raytracinggltf.cpp

/*
 * Vulkan Example - Rendering a glTF model using hardware accelerated ray tracing example (for proper transparency, this sample does frame accumulation)
 *
 * Copyright (C) 2023-2025 by Sascha Willems - www.saschawillems.de
 *
 * This code is licensed under the MIT license (MIT) (http://opensource.org/licenses/MIT)
 */

#include "VulkanRaytracingSample.h"
#define VK_GLTF_MATERIAL_IDS
#include "VulkanglTFModel.h"

class VulkanExample : public VulkanRaytracingSample
{
public:
	AccelerationStructure bottomLevelAS{};
	AccelerationStructure topLevelAS{};

	vks::Buffer vertexBuffer;
	vks::Buffer indexBuffer;
	uint32_t indexCount{ 0 };
	vks::Buffer transformBuffer;
	
	struct GeometryNode {
		uint64_t vertexBufferDeviceAddress;
		uint64_t indexBufferDeviceAddress;
		int32_t textureIndexBaseColor;
		int32_t textureIndexOcclusion;
	};
	vks::Buffer geometryNodesBuffer;

	std::vector<VkRayTracingShaderGroupCreateInfoKHR> shaderGroups{};
	struct ShaderBindingTables {
		ShaderBindingTable raygen;
		ShaderBindingTable miss;
		ShaderBindingTable hit;
	} shaderBindingTables;

	vks::Texture2D texture;

	struct UniformData {
		glm::mat4 viewInverse;
		glm::mat4 projInverse;
		uint32_t frame{ 0 };
	} uniformData;
	std::array<vks::Buffer, maxConcurrentFrames> uniformBuffers;

	VkPipeline pipeline{ VK_NULL_HANDLE };
	VkPipelineLayout pipelineLayout{ VK_NULL_HANDLE };
	VkDescriptorSetLayout descriptorSetLayout{ VK_NULL_HANDLE };
	std::array<VkDescriptorSet, maxConcurrentFrames> descriptorSets{};

	vkglTF::Model model;

	VkPhysicalDeviceDescriptorIndexingFeaturesEXT physicalDeviceDescriptorIndexingFeatures{};

	VulkanExample() : VulkanRaytracingSample()
	{
		title = "Ray tracing glTF model";
		camera.type = Camera::CameraType::lookat;
		camera.setPerspective(60.0f, (float)width / (float)height, 0.1f, 512.0f);
		camera.setRotation(glm::vec3(0.0f, 0.0f, 0.0f));
		camera.setTranslation(glm::vec3(0.0f, -0.1f, -1.0f));

		enableExtensions();

		// Buffer device address requires the 64-bit integer feature to be enabled
		enabledFeatures.shaderInt64 = VK_TRUE;
		// Format for the storage image is decided at runtime, so we can't explicilily state it in the shader
		enabledFeatures.shaderStorageImageReadWithoutFormat = VK_TRUE;
		enabledFeatures.shaderStorageImageWriteWithoutFormat = VK_TRUE;

		enabledDeviceExtensions.push_back(VK_KHR_MAINTENANCE3_EXTENSION_NAME);
		enabledDeviceExtensions.push_back(VK_EXT_DESCRIPTOR_INDEXING_EXTENSION_NAME);
	}

	~VulkanExample()
	{
		if (device) {
			vkDestroyPipeline(device, pipeline, nullptr);
			vkDestroyPipelineLayout(device, pipelineLayout, nullptr);
			vkDestroyDescriptorSetLayout(device, descriptorSetLayout, nullptr);
			deleteStorageImage();
			deleteAccelerationStructure(bottomLevelAS);
			deleteAccelerationStructure(topLevelAS);
			vertexBuffer.destroy();
			indexBuffer.destroy();
			transformBuffer.destroy();
			shaderBindingTables.raygen.destroy();
			shaderBindingTables.miss.destroy();
			shaderBindingTables.hit.destroy();
			geometryNodesBuffer.destroy();
			for (auto& buffer : uniformBuffers) {
				buffer.destroy();
			}
		}
	}

