
+    private static uint Snorm2(float f) => f < 0f ? 3u : 1u;
 }
 
 public unsafe struct PerMaterialMeshData : IPerMaterialData
@@ -189,7 +190,7 @@ public Geometry(MeshVertex[] vertices, uint[] indices, FColor[]? colors, FMeshUV
             {
                 var vertex = vertices[i];
                 var position = new Vector3(vertex.Position.X, vertex.Position.Z, vertex.Position.Y) * Settings.GlobalScale;
-                var normal = new Vector4(vertex.Normal.X, vertex.Normal.Z, vertex.Normal.Y, vertex.Normal.W);
+                var normal = new Vector4(vertex.Normal.X, vertex.Normal.Z, vertex.Normal.Y, -vertex.Normal.W);
                 var tangent = new Vector3(vertex.Tangent.X, vertex.Tangent.Z, vertex.Tangent.Y);
                 var texCoord = new Vector2(vertex.Uv.U, vertex.Uv.V);
                 var texLayer = extraUvs != null ? (uint)Math.Floor(extraUvs[i].U) : 0u;
@@ -219,7 +220,7 @@ public Geometry(SkinnedMeshVertex[] vertices, uint[] indices, FColor[]? colors,
             {
                 var vertex = vertices[i];
                 var position = new Vector3(vertex.Position.X, vertex.Position.Z, vertex.Position.Y) * Settings.GlobalScale;
-                var normal = new Vector4(vertex.Normal.X, vertex.Normal.Z, vertex.Normal.Y, vertex.Normal.W);
+                var normal = new Vector4(vertex.Normal.X, vertex.Normal.Z, vertex.Normal.Y, -vertex.Normal.W);
                 var tangent = new Vector3(vertex.Tangent.X, vertex.Tangent.Z, vertex.Tangent.Y);
                 var texCoord = new Vector2(vertex.Uv.U, vertex.Uv.V);
                 var texLayer = extraUvs != null ? (uint)Math.Floor(extraUvs[i].U) : 0u;

Snooper/Rendering/Managers/PostProcessor.cs

@@ -95,7 +95,7 @@ public override void Generate()
                     shader.SetUniform("ssao", 4);
                 }
 
-                if (ctx.LightSystem is { IsEnabled: true } system)
+                if (ctx.LightSystem is { IsEnabled: true, HasClusters: true } system)
                 {
                     system.BindForRendering();
                     shader.SetUniform("useLighting", true);
@@ -191,7 +191,7 @@ public override void Generate()
                 shader.SetUniform("uShowGrid", ctx.ShowGrid);
                 shader.SetUniform("uMaxLightsPerCluster", ClusteredLightSystem.MaxLightsPerClusterLimit);
 
-                if (ctx.LightSystem is { IsEnabled: true } system)
+                if (ctx.LightSystem is { IsEnabled: true, HasClusters: true } system)
                 {
                     system.BindForRendering();
                     shader.SetUniform("uHasLights", true);

Snooper/Rendering/Systems/BillboardSystem.cs

@@ -16,10 +16,5 @@ public class BillboardSystem : PrimitiveSystem<Vector2, BillboardComponent, PerI
         [CommandBufferType.Transparent] = new EmbeddedShader("billboard")
     };
 
-    protected override Action<uint> VertexLayout { get; } = vao =>
-    {
-        GL.VertexArrayAttribFormat(vao, 0, 2, VertexAttribType.Float, false, 0);
-        GL.EnableVertexArrayAttrib(vao, 0);
-        GL.VertexArrayAttribBinding(vao, 0, 0);
-    };
+    protected override Action<VertexArrayLayout> VertexLayout { get; } = layout => layout.Float(0, 2);
 }

Snooper/Rendering/Systems/ClusteredLightSystem.cs

@@ -91,6 +91,7 @@ internal abstract class LightBindings : Bindings
     public int GridDimensionX { get; private set; }
     public int GridDimensionY { get; private set; }
     public int GridDimensionZ => 16;
+    public bool HasClusters => _numClusters > 0;
 
     private int _numClusters;
     private int _numWorkGroups;
@@ -161,7 +162,7 @@ private bool ConsumeClustersDirty(CameraComponent camera)
 
     private void BuildClusters(CameraComponent camera)
     {
-        if (_numClusters == 0) return;
+        if (!HasClusters) return;
 
         _clusterBuildProgram.Use();
         _clusterBuildProgram.SetUniform("uScreenWidth", _screenWidth);
@@ -183,7 +184,7 @@ private void BuildClusters(CameraComponent camera)
 
     private void CullLights(CameraComponent camera)
     {
-        if (_numClusters == 0 || _lightDataBuffer.Count == 0)
+        if (!HasClusters || _lightDataBuffer.Count == 0)
         {
             return;
         }
@@ -255,6 +256,7 @@ public void Resize(int newWidth, int newHeight)
         GridDimensionY = (_screenHeight + TileSize - 1) / TileSize;
         _numClusters = GridDimensionX * GridDimensionY * GridDimensionZ;
         _numWorkGroups = (_numClusters + WorkGroupSize - 1) / WorkGroupSize;
+        if (!HasClusters) return;
 
         _clusterAABBBuffer.Reallocate(_numClusters);
         _clusterDataBuffer.Reallocate(_numClusters);