Draft: Compute pipeline support

src/cs/examples/Examples.Gpu/Examples/E004_BasicVertexBuffer.cs

@@ -51,8 +51,7 @@ public override bool OnStart()
         vertexShader.Dispose();
         fragmentShader.Dispose();
 
-        if (!Device.TryCreateDataBuffer<VertexPositionColor>(
-                3, out _vertexBuffer))
+        if (!Device.TryCreateDataBuffer<VertexPositionColor>(GpuBufferUsageFlags.Vertex, 3, out _vertexBuffer))
         {
             return false;
         }

src/cs/examples/Examples.Gpu/Examples/E005_CullMode.cs

@@ -75,12 +75,12 @@ public override bool OnStart()
         fragmentShader.Dispose();
 
         // Create the vertex buffers. They're the same except for the vertex order.
-        if (!Device.TryCreateDataBuffer<VertexPositionColor>(3, out _vertexBufferCw))
+        if (!Device.TryCreateDataBuffer<VertexPositionColor>(GpuBufferUsageFlags.Vertex, 3, out _vertexBufferCw))
         {
             return false;
         }
 
-        if (!Device.TryCreateDataBuffer<VertexPositionColor>(3, out _vertexBufferCcw))
+        if (!Device.TryCreateDataBuffer<VertexPositionColor>(GpuBufferUsageFlags.Vertex, 3, out _vertexBufferCcw))
         {
             return false;
         }

src/cs/examples/Examples.Gpu/Examples/E006_BasicStencil.cs

@@ -110,7 +110,7 @@ public override bool OnStart()
         vertexShader.Dispose();
         fragmentShader.Dispose();
 
-        if (!Device.TryCreateDataBuffer<VertexPositionColor>(6, out _vertexBuffer))
+        if (!Device.TryCreateDataBuffer<VertexPositionColor>(GpuBufferUsageFlags.Vertex, 6, out _vertexBuffer))
         {
             return false;
         }

src/cs/examples/Examples.Gpu/Examples/E007_InstancedIndexed.cs

@@ -54,12 +54,12 @@ public override bool OnStart()
         vertexShader.Dispose();
         fragmentShader.Dispose();
 
-        if (!Device.TryCreateDataBuffer<VertexPositionColor>(9, out _vertexBuffer))
+        if (!Device.TryCreateDataBuffer<VertexPositionColor>(GpuBufferUsageFlags.Vertex, 9, out _vertexBuffer))
         {
             return false;
         }
 
-        if (!Device.TryCreateDataBuffer<ushort>(6, out _indexBuffer))
+        if (!Device.TryCreateDataBuffer<ushort>(GpuBufferUsageFlags.Index, 6, out _indexBuffer))
         {
             return false;
         }

src/cs/examples/Examples.Gpu/Examples/E008_TexturedQuad.cs

@@ -169,13 +169,12 @@ public override bool OnStart()
             return false;
         }
 
-        if (!Device.TryCreateDataBuffer<VertexPositionTexture>(
-                4, out _vertexBuffer, "Ravioli Vertex Buffer 🥣"))
+        if (!Device.TryCreateDataBuffer<VertexPositionTexture>(GpuBufferUsageFlags.Vertex, 4, out _vertexBuffer, "Ravioli Vertex Buffer 🥣"))
         {
             return false;
         }
 
-        if (!Device.TryCreateDataBuffer<ushort>(6, out _indexBuffer))
+        if (!Device.TryCreateDataBuffer<ushort>(GpuBufferUsageFlags.Index, 6, out _indexBuffer))
         {
             return false;
         }

src/cs/examples/Examples.Gpu/Examples/E009_TexturedAnimatedQuad.cs

@@ -98,12 +98,12 @@ public override bool OnStart()
             return false;
         }
 
-        if (!Device.TryCreateDataBuffer<VertexPositionTexture>(4, out _vertexBuffer))
+        if (!Device.TryCreateDataBuffer<VertexPositionTexture>(GpuBufferUsageFlags.Vertex, 4, out _vertexBuffer))
         {
             return false;
         }
 
-        if (!Device.TryCreateDataBuffer<ushort>(6, out _indexBuffer))
+        if (!Device.TryCreateDataBuffer<ushort>(GpuBufferUsageFlags.Index, 6, out _indexBuffer))
         {
             return false;
         }

src/cs/examples/Examples.Gpu/Examples/E011_BasicCompute.cs

@@ -95,7 +95,7 @@ public override bool OnStart()
 
         var verticesCount = 6;
         var verticesBytesCount = VertexPositionTexture.SizeOf * verticesCount;
-        if (!Device.TryCreateDataBuffer<VertexPositionTexture>(verticesCount, out _vertexBuffer!))
+        if (!Device.TryCreateDataBuffer<VertexPositionTexture>(GpuBufferUsageFlags.Vertex, verticesCount, out _vertexBuffer!))
         {
             return false;
         }

src/cs/games/Games.MinecraftClone/ECS/Systems/Draw/RenderSystem.cs

@@ -156,7 +156,7 @@ private bool TryCreateVertexBuffer()
         const int vertexCount = 24;
         var bufferSize = Unsafe.SizeOf<Vertex>() * vertexCount;
 
-        if (!_device.TryCreateDataBuffer<Vertex>(vertexCount, out _vertexBuffer))
+        if (!_device.TryCreateDataBuffer<Vertex>(GpuBufferUsageFlags.Vertex, vertexCount, out _vertexBuffer))
         {
             return false;
         }
@@ -344,7 +344,7 @@ private bool TryCreateIndexBuffer()
         const int indexCount = 36;
         var bufferSize = Unsafe.SizeOf<ushort>() * indexCount;
 
-        if (!_device.TryCreateDataBuffer<ushort>(indexCount, out _indexBuffer))
+        if (!_device.TryCreateDataBuffer<ushort>(GpuBufferUsageFlags.Index, indexCount, out _indexBuffer))
         {
             return false;
         }

src/cs/production/SDL/GPU/GpuBufferUsageFlags.cs

@@ -0,0 +1,41 @@
+// Copyright (c) Bottlenose Labs Inc. (https://github.com/bottlenoselabs). All rights reserved.
+// Licensed under the MIT license. See LICENSE file in the Git repository root directory for full license information.
+
+namespace bottlenoselabs.SDL;
+
+/// <summary>
+///     Specifies how a buffer is intended to be used by the client.
+/// </summary>
+[Flags]
+public enum GpuBufferUsageFlags
+{
+    /// <summary>
+    ///     Buffer is a vertex buffer.
+    /// </summary>
+    Vertex = 1 << 0,
+
+    /// <summary>
+    ///     Buffer is an index buffer.
+    /// </summary>
+    Index = 1 << 1,
+
+    /// <summary>
+    ///     Buffer is an indirect buffer.
+    /// </summary>
+    Indirect = 1 << 2,
+
+    /// <summary>
+    ///     Buffer supports storage reads in graphics stages.
+    /// </summary>
+    GraphicsStorageRead = 1 << 3,
+
+    /// <summary>
+    ///     Buffer supports storage reads in the compute stage.
+    /// </summary>
+    ComputeStorageRead = 1 << 4,
+
+    /// <summary>
+    ///     Buffer supports storage writes in the compute stage.
+    /// </summary>
+    ComputeStorageWrite = 1 << 5
+}

src/cs/production/SDL/GPU/GpuCommandBuffer.cs

@@ -264,6 +264,22 @@ public GpuComputePass BeginComputePass(in GpuComputePassParameters parameters)
         return computePass;
     }
 
+    /// <summary>
+    ///     Pushes the specified data to a vertex shader uniform slot. Subsequent draw calls will
+    ///     use this uniform data.
+    /// </summary>
+    /// <param name="data">TODO.</param>
+    /// <param name="startIndex">Index of the uniform slot to push data to.</param>
+    /// <typeparam name="T">Data type. It must respect std140 layout conventions.</typeparam>
+    public void PushVertexShaderUniformData<T>(in T data, int startIndex = 0)
+        where T : unmanaged
+    {
+        fixed (T* pointer = &data)
+        {
+            SDL_PushGPUVertexUniformData(HandleTyped, (uint)startIndex, pointer, (uint)sizeof(T));
+        }
+    }
+
     /// <summary>
     ///     Pushes the specified <see cref="Matrix4x4" /> to a vertex shader uniform slot. Subsequent draw calls will
     ///     use this uniform data.
@@ -272,9 +288,22 @@ public GpuComputePass BeginComputePass(in GpuComputePassParameters parameters)
     /// <param name="slotIndex">The vertex shader uniform slot to push data to.</param>
     public void PushVertexShaderUniformMatrix(in Matrix4x4 matrix, int slotIndex = 0)
     {
-        fixed (Matrix4x4* pointer = &matrix)
+        PushVertexShaderUniformData(matrix, slotIndex);
+    }
+
+    /// <summary>
+    ///     Pushes the specified data to a fragment shader uniform slot. Subsequent draw calls will
+    ///     use this uniform data.
+    /// </summary>
+    /// <param name="data">TODO.</param>
+    /// <param name="startIndex">Index of the uniform slot to push data to.</param>
+    /// <typeparam name="T">Data type. It must respect std140 layout conventions.</typeparam>
+    public void PushFragmentShaderUniformData<T>(in T data, int startIndex = 0)
+        where T : unmanaged
+    {
+        fixed (T* pointer = &data)
         {
-            SDL_PushGPUVertexUniformData(HandleTyped, (uint)slotIndex, pointer, (uint)sizeof(Matrix4x4));
+            SDL_PushGPUFragmentUniformData(HandleTyped, (uint)startIndex, pointer, (uint)sizeof(T));
         }
     }
 
@@ -286,9 +315,21 @@ public void PushVertexShaderUniformMatrix(in Matrix4x4 matrix, int slotIndex = 0
     /// <param name="slotIndex">The fragment shader uniform slot to push data to.</param>
     public void PushFragmentShaderUniformColor(in Rgba32F color, int slotIndex = 0)
     {
-        fixed (Rgba32F* pointer = &color)
+        PushFragmentShaderUniformData(color, slotIndex);
+    }
+
+    /// <summary>
+    ///     Pushes data to a uniform slot on the command buffer.
+    /// </summary>
+    /// <param name="data">TODO.</param>
+    /// <param name="startIndex">Index of the uniform slot to push data to.</param>
+    /// <typeparam name="T">Data type. It must respect std140 layout conventions.</typeparam>
+    public void PushComputeShaderUniformData<T>(in T data, int startIndex = 0)
+        where T : unmanaged
+    {
+        fixed (T* pointer = &data)
         {
-            SDL_PushGPUFragmentUniformData(HandleTyped, (uint)slotIndex, pointer, (uint)sizeof(Rgba32F));
+            SDL_PushGPUComputeUniformData(HandleTyped, (uint)startIndex, pointer, (uint)sizeof(T));
         }
     }
 

src/cs/production/SDL/GPU/GpuComputePass.cs

@@ -13,6 +13,8 @@ public sealed unsafe class GpuComputePass : Poolable<GpuComputePass>
     internal SDL_GPUComputePass* Handle;
 #pragma warning restore SA1401
 
+    private bool _isPipelineBound;
+
     /// <summary>
     ///     Gets the <see cref="GpuDevice" /> instance associated with the compute pass.
     /// </summary>
@@ -37,6 +39,7 @@ internal GpuComputePass(GpuDevice device)
     public void BindShader(GpuComputeShader computeShader)
     {
         SDL_BindGPUComputePipeline(Handle, computeShader.HandleTyped);
+        _isPipelineBound = true;
     }
 
     /// <summary>
@@ -50,12 +53,97 @@ public void Dispatch(
         int workGroupsCountY,
         int workGroupsCountZ)
     {
+        if (!_isPipelineBound)
+        {
+            throw new InvalidOperationException("A compute shader must be bound before dispatching.");
+        }
+
         SDL_DispatchGPUCompute(
             Handle, (uint)workGroupsCountX, (uint)workGroupsCountY, (uint)workGroupsCountZ);
     }
 
     /// <summary>
-    ///     Ends the compute pass.
+    ///     Dispatches compute work with parameters set from a buffer.
+    /// </summary>
+    /// <param name="buffer">The buffer containing dispatch parameters.</param>
+    /// <param name="offset">the offset to start reading from the dispatch buffer.</param>
+    public void DispatchIndirect(GpuDataBuffer buffer, int offset)
+    {
+        if (!_isPipelineBound)
+        {
+            throw new InvalidOperationException("A compute shader must be bound before dispatching.");
+        }
+
+        SDL_DispatchGPUComputeIndirect(Handle, buffer.HandleTyped, (uint)offset);
+    }
+
+    /// <summary>
+    ///     Binds storage textures as readonly for use on the compute pipeline.
+    /// </summary>
+    /// <param name="startIndex">Index of the slot to begin binding from.</param>
+    /// <param name="textures">An array of <see cref="GpuTexture"/>s to bind.</param>
+    public void BindStorageTextures(int startIndex, params ReadOnlySpan<GpuTexture> textures)
+    {
+        var handles = stackalloc SDL_GPUTexture*[textures.Length];
+        for (var i = 0; i < textures.Length; i++)
+        {
+            handles[i] = (SDL_GPUTexture*)textures[i].Handle;
+        }
+
+        SDL_BindGPUComputeStorageTextures(
+            Handle,
+            (uint)startIndex,
+            handles,
+            (uint)textures.Length);
+    }
+
+    /// <summary>
+    ///     Binds storage buffers as readonly for use on the compute pipeline.
+    /// </summary>
+    /// <param name="startIndex">Index of the slot to begin binding from.</param>
+    /// <param name="buffers">An array of <see cref="GpuDataBuffer"/>s to bind.</param>
+    public void BindStorageBuffers(int startIndex, params ReadOnlySpan<GpuDataBuffer> buffers)
+    {
+        var handles = stackalloc SDL_GPUBuffer*[buffers.Length];
+        for (var i = 0; i < buffers.Length; i++)
+        {
+            handles[i] = (SDL_GPUBuffer*)buffers[i].Handle;
+        }
+
+        SDL_BindGPUComputeStorageBuffers(
+            Handle,
+            (uint)startIndex,
+            handles,
+            (uint)buffers.Length);
+    }
+
+    /// <summary>
+    ///     Binds texture-sampler pairs for use on the compute shader pipeline.
+    /// </summary>
+    /// <param name="startIndex">Index of the slot to begin binding from.</param>
+    /// <param name="samplers">An array of <see cref="GpuTexture"/> and <see cref="GpuSampler"/> pairs to bind.</param>
+    public void BindSamplers(
+        int startIndex,
+        params ReadOnlySpan<(GpuTexture Texture, GpuSampler Sampler)> samplers)
+    {
+        var bindings = stackalloc SDL_GPUTextureSamplerBinding[samplers.Length];
+        for (var i = 0; i < samplers.Length; i++)
+        {
+            var src = samplers[i];
+            ref var dst = ref bindings[i];
+            dst.texture = (SDL_GPUTexture*)src.Texture.Handle;
+            dst.sampler = (SDL_GPUSampler*)src.Sampler.Handle;
+        }
+
+        SDL_BindGPUComputeSamplers(
+            Handle,
+            (uint)startIndex,
+            bindings,
+            (uint)samplers.Length);
+    }
+
+    /// <summary>
+    ///     Ends a render pass.
     /// </summary>
     /// <exception cref="InvalidOperationException">The associated command buffer was submitted.</exception>
     public void End()
@@ -68,5 +156,6 @@ public void End()
     protected override void Reset()
     {
         Device.EndComputePassTryInternal(this);
+        _isPipelineBound = false;
     }
 }

src/cs/production/SDL/GPU/GpuDevice.cs

@@ -152,7 +152,7 @@ internal GpuDevice(ILogger<GpuDevice> logger, GpuDeviceOptions? options)
         PoolRenderPass = new Pool<GpuRenderPass>(
             _logger, () => new GpuRenderPass(this), "GpuRenderPasses");
         PoolComputePass = new Pool<GpuComputePass>(
-            _logger, () => new GpuComputePass(this), "GpuComputePass");
+            _logger, () => new GpuComputePass(this), "GpuComputePasses");
 
         SupportedShaderFormats = (GpuShaderFormats)(uint)SDL_GetGPUShaderFormats(HandleTyped);
 
@@ -442,6 +442,7 @@ public bool TryCreateGraphicsPipeline(
     /// <summary>
     ///     Attempts to create a new <see cref="GpuDataBuffer" /> instance.
     /// </summary>
+    /// <param name="usage">How the buffer is intended to be used by the client.</param>
     /// <param name="elementCount">
     ///     The maximum number of <typeparamref name="TElement" /> elements the buffer can hold.
     /// </param>
@@ -452,13 +453,14 @@ public bool TryCreateGraphicsPipeline(
     /// <typeparam name="TElement">The type of data buffer element.</typeparam>
     /// <returns><c>true</c> if the data buffer was successfully created; otherwise, <c>false</c>.</returns>
     public bool TryCreateDataBuffer<TElement>(
+        GpuBufferUsageFlags usage,
         int elementCount,
         [NotNullWhen(true)] out GpuDataBuffer? buffer,
         string? name = null)
         where TElement : unmanaged
     {
         var bufferCreateInfo = default(SDL_GPUBufferCreateInfo);
-        bufferCreateInfo.usage = SDL_GPU_BUFFERUSAGE_VERTEX;
+        bufferCreateInfo.usage = (uint)usage;
         bufferCreateInfo.size = (uint)(sizeof(TElement) * elementCount);
         var handle = SDL_CreateGPUBuffer(HandleTyped, &bufferCreateInfo);
         if (handle == null)