	void createAccelerationStructureBuffer(AccelerationStructure &accelerationStructure, VkAccelerationStructureBuildSizesInfoKHR buildSizeInfo)
	{
		VkBufferCreateInfo bufferCreateInfo{};
		bufferCreateInfo.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO;
		bufferCreateInfo.size = buildSizeInfo.accelerationStructureSize;
		bufferCreateInfo.usage = VK_BUFFER_USAGE_ACCELERATION_STRUCTURE_STORAGE_BIT_KHR | VK_BUFFER_USAGE_SHADER_DEVICE_ADDRESS_BIT;
		VK_CHECK_RESULT(vkCreateBuffer(device, &bufferCreateInfo, nullptr, &accelerationStructure.buffer));
		VkMemoryRequirements memoryRequirements{};
		vkGetBufferMemoryRequirements(device, accelerationStructure.buffer, &memoryRequirements);
		VkMemoryAllocateFlagsInfo memoryAllocateFlagsInfo{};
		memoryAllocateFlagsInfo.sType = VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_FLAGS_INFO;
		memoryAllocateFlagsInfo.flags = VK_MEMORY_ALLOCATE_DEVICE_ADDRESS_BIT_KHR;
		VkMemoryAllocateInfo memoryAllocateInfo{};
		memoryAllocateInfo.sType = VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO;
		memoryAllocateInfo.pNext = &memoryAllocateFlagsInfo;
		memoryAllocateInfo.allocationSize = memoryRequirements.size;
		memoryAllocateInfo.memoryTypeIndex = vulkanDevice->getMemoryType(memoryRequirements.memoryTypeBits, VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT);
		VK_CHECK_RESULT(vkAllocateMemory(device, &memoryAllocateInfo, nullptr, &accelerationStructure.memory));
		VK_CHECK_RESULT(vkBindBufferMemory(device, accelerationStructure.buffer, accelerationStructure.memory, 0));
	}

	/*
		Create the bottom level acceleration structure that contains the scene's actual geometry (vertices, triangles)
	*/
	void createBottomLevelAccelerationStructure()
	{
		// Use transform matrices from the glTF nodes
		std::vector<VkTransformMatrixKHR> transformMatrices{};
		for (auto node : model.linearNodes) {
			if (node->mesh) {
				for (auto primitive : node->mesh->primitives) {
					if (primitive->indexCount > 0) {
						VkTransformMatrixKHR transformMatrix{};
						auto m = glm::mat3x4(glm::transpose(node->getMatrix()));
						memcpy(&transformMatrix, (void*)&m, sizeof(glm::mat3x4));
						transformMatrices.push_back(transformMatrix);
					}
				}
			}
		}

		// Transform buffer
		VK_CHECK_RESULT(vulkanDevice->createBuffer(
			VK_BUFFER_USAGE_SHADER_DEVICE_ADDRESS_BIT | VK_BUFFER_USAGE_ACCELERATION_STRUCTURE_BUILD_INPUT_READ_ONLY_BIT_KHR,
			VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT | VK_MEMORY_PROPERTY_HOST_COHERENT_BIT,
			&transformBuffer,
			static_cast<uint32_t>(transformMatrices.size()) * sizeof(VkTransformMatrixKHR),
			transformMatrices.data()));
		
		// Build
		// One geometry per glTF node, so we can index materials using gl_GeometryIndexEXT
		std::vector<uint32_t> maxPrimitiveCounts{};
		std::vector<VkAccelerationStructureGeometryKHR> geometries{};
		std::vector<VkAccelerationStructureBuildRangeInfoKHR> buildRangeInfos{};
		std::vector<VkAccelerationStructureBuildRangeInfoKHR*> pBuildRangeInfos{};
		std::vector<GeometryNode> geometryNodes{};
		for (auto node : model.linearNodes) {
			if (node->mesh) {
				for (auto primitive : node->mesh->primitives) {
					if (primitive->indexCount > 0) {
						VkDeviceOrHostAddressConstKHR vertexBufferDeviceAddress{};
						VkDeviceOrHostAddressConstKHR indexBufferDeviceAddress{};
						VkDeviceOrHostAddressConstKHR transformBufferDeviceAddress{};

						vertexBufferDeviceAddress.deviceAddress = getBufferDeviceAddress(model.vertices.buffer);// +primitive->firstVertex * sizeof(vkglTF::Vertex);
						indexBufferDeviceAddress.deviceAddress = getBufferDeviceAddress(model.indices.buffer) + primitive->firstIndex * sizeof(uint32_t);
						transformBufferDeviceAddress.deviceAddress = getBufferDeviceAddress(transformBuffer.buffer) + static_cast<uint32_t>(geometryNodes.size()) * sizeof(VkTransformMatrixKHR);

						VkAccelerationStructureGeometryKHR geometry{};
						geometry.sType = VK_STRUCTURE_TYPE_ACCELERATION_STRUCTURE_GEOMETRY_KHR;
						geometry.geometryType = VK_GEOMETRY_TYPE_TRIANGLES_KHR;
						geometry.geometry.triangles.sType = VK_STRUCTURE_TYPE_ACCELERATION_STRUCTURE_GEOMETRY_TRIANGLES_DATA_KHR;
						geometry.geometry.triangles.vertexFormat = VK_FORMAT_R32G32B32_SFLOAT;
						geometry.geometry.triangles.vertexData = vertexBufferDeviceAddress;
						geometry.geometry.triangles.maxVertex = model.vertices.count;
						//geometry.geometry.triangles.maxVertex = primitive->vertexCount;
						geometry.geometry.triangles.vertexStride = sizeof(vkglTF::Vertex);
						geometry.geometry.triangles.indexType = VK_INDEX_TYPE_UINT32;
						geometry.geometry.triangles.indexData = indexBufferDeviceAddress;
						geometry.geometry.triangles.transformData = transformBufferDeviceAddress;
						geometries.push_back(geometry);
						maxPrimitiveCounts.push_back(primitive->indexCount / 3);

						VkAccelerationStructureBuildRangeInfoKHR buildRangeInfo{};
						buildRangeInfo.firstVertex = 0;
						buildRangeInfo.primitiveOffset = 0; // primitive->firstIndex * sizeof(uint32_t);
						buildRangeInfo.primitiveCount = primitive->indexCount / 3;
						buildRangeInfo.transformOffset = 0;
						buildRangeInfos.push_back(buildRangeInfo);

						GeometryNode geometryNode{};
						geometryNode.vertexBufferDeviceAddress = vertexBufferDeviceAddress.deviceAddress;
						geometryNode.indexBufferDeviceAddress = indexBufferDeviceAddress.deviceAddress;
						geometryNode.textureIndexBaseColor = primitive->material.baseColorTexture->index;
						geometryNode.textureIndexOcclusion = primitive->material.occlusionTexture ? primitive->material.occlusionTexture->index : -1;
						geometryNodes.push_back(geometryNode);
					}
				}
			}
		}
		for (auto& rangeInfo : buildRangeInfos) {
			pBuildRangeInfos.push_back(&rangeInfo);
		}

		vks::Buffer stagingBuffer;

		VK_CHECK_RESULT(vulkanDevice->createBuffer(
			VK_BUFFER_USAGE_TRANSFER_SRC_BIT,
			VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT | VK_MEMORY_PROPERTY_HOST_COHERENT_BIT,
			&stagingBuffer,
			static_cast<uint32_t>(geometryNodes.size()) * sizeof(GeometryNode),
			geometryNodes.data()));

		VK_CHECK_RESULT(vulkanDevice->createBuffer(
			VK_BUFFER_USAGE_SHADER_DEVICE_ADDRESS_BIT | VK_BUFFER_USAGE_STORAGE_BUFFER_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT,
			VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT,
			&geometryNodesBuffer,
			static_cast<uint32_t>(geometryNodes.size()) * sizeof(GeometryNode)));

		vulkanDevice->copyBuffer(&stagingBuffer, &geometryNodesBuffer, queue);

		stagingBuffer.destroy();
	
		// Get size info
		VkAccelerationStructureBuildGeometryInfoKHR accelerationStructureBuildGeometryInfo{};
		accelerationStructureBuildGeometryInfo.sType = VK_STRUCTURE_TYPE_ACCELERATION_STRUCTURE_BUILD_GEOMETRY_INFO_KHR;
		accelerationStructureBuildGeometryInfo.type = VK_ACCELERATION_STRUCTURE_TYPE_BOTTOM_LEVEL_KHR;
		accelerationStructureBuildGeometryInfo.flags = VK_BUILD_ACCELERATION_STRUCTURE_PREFER_FAST_TRACE_BIT_KHR;
		accelerationStructureBuildGeometryInfo.geometryCount = static_cast<uint32_t>(geometries.size());
		accelerationStructureBuildGeometryInfo.pGeometries = geometries.data();
		
		VkAccelerationStructureBuildSizesInfoKHR accelerationStructureBuildSizesInfo{};
		accelerationStructureBuildSizesInfo.sType = VK_STRUCTURE_TYPE_ACCELERATION_STRUCTURE_BUILD_SIZES_INFO_KHR;
		vkGetAccelerationStructureBuildSizesKHR(
			device,
			VK_ACCELERATION_STRUCTURE_BUILD_TYPE_DEVICE_KHR,
			&accelerationStructureBuildGeometryInfo,
			maxPrimitiveCounts.data(),
			&accelerationStructureBuildSizesInfo);

		createAccelerationStructureBuffer(bottomLevelAS, accelerationStructureBuildSizesInfo);

		VkAccelerationStructureCreateInfoKHR accelerationStructureCreateInfo{};
		accelerationStructureCreateInfo.sType = VK_STRUCTURE_TYPE_ACCELERATION_STRUCTURE_CREATE_INFO_KHR;
		accelerationStructureCreateInfo.buffer = bottomLevelAS.buffer;
		accelerationStructureCreateInfo.size = accelerationStructureBuildSizesInfo.accelerationStructureSize;
		accelerationStructureCreateInfo.type = VK_ACCELERATION_STRUCTURE_TYPE_BOTTOM_LEVEL_KHR;
		vkCreateAccelerationStructureKHR(device, &accelerationStructureCreateInfo, nullptr, &bottomLevelAS.handle);

		// Create a small scratch buffer used during build of the bottom level acceleration structure
		ScratchBuffer scratchBuffer = createScratchBuffer(accelerationStructureBuildSizesInfo.buildScratchSize);

		accelerationStructureBuildGeometryInfo.mode = VK_BUILD_ACCELERATION_STRUCTURE_MODE_BUILD_KHR;
		accelerationStructureBuildGeometryInfo.dstAccelerationStructure = bottomLevelAS.handle;
		accelerationStructureBuildGeometryInfo.scratchData.deviceAddress = scratchBuffer.deviceAddress;

		const VkAccelerationStructureBuildRangeInfoKHR* buildOffsetInfo = buildRangeInfos.data();

		// Build the acceleration structure on the device via a one-time command buffer submission
		// Some implementations may support acceleration structure building on the host (VkPhysicalDeviceAccelerationStructureFeaturesKHR->accelerationStructureHostCommands), but we prefer device builds
		VkCommandBuffer commandBuffer = vulkanDevice->createCommandBuffer(VK_COMMAND_BUFFER_LEVEL_PRIMARY, true);
		vkCmdBuildAccelerationStructuresKHR(
			commandBuffer,
			1,
			&accelerationStructureBuildGeometryInfo,
			pBuildRangeInfos.data());
		vulkanDevice->flushCommandBuffer(commandBuffer, queue);

		VkAccelerationStructureDeviceAddressInfoKHR accelerationDeviceAddressInfo{};
		accelerationDeviceAddressInfo.sType = VK_STRUCTURE_TYPE_ACCELERATION_STRUCTURE_DEVICE_ADDRESS_INFO_KHR;
		accelerationDeviceAddressInfo.accelerationStructure = bottomLevelAS.handle;
		bottomLevelAS.deviceAddress = vkGetAccelerationStructureDeviceAddressKHR(device, &accelerationDeviceAddressInfo);

		deleteScratchBuffer(scratchBuffer);
	}

	/*
		The top level acceleration structure contains the scene's object instances
	*/
	void createTopLevelAccelerationStructure()
	{
		// We flip the matrix [1][1] = -1.0f to accomodate for the glTF up vector
		VkTransformMatrixKHR transformMatrix = {
			1.0f, 0.0f, 0.0f, 0.0f,
			0.0f, -1.0f, 0.0f, 0.0f,
			0.0f, 0.0f, 1.0f, 0.0f };

		VkAccelerationStructureInstanceKHR instance{};
		instance.transform = transformMatrix;
		instance.instanceCustomIndex = 0;
		instance.mask = 0xFF;
		instance.instanceShaderBindingTableRecordOffset = 0;
		instance.flags = VK_GEOMETRY_INSTANCE_TRIANGLE_FACING_CULL_DISABLE_BIT_KHR;
		instance.accelerationStructureReference = bottomLevelAS.deviceAddress;

		// Buffer for instance data
		vks::Buffer instancesBuffer;
		VK_CHECK_RESULT(vulkanDevice->createBuffer(
			VK_BUFFER_USAGE_SHADER_DEVICE_ADDRESS_BIT | VK_BUFFER_USAGE_ACCELERATION_STRUCTURE_BUILD_INPUT_READ_ONLY_BIT_KHR,
			VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT | VK_MEMORY_PROPERTY_HOST_COHERENT_BIT,
			&instancesBuffer,
			sizeof(VkAccelerationStructureInstanceKHR),
			&instance));

		VkDeviceOrHostAddressConstKHR instanceDataDeviceAddress{};
		instanceDataDeviceAddress.deviceAddress = getBufferDeviceAddress(instancesBuffer.buffer);

		VkAccelerationStructureGeometryKHR accelerationStructureGeometry{};
		accelerationStructureGeometry.sType = VK_STRUCTURE_TYPE_ACCELERATION_STRUCTURE_GEOMETRY_KHR;
		accelerationStructureGeometry.geometryType = VK_GEOMETRY_TYPE_INSTANCES_KHR;
		accelerationStructureGeometry.flags = VK_GEOMETRY_OPAQUE_BIT_KHR;
		accelerationStructureGeometry.geometry.instances.sType = VK_STRUCTURE_TYPE_ACCELERATION_STRUCTURE_GEOMETRY_INSTANCES_DATA_KHR;
		accelerationStructureGeometry.geometry.instances.arrayOfPointers = VK_FALSE;
		accelerationStructureGeometry.geometry.instances.data = instanceDataDeviceAddress;

		// Get size info
		/*
		The pSrcAccelerationStructure, dstAccelerationStructure, and mode members of pBuildInfo are ignored. Any VkDeviceOrHostAddressKHR members of pBuildInfo are ignored by this command, except that the hostAddress member of VkAccelerationStructureGeometryTrianglesDataKHR::transformData will be examined to check if it is NULL.*
		*/
		VkAccelerationStructureBuildGeometryInfoKHR accelerationStructureBuildGeometryInfo{};
		accelerationStructureBuildGeometryInfo.sType = VK_STRUCTURE_TYPE_ACCELERATION_STRUCTURE_BUILD_GEOMETRY_INFO_KHR;
		accelerationStructureBuildGeometryInfo.type = VK_ACCELERATION_STRUCTURE_TYPE_TOP_LEVEL_KHR;
		accelerationStructureBuildGeometryInfo.flags = VK_BUILD_ACCELERATION_STRUCTURE_PREFER_FAST_TRACE_BIT_KHR;
		accelerationStructureBuildGeometryInfo.geometryCount = 1;
		accelerationStructureBuildGeometryInfo.pGeometries = &accelerationStructureGeometry;

		uint32_t primitive_count = 1;

		VkAccelerationStructureBuildSizesInfoKHR accelerationStructureBuildSizesInfo{};
		accelerationStructureBuildSizesInfo.sType = VK_STRUCTURE_TYPE_ACCELERATION_STRUCTURE_BUILD_SIZES_INFO_KHR;
		vkGetAccelerationStructureBuildSizesKHR(
			device, 
			VK_ACCELERATION_STRUCTURE_BUILD_TYPE_DEVICE_KHR,
			&accelerationStructureBuildGeometryInfo,
			&primitive_count,
			&accelerationStructureBuildSizesInfo);

		createAccelerationStructureBuffer(topLevelAS, accelerationStructureBuildSizesInfo);

		VkAccelerationStructureCreateInfoKHR accelerationStructureCreateInfo{};
		accelerationStructureCreateInfo.sType = VK_STRUCTURE_TYPE_ACCELERATION_STRUCTURE_CREATE_INFO_KHR;
		accelerationStructureCreateInfo.buffer = topLevelAS.buffer;
		accelerationStructureCreateInfo.size = accelerationStructureBuildSizesInfo.accelerationStructureSize;
		accelerationStructureCreateInfo.type = VK_ACCELERATION_STRUCTURE_TYPE_TOP_LEVEL_KHR;
		vkCreateAccelerationStructureKHR(device, &accelerationStructureCreateInfo, nullptr, &topLevelAS.handle);

		// Create a small scratch buffer used during build of the top level acceleration structure
		ScratchBuffer scratchBuffer = createScratchBuffer(accelerationStructureBuildSizesInfo.buildScratchSize);

		VkAccelerationStructureBuildGeometryInfoKHR accelerationBuildGeometryInfo{};
		accelerationBuildGeometryInfo.sType = VK_STRUCTURE_TYPE_ACCELERATION_STRUCTURE_BUILD_GEOMETRY_INFO_KHR;
		accelerationBuildGeometryInfo.type = VK_ACCELERATION_STRUCTURE_TYPE_TOP_LEVEL_KHR;
		accelerationBuildGeometryInfo.flags = VK_BUILD_ACCELERATION_STRUCTURE_PREFER_FAST_TRACE_BIT_KHR;
		accelerationBuildGeometryInfo.mode = VK_BUILD_ACCELERATION_STRUCTURE_MODE_BUILD_KHR;
		accelerationBuildGeometryInfo.dstAccelerationStructure = topLevelAS.handle;
		accelerationBuildGeometryInfo.geometryCount = 1;
		accelerationBuildGeometryInfo.pGeometries = &accelerationStructureGeometry;
		accelerationBuildGeometryInfo.scratchData.deviceAddress = scratchBuffer.deviceAddress;

		VkAccelerationStructureBuildRangeInfoKHR accelerationStructureBuildRangeInfo{};
		accelerationStructureBuildRangeInfo.primitiveCount = 1;
		accelerationStructureBuildRangeInfo.primitiveOffset = 0;
		accelerationStructureBuildRangeInfo.firstVertex = 0;
		accelerationStructureBuildRangeInfo.transformOffset = 0;
		std::vector<VkAccelerationStructureBuildRangeInfoKHR*> accelerationBuildStructureRangeInfos = { &accelerationStructureBuildRangeInfo };

		// Build the acceleration structure on the device via a one-time command buffer submission
		// Some implementations may support acceleration structure building on the host (VkPhysicalDeviceAccelerationStructureFeaturesKHR->accelerationStructureHostCommands), but we prefer device builds
		VkCommandBuffer commandBuffer = vulkanDevice->createCommandBuffer(VK_COMMAND_BUFFER_LEVEL_PRIMARY, true);
		vkCmdBuildAccelerationStructuresKHR(
			commandBuffer,
			1,
			&accelerationBuildGeometryInfo,
			accelerationBuildStructureRangeInfos.data());
		vulkanDevice->flushCommandBuffer(commandBuffer, queue);

		VkAccelerationStructureDeviceAddressInfoKHR accelerationDeviceAddressInfo{};
		accelerationDeviceAddressInfo.sType = VK_STRUCTURE_TYPE_ACCELERATION_STRUCTURE_DEVICE_ADDRESS_INFO_KHR;
		accelerationDeviceAddressInfo.accelerationStructure = topLevelAS.handle;

		deleteScratchBuffer(scratchBuffer);
		instancesBuffer.destroy();
	}

	/*
		Create the Shader Binding Tables that binds the programs and top-level acceleration structure

		SBT Layout used in this sample:

			/-----------\
			| raygen    |
			|-----------|
			| miss + shadow     |
			|-----------|
			| hit + any |
			\-----------/

	*/
	void createShaderBindingTables() {
		const uint32_t handleSize = rayTracingPipelineProperties.shaderGroupHandleSize;
		const uint32_t handleSizeAligned = vks::tools::alignedSize(rayTracingPipelineProperties.shaderGroupHandleSize, rayTracingPipelineProperties.shaderGroupHandleAlignment);
		const uint32_t groupCount = static_cast<uint32_t>(shaderGroups.size());
		const uint32_t sbtSize = groupCount * handleSizeAligned;

		std::vector<uint8_t> shaderHandleStorage(sbtSize);
		VK_CHECK_RESULT(vkGetRayTracingShaderGroupHandlesKHR(device, pipeline, 0, groupCount, sbtSize, shaderHandleStorage.data()));

		createShaderBindingTable(shaderBindingTables.raygen, 1);
		createShaderBindingTable(shaderBindingTables.miss, 2);
		createShaderBindingTable(shaderBindingTables.hit, 1);

		// Copy handles
		memcpy(shaderBindingTables.raygen.mapped, shaderHandleStorage.data(), handleSize);
		// We are using two miss shaders, so we need to get two handles for the miss shader binding table
		memcpy(shaderBindingTables.miss.mapped, shaderHandleStorage.data() + handleSizeAligned, handleSize * 2);
		memcpy(shaderBindingTables.hit.mapped, shaderHandleStorage.data() + handleSizeAligned * 3, handleSize);
	}

	/*
		Create our ray tracing pipeline
	*/
	void createRayTracingPipeline()
	{
		const uint32_t imageCount = static_cast<uint32_t>(model.textures.size());

		std::vector<VkDescriptorSetLayoutBinding> setLayoutBindings = {
			// Binding 0: Top level acceleration structure
			vks::initializers::descriptorSetLayoutBinding(VK_DESCRIPTOR_TYPE_ACCELERATION_STRUCTURE_KHR, VK_SHADER_STAGE_RAYGEN_BIT_KHR | VK_SHADER_STAGE_CLOSEST_HIT_BIT_KHR, 0),
			// Binding 1: Ray tracing result image
			vks::initializers::descriptorSetLayoutBinding(VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, VK_SHADER_STAGE_RAYGEN_BIT_KHR, 1),
			// Binding 2: Uniform buffer
			vks::initializers::descriptorSetLayoutBinding(VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER, VK_SHADER_STAGE_RAYGEN_BIT_KHR | VK_SHADER_STAGE_CLOSEST_HIT_BIT_KHR | VK_SHADER_STAGE_MISS_BIT_KHR, 2),
			// Binding 3: Texture image
			vks::initializers::descriptorSetLayoutBinding(VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER, VK_SHADER_STAGE_CLOSEST_HIT_BIT_KHR | VK_SHADER_STAGE_ANY_HIT_BIT_KHR, 3),
			// Binding 4: Geometry node information SSBO
			vks::initializers::descriptorSetLayoutBinding(VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, VK_SHADER_STAGE_CLOSEST_HIT_BIT_KHR | VK_SHADER_STAGE_ANY_HIT_BIT_KHR, 4),
			// Binding 5: All images used by the glTF model
			vks::initializers::descriptorSetLayoutBinding(VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER, VK_SHADER_STAGE_CLOSEST_HIT_BIT_KHR | VK_SHADER_STAGE_ANY_HIT_BIT_KHR, 5, imageCount)
		};

		// Unbound set
		VkDescriptorSetLayoutBindingFlagsCreateInfoEXT setLayoutBindingFlags{};
		setLayoutBindingFlags.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_LAYOUT_BINDING_FLAGS_CREATE_INFO_EXT;
		setLayoutBindingFlags.bindingCount = 6;
		std::vector<VkDescriptorBindingFlagsEXT> descriptorBindingFlags = {
			0,
			0,
			0,
			0,
			0,
			VK_DESCRIPTOR_BINDING_VARIABLE_DESCRIPTOR_COUNT_BIT_EXT
		};
		setLayoutBindingFlags.pBindingFlags = descriptorBindingFlags.data();

		VkDescriptorSetLayoutCreateInfo descriptorSetLayoutCI = vks::initializers::descriptorSetLayoutCreateInfo(setLayoutBindings);
		descriptorSetLayoutCI.pNext = &setLayoutBindingFlags;
		VK_CHECK_RESULT(vkCreateDescriptorSetLayout(device, &descriptorSetLayoutCI, nullptr, &descriptorSetLayout));

		VkPipelineLayoutCreateInfo pipelineLayoutCI = vks::initializers::pipelineLayoutCreateInfo(&descriptorSetLayout, 1);
		VK_CHECK_RESULT(vkCreatePipelineLayout(device, &pipelineLayoutCI, nullptr, &pipelineLayout));

		/*
			Setup ray tracing shader groups
		*/
		std::vector<VkPipelineShaderStageCreateInfo> shaderStages;

		// Ray generation group
		{
			shaderStages.push_back(loadShader(getShadersPath() + "raytracinggltf/raygen.rgen.spv", VK_SHADER_STAGE_RAYGEN_BIT_KHR));
			VkRayTracingShaderGroupCreateInfoKHR shaderGroup{};
			shaderGroup.sType = VK_STRUCTURE_TYPE_RAY_TRACING_SHADER_GROUP_CREATE_INFO_KHR;
			shaderGroup.type = VK_RAY_TRACING_SHADER_GROUP_TYPE_GENERAL_KHR;
			shaderGroup.generalShader = static_cast<uint32_t>(shaderStages.size()) - 1;
			shaderGroup.closestHitShader = VK_SHADER_UNUSED_KHR;
			shaderGroup.anyHitShader = VK_SHADER_UNUSED_KHR;
			shaderGroup.intersectionShader = VK_SHADER_UNUSED_KHR;
			shaderGroups.push_back(shaderGroup);
		}

		// Miss group
		{
			shaderStages.push_back(loadShader(getShadersPath() + "raytracinggltf/miss.rmiss.spv", VK_SHADER_STAGE_MISS_BIT_KHR));
			VkRayTracingShaderGroupCreateInfoKHR shaderGroup{};
			shaderGroup.sType = VK_STRUCTURE_TYPE_RAY_TRACING_SHADER_GROUP_CREATE_INFO_KHR;
			shaderGroup.type = VK_RAY_TRACING_SHADER_GROUP_TYPE_GENERAL_KHR;
			shaderGroup.generalShader = static_cast<uint32_t>(shaderStages.size()) - 1;
			shaderGroup.closestHitShader = VK_SHADER_UNUSED_KHR;
			shaderGroup.anyHitShader = VK_SHADER_UNUSED_KHR;
			shaderGroup.intersectionShader = VK_SHADER_UNUSED_KHR;
			shaderGroups.push_back(shaderGroup);
			// Second shader for shadows
			shaderStages.push_back(loadShader(getShadersPath() + "raytracinggltf/shadow.rmiss.spv", VK_SHADER_STAGE_MISS_BIT_KHR));
			shaderGroup.generalShader = static_cast<uint32_t>(shaderStages.size()) - 1;
			shaderGroups.push_back(shaderGroup);
		}

		// Closest hit group for doing texture lookups
		{
			shaderStages.push_back(loadShader(getShadersPath() + "raytracinggltf/closesthit.rchit.spv", VK_SHADER_STAGE_CLOSEST_HIT_BIT_KHR));
			VkRayTracingShaderGroupCreateInfoKHR shaderGroup{};
			shaderGroup.sType = VK_STRUCTURE_TYPE_RAY_TRACING_SHADER_GROUP_CREATE_INFO_KHR;
			shaderGroup.type = VK_RAY_TRACING_SHADER_GROUP_TYPE_TRIANGLES_HIT_GROUP_KHR;
			shaderGroup.generalShader = VK_SHADER_UNUSED_KHR;
			shaderGroup.closestHitShader = static_cast<uint32_t>(shaderStages.size()) - 1;
			shaderGroup.intersectionShader = VK_SHADER_UNUSED_KHR;
			// This group also uses an anyhit shader for doing transparency (see anyhit.rahit for details)
			shaderStages.push_back(loadShader(getShadersPath() + "raytracinggltf/anyhit.rahit.spv", VK_SHADER_STAGE_ANY_HIT_BIT_KHR));
			shaderGroup.anyHitShader = static_cast<uint32_t>(shaderStages.size()) - 1;
			shaderGroups.push_back(shaderGroup);
		}

		/*
			Create the ray tracing pipeline
		*/
		VkRayTracingPipelineCreateInfoKHR rayTracingPipelineCI{};
		rayTracingPipelineCI.sType = VK_STRUCTURE_TYPE_RAY_TRACING_PIPELINE_CREATE_INFO_KHR;
		rayTracingPipelineCI.stageCount = static_cast<uint32_t>(shaderStages.size());
		rayTracingPipelineCI.pStages = shaderStages.data();
		rayTracingPipelineCI.groupCount = static_cast<uint32_t>(shaderGroups.size());
		rayTracingPipelineCI.pGroups = shaderGroups.data();
		rayTracingPipelineCI.maxPipelineRayRecursionDepth = 1;
		rayTracingPipelineCI.layout = pipelineLayout;
		VK_CHECK_RESULT(vkCreateRayTracingPipelinesKHR(device, VK_NULL_HANDLE, VK_NULL_HANDLE, 1, &rayTracingPipelineCI, nullptr, &pipeline));
	}

	/*
		Create the descriptor sets used for the ray tracing dispatch
	*/
	void createDescriptorSets()
	{
		uint32_t imageCount = static_cast<uint32_t>(model.textures.size());
		std::vector<VkDescriptorPoolSize> poolSizes = {
			{ VK_DESCRIPTOR_TYPE_ACCELERATION_STRUCTURE_KHR, maxConcurrentFrames },
			{ VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, maxConcurrentFrames },
			{ VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER, maxConcurrentFrames },
			{ VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER, maxConcurrentFrames },
			{ VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, maxConcurrentFrames },
			{ VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER, static_cast<uint32_t>(model.textures.size()) * maxConcurrentFrames }
		};
		VkDescriptorPoolCreateInfo descriptorPoolCreateInfo = vks::initializers::descriptorPoolCreateInfo(poolSizes, maxConcurrentFrames);
		VK_CHECK_RESULT(vkCreateDescriptorPool(device, &descriptorPoolCreateInfo, nullptr, &descriptorPool));

		VkDescriptorSetVariableDescriptorCountAllocateInfoEXT variableDescriptorCountAllocInfo{};
		uint32_t variableDescCounts[] = { imageCount };
		variableDescriptorCountAllocInfo.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_VARIABLE_DESCRIPTOR_COUNT_ALLOCATE_INFO_EXT;
		variableDescriptorCountAllocInfo.descriptorSetCount = 1;
		variableDescriptorCountAllocInfo.pDescriptorCounts = variableDescCounts;

		// Sets per frame, just like the buffers themselves
		// Acceleration structure and images do not need to be duplicated per frame, we use the same for each descriptor to keep things simple
		VkDescriptorSetAllocateInfo descriptorSetAllocateInfo = vks::initializers::descriptorSetAllocateInfo(descriptorPool, &descriptorSetLayout, 1);
		// Required for the variable no. of images used by the glTF model
		descriptorSetAllocateInfo.pNext = &variableDescriptorCountAllocInfo;
		for (auto i = 0; i < maxConcurrentFrames; i++) {
			VK_CHECK_RESULT(vkAllocateDescriptorSets(device, &descriptorSetAllocateInfo, &descriptorSets[i]));

			VkWriteDescriptorSetAccelerationStructureKHR descriptorAccelerationStructureInfo = vks::initializers::writeDescriptorSetAccelerationStructureKHR();
			descriptorAccelerationStructureInfo.accelerationStructureCount = 1;
			descriptorAccelerationStructureInfo.pAccelerationStructures = &topLevelAS.handle;

			VkWriteDescriptorSet accelerationStructureWrite{};
			accelerationStructureWrite.sType = VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET;
			// The specialized acceleration structure descriptor has to be chained
			accelerationStructureWrite.pNext = &descriptorAccelerationStructureInfo;
			accelerationStructureWrite.dstSet = descriptorSets[i];
			accelerationStructureWrite.dstBinding = 0;
			accelerationStructureWrite.descriptorCount = 1;
			accelerationStructureWrite.descriptorType = VK_DESCRIPTOR_TYPE_ACCELERATION_STRUCTