wip: render primitive

filament/backend/src/webgpu/WebGPUDriver.cpp

@@ -464,13 +464,21 @@ void WebGPUDriver::createSwapChainHeadlessR(Handle<HwSwapChain> sch, uint32_t wi
         uint32_t height, uint64_t flags) {}
 
 void WebGPUDriver::createVertexBufferInfoR(Handle<HwVertexBufferInfo> vbih, uint8_t bufferCount,
-        uint8_t attributeCount, AttributeArray attributes) {}
+        uint8_t attributeCount, AttributeArray attributes) {
+    constructHandle<WGPUVertexBufferInfo>(vbih, bufferCount, attributeCount, attributes);
+}
 
 void WebGPUDriver::createVertexBufferR(Handle<HwVertexBuffer> vbh, uint32_t vertexCount,
-        Handle<HwVertexBufferInfo> vbih) {}
+        Handle<HwVertexBufferInfo> vbih) {
+    auto* vertexBufferInfo = handleCast<WGPUVertexBufferInfo>(vbih);
+    constructHandle<WGPUVertexBuffer>(vbh, mDevice, vertexCount,vertexBufferInfo->bufferCount, vbih);
+}
 
 void WebGPUDriver::createIndexBufferR(Handle<HwIndexBuffer> ibh, ElementType elementType,
-        uint32_t indexCount, BufferUsage usage) {}
+        uint32_t indexCount, BufferUsage usage) {
+    auto elementSize = (uint8_t)getElementTypeSize(elementType);
+    constructHandle<WGPUIndexBuffer>(ibh, mDevice, elementSize, indexCount);
+}
 
 void WebGPUDriver::createBufferObjectR(Handle<HwBufferObject> boh, uint32_t byteCount,
         BufferObjectBinding bindingType, BufferUsage usage) {}
@@ -503,7 +511,18 @@ void WebGPUDriver::importTextureR(Handle<HwTexture> th, intptr_t id, SamplerType
         uint32_t depth, TextureUsage usage) {}
 
 void WebGPUDriver::createRenderPrimitiveR(Handle<HwRenderPrimitive> rph, Handle<HwVertexBuffer> vbh,
-        Handle<HwIndexBuffer> ibh, PrimitiveType pt) {}
+        Handle<HwIndexBuffer> ibh, PrimitiveType pt) {
+    assert_invariant(mDevice);
+
+    auto* renderPrimitive = handleCast<WGPURenderPrimitive>(rph);
+    auto* vertexBuffer = handleCast<WGPUVertexBuffer>(vbh);
+    auto* indexBuffer = handleCast<WGPUIndexBuffer>(ibh);
+    // auto* vertexBufferInfo = handleCast<WGPUVertexBufferInfo>(vertexBuffer->vbih);
+    // renderPrimitive->setBuffers(vertexBufferInfo, vertexBuffer, indexBuffer);
+    renderPrimitive->vertexBuffer = vertexBuffer;
+    renderPrimitive->indexBuffer = indexBuffer;
+    renderPrimitive->type = pt;
+}
 
 void WebGPUDriver::createProgramR(Handle<HwProgram> ph, Program&& program) {}
 
@@ -715,6 +734,7 @@ void WebGPUDriver::beginRenderPass(Handle<HwRenderTarget> rth, const RenderPassP
     };
     mCommandEncoder = mDevice.CreateCommandEncoder(&commandEncoderDescriptor);
     assert_invariant(mCommandEncoder);
+
     // TODO: Remove this code once WebGPU pipeline is implemented
     static float red = 1.0f;
     if (red - 0.01 > 0) {
@@ -828,6 +848,70 @@ void WebGPUDriver::bindPipeline(PipelineState const& pipelineState) {
 }
 
 void WebGPUDriver::bindRenderPrimitive(Handle<HwRenderPrimitive> rph) {
+    // VulkanCommandBuffer* commands = mCurrentRenderPass.commandBuffer;
+    // VkCommandBuffer cmdbuffer = commands->buffer();
+    // auto prim = resource_ptr<VulkanRenderPrimitive>::cast(&mResourceManager, rph);
+    auto* renderPrimitive = handleCast<WGPURenderPrimitive>(rph);
+    // commands->acquire(prim);
+
+    // This *must* match the VulkanVertexBufferInfo that was bound in bindPipeline(). But we want
+    // to allow to call this before bindPipeline(), so the validation can only happen in draw()
+    auto vbi = handleCast<WGPUVertexBufferInfo>(renderPrimitive->vertexBuffer->vbih);
+    (void) vbi;
+
+    // mRenderPassEncoder.SetVertexBuffer(uint32_t slot, Buffer const& buffer = nullptr, uint64_t offset = 0, uint64_t size = kWholeSize);
+    mRenderPassEncoder.SetIndexBuffer(renderPrimitive->indexBuffer->buffer, renderPrimitive->indexBuffer->indexFormat);
+
+    // uint32_t const bufferCount = vbi->getAttributeCount();
+    // VkDeviceSize const* offsets = vbi->getOffsets();
+    // VkBuffer const* buffers = prim->vertexBuffer->getVkBuffers();
+    //
+    // // Next bind the vertex buffers and index buffer. One potential performance improvement is to
+    // // avoid rebinding these if they are already bound, but since we do not (yet) support subranges
+    // // it would be rare for a client to make consecutive draw calls with the same render primitive.
+    // vkCmdBindVertexBuffers(cmdbuffer, 0, bufferCount, buffers, offsets);
+    // vkCmdBindIndexBuffer(cmdbuffer, prim->indexBuffer->buffer.getGpuBuffer(), 0,
+    //         prim->indexBuffer->indexType);
+
+
+    // METAL
+    // if (UTILS_UNLIKELY(mContext->currentRenderPassAbandoned)) {
+    //     return;
+    // }
+    // FILAMENT_CHECK_PRECONDITION(mContext->currentRenderPassEncoder != nullptr)
+    //         << "bindRenderPrimitive() without a valid command encoder.";
+    //
+    // // Bind the user vertex buffers.
+    // MetalBuffer* vertexBuffers[MAX_VERTEX_BUFFER_COUNT] = {};
+    // size_t vertexBufferOffsets[MAX_VERTEX_BUFFER_COUNT] = {};
+    // size_t maxBufferIndex = 0;
+    //
+    // MetalRenderPrimitive const* const primitive = handle_cast<MetalRenderPrimitive>(rph);
+    // MetalVertexBufferInfo const* const vbi =
+    //         handle_cast<MetalVertexBufferInfo>(primitive->vertexBuffer->vbih);
+    //
+    // mContext->currentRenderPrimitive = rph;
+    //
+    // auto vb = primitive->vertexBuffer;
+    // for (auto m : vbi->bufferMapping) {
+    //     assert_invariant(
+    //             m.bufferArgumentIndex >= USER_VERTEX_BUFFER_BINDING_START &&
+    //             m.bufferArgumentIndex < USER_VERTEX_BUFFER_BINDING_START + MAX_VERTEX_BUFFER_COUNT);
+    //     size_t const vertexBufferIndex = m.bufferArgumentIndex - USER_VERTEX_BUFFER_BINDING_START;
+    //     vertexBuffers[vertexBufferIndex] = vb->buffers[m.sourceBufferIndex];
+    //     maxBufferIndex = std::max(maxBufferIndex, vertexBufferIndex);
+    // }
+    //
+    // const auto bufferCount = maxBufferIndex + 1;
+    // MetalBuffer::bindBuffers(getPendingCommandBuffer(mContext), mContext->currentRenderPassEncoder,
+    //         USER_VERTEX_BUFFER_BINDING_START, MetalBuffer::Stage::VERTEX, vertexBuffers,
+    //         vertexBufferOffsets, bufferCount);
+    //
+    // // Bind the zero buffer, used for missing vertex attributes.
+    // static const char bytes[16] = { 0 };
+    // [mContext->currentRenderPassEncoder setVertexBytes:bytes
+    //                                             length:16
+    //                                            atIndex:ZERO_VERTEX_BUFFER_BINDING];
 }
 
 void WebGPUDriver::draw2(uint32_t indexOffset, uint32_t indexCount, uint32_t instanceCount) {

filament/backend/src/webgpu/WebGPUDriver.h

@@ -112,7 +112,7 @@ private:
     template<typename D, typename B>
     void destructHandle(Handle<B>& handle) noexcept {
         auto* p = mHandleAllocator.handle_cast<D*>(handle);
-        return mHandleAllocator.deallocate(handle, p);
+        mHandleAllocator.deallocate(handle, p);
     }
 };
 

filament/backend/src/webgpu/WebGPUHandles.cpp

@@ -27,18 +27,126 @@ wgpu::Buffer createIndexBuffer(wgpu::Device const& device, uint8_t elementSize,
         .mappedAtCreation = false };
     return device.CreateBuffer(&descriptor);
 }
+
+wgpu::VertexFormat getVertexFormat(filament::backend::ElementType type, bool normalized, bool integer) {
+    using ElementType = filament::backend::ElementType;
+    // using VertexFormat = wgpu::VertexFormat;
+    if (normalized) {
+        switch (type) {
+            // Single Component Types
+            case ElementType::BYTE: return wgpu::VertexFormat::Snorm8;
+            case ElementType::UBYTE: return wgpu::VertexFormat::Unorm8;
+            case ElementType::SHORT: return wgpu::VertexFormat::Snorm16;
+            case ElementType::USHORT: return wgpu::VertexFormat::Unorm16;
+            // Two Component Types
+            case ElementType::BYTE2: return wgpu::VertexFormat::Snorm8x2;
+            case ElementType::UBYTE2: return wgpu::VertexFormat::Unorm8x2;
+            case ElementType::SHORT2: return wgpu::VertexFormat::Snorm16x2;
+            case ElementType::USHORT2: return wgpu::VertexFormat::Unorm16x2;
+            // Three Component Types
+            // There is no vertex format type for 3 byte data in webgpu. Use
+            // 4 byte signed normalized type and ignore the last byte.
+            // TODO: This is to be verified.
+            case ElementType::BYTE3: return wgpu::VertexFormat::Snorm8x4;      // NOT MINSPEC
+            case ElementType::UBYTE3: return wgpu::VertexFormat::Unorm8x4;     // NOT MINSPEC
+            case ElementType::SHORT3: return wgpu::VertexFormat::Snorm16x4;  // NOT MINSPEC
+            case ElementType::USHORT3: return wgpu::VertexFormat::Unorm16x4; // NOT MINSPEC
+            // Four Component Types
+            case ElementType::BYTE4: return wgpu::VertexFormat::Snorm8x4;
+            case ElementType::UBYTE4: return wgpu::VertexFormat::Unorm8x4;
+            case ElementType::SHORT4: return wgpu::VertexFormat::Snorm16x4;
+            case ElementType::USHORT4: return wgpu::VertexFormat::Unorm8x4;
+            default:
+                FILAMENT_CHECK_POSTCONDITION(false) << "Normalized format does not exist.";
+                return wgpu::VertexFormat::Float32x3;
+        }
+    }
+    switch (type) {
+        // Single Component Types
+        // There is no direct alternative for SSCALED in webgpu. Convert them to Float32 directly.
+        // This will result in increased memory on the cpu side.
+        // TODO: Is Float16 acceptable instead with some potential accuracy errors?
+        case ElementType::BYTE: return integer ? wgpu::VertexFormat::Sint8 : wgpu::VertexFormat::Float32;
+        case ElementType::UBYTE: return integer ? wgpu::VertexFormat::Uint8 : wgpu::VertexFormat::Float32;
+        case ElementType::SHORT: return integer ? wgpu::VertexFormat::Sint16 : wgpu::VertexFormat::Float32;
+        case ElementType::USHORT: return integer ? wgpu::VertexFormat::Uint16 : wgpu::VertexFormat::Float32;
+        case ElementType::HALF: return wgpu::VertexFormat::Float16;
+        case ElementType::INT: return wgpu::VertexFormat::Sint32;
+        case ElementType::UINT: return wgpu::VertexFormat::Uint32;
+        case ElementType::FLOAT: return wgpu::VertexFormat::Float32;
+        // Two Component Types
+        case ElementType::BYTE2: return integer ? wgpu::VertexFormat::Sint8x2 : wgpu::VertexFormat::Float32x2;
+        case ElementType::UBYTE2: return integer ? wgpu::VertexFormat::Uint8x2 : wgpu::VertexFormat::Float32x2;
+        case ElementType::SHORT2: return integer ? wgpu::VertexFormat::Sint16x2 : wgpu::VertexFormat::Float32x2;
+        case ElementType::USHORT2: return integer ? wgpu::VertexFormat::Uint16x2 : wgpu::VertexFormat::Float32x2;
+        case ElementType::HALF2: return wgpu::VertexFormat::Float16x2;
+        case ElementType::FLOAT2: return wgpu::VertexFormat::Float32x2;
+        // Three Component Types
+        case ElementType::BYTE3: return wgpu::VertexFormat::Sint8x4;      // NOT MINSPEC
+        case ElementType::UBYTE3: return wgpu::VertexFormat::Uint8x4;     // NOT MINSPEC
+        case ElementType::SHORT3: return wgpu::VertexFormat::Sint16x4;  // NOT MINSPEC
+        case ElementType::USHORT3: return wgpu::VertexFormat::Uint16x4; // NOT MINSPEC
+        case ElementType::HALF3: return wgpu::VertexFormat::Float16x4; // NOT MINSPEC
+        case ElementType::FLOAT3: return wgpu::VertexFormat::Float32x3;
+        // Four Component Types
+        case ElementType::BYTE4: return integer ? wgpu::VertexFormat::Sint8x4 : wgpu::VertexFormat::Float32x4;
+        case ElementType::UBYTE4: return integer ? wgpu::VertexFormat::Uint8x4 : wgpu::VertexFormat::Float32x4;
+        case ElementType::SHORT4: return integer ? wgpu::VertexFormat::Sint16x4 : wgpu::VertexFormat::Float32x4;
+        case ElementType::USHORT4: return integer ? wgpu::VertexFormat::Uint16x4 : wgpu::VertexFormat::Float32x4;
+        case ElementType::HALF4: return wgpu::VertexFormat::Float16x4;
+        case ElementType::FLOAT4: return wgpu::VertexFormat::Float32x4;
+    }
+    FILAMENT_CHECK_POSTCONDITION(false) << "Vertex format should always be defined.";
+    return wgpu::VertexFormat::Float32x3;
+}
+
 } // namespace
 
 namespace filament::backend {
 
+WGPUVertexBufferInfo::WGPUVertexBufferInfo(uint8_t bufferCount, uint8_t attributeCount,
+        AttributeArray const& attributes)
+    : HwVertexBufferInfo(bufferCount, attributeCount),
+      mVertexBufferLayout(bufferCount),
+      mAttributes(bufferCount) {
+    for (uint32_t attribIndex = 0; attribIndex < attributes.size(); attribIndex++) {
+        Attribute attrib = attributes[attribIndex];
+        // Ignore the attributes which are not bind to vertex buffers.
+        if (attrib.buffer == Attribute::BUFFER_UNUSED) {
+            continue;
+        }
+
+        assert_invariant(attrib.buffer < bufferCount);
+        bool const isInteger = attrib.flags & Attribute::FLAG_INTEGER_TARGET;
+        bool const isNormalized = attrib.flags & Attribute::FLAG_NORMALIZED;
+        wgpu::VertexFormat vertexFormat = getVertexFormat(attrib.type, isNormalized, isInteger);
+
+        mAttributes[attrib.buffer].push_back({
+            .format = vertexFormat,
+            .offset = attrib.offset,
+            .shaderLocation = attribIndex,
+        });
+        mVertexBufferLayout[attrib.buffer] = {
+            .arrayStride = attrib.stride,
+        };
+    }
+
+    for (uint32_t bufferIndex = 0; bufferIndex < bufferCount; bufferIndex++) {
+        mVertexBufferLayout[bufferIndex] = {
+            .attributeCount = mAttributes[bufferIndex].size(),
+            .attributes = mAttributes[bufferIndex].data(),
+        };
+    }
+}
 
 WGPUIndexBuffer::WGPUIndexBuffer(wgpu::Device const& device, uint8_t elementSize,
         uint32_t indexCount)
-    : buffer(createIndexBuffer(device, elementSize, indexCount)) {}
+    : buffer(createIndexBuffer(device, elementSize, indexCount)),
+      indexFormat(elementSize == 2 ? wgpu::IndexFormat::Uint16 : wgpu::IndexFormat::Uint32) {}
 
 
 WGPUVertexBuffer::WGPUVertexBuffer(wgpu::Device const &device, uint32_t vextexCount, uint32_t bufferCount,
-                                   Handle<WGPUVertexBufferInfo> vbih)
+                                   Handle<HwVertexBufferInfo> vbih)
         : HwVertexBuffer(vextexCount),
           vbih(vbih),
           buffers(bufferCount) {
@@ -150,4 +258,10 @@ WebGPUDescriptorSetLayout::WebGPUDescriptorSetLayout(DescriptorSetLayout const&
     mLayout = device->CreateBindGroupLayout(&layoutDescriptor);
 }
 WebGPUDescriptorSetLayout::~WebGPUDescriptorSetLayout() {}
+
+void WGPURenderPrimitive::setBuffers(WGPUVertexBufferInfo const* const vbi,
+        WGPUVertexBuffer* vertexBuffer, WGPUIndexBuffer* indexBuffer) {
+
+    }
+
 }// namespace filament::backend

filament/backend/src/webgpu/WebGPUHandles.h

@@ -24,6 +24,7 @@
 #include <backend/Handle.h>
 
 #include <utils/FixedCapacityVector.h>
+// #include <utils/StructureOfArrays.h>
 
 #include <webgpu/webgpu_cpp.h>
 
@@ -32,23 +33,40 @@
 namespace filament::backend {
 
 struct WGPUBufferObject;
-// TODO: Currently WGPUVertexBufferInfo is not used by WebGPU for useful task.
-// Update the struct when used by WebGPU driver.
+
+// VertexBufferInfo contains layout info for Vertex Buffer based on WebGPU structs. In WebGPU each
+// VertexBufferLayout is associated with a single vertex buffer. So number of mVertexBufferLayout
+// is equal to bufferCount. Each VertexBufferLayout can contain multiple VertexAttribute. Bind index
+// of vertex buffer is implicitly calculated by the position of VertexBufferLayout in an array.
 struct WGPUVertexBufferInfo : public HwVertexBufferInfo {
+
     WGPUVertexBufferInfo(uint8_t bufferCount, uint8_t attributeCount,
-            AttributeArray const& attributes)
-        : HwVertexBufferInfo(bufferCount, attributeCount),
-          attributes(attributes) {}
-    AttributeArray attributes;
+            AttributeArray const& attributes);
+    inline  wgpu::VertexBufferLayout const* getVertexBufferLayout() const {
+        return mVertexBufferLayout.data();
+    }
+    inline uint32_t getVertexBufferLayoutSize() const {
+        return mVertexBufferLayout.size();
+    }
+
+    inline wgpu::VertexAttribute const* getVertexAttributeForIndex(uint32_t index) const {
+        return mAttributes[index].data();
+    }
+    inline uint32_t getVertexAttributeSize(uint32_t index) const {
+        return mAttributes[index].size();
+    }
+private:
+    std::vector<wgpu::VertexBufferLayout> mVertexBufferLayout;
+    std::vector<std::vector<wgpu::VertexAttribute>> mAttributes;
 };
 
 struct WGPUVertexBuffer : public HwVertexBuffer {
     WGPUVertexBuffer(wgpu::Device const &device, uint32_t vextexCount, uint32_t bufferCount,
-                     Handle<WGPUVertexBufferInfo> vbih);
+                     Handle<HwVertexBufferInfo> vbih);
 
     void setBuffer(WGPUBufferObject *bufferObject, uint32_t index);
 
-    Handle<WGPUVertexBufferInfo> vbih;
+    Handle<HwVertexBufferInfo> vbih;
     utils::FixedCapacityVector<wgpu::Buffer> buffers;
 };
 
@@ -57,9 +75,10 @@ struct WGPUIndexBuffer : public HwIndexBuffer {
                     uint32_t indexCount);
 
     wgpu::Buffer buffer;
+    wgpu::IndexFormat indexFormat;
 };
 
-// TODO: Currently WGPUVertexBufferInfo is not used by WebGPU for useful task.
+// TODO: Currently WGPUBufferObject is not used by WebGPU for useful task.
 // Update the struct when used by WebGPU driver.
 struct WGPUBufferObject : HwBufferObject {
     WGPUBufferObject(BufferObjectBinding bindingType, uint32_t byteCount);
@@ -67,6 +86,7 @@ struct WGPUBufferObject : HwBufferObject {
     wgpu::Buffer buffer;
     const BufferObjectBinding bufferObjectBinding;
 };
+
 class WebGPUDescriptorSetLayout : public HwDescriptorSetLayout {
 public:
     WebGPUDescriptorSetLayout(DescriptorSetLayout const& layout, wgpu::Device const* device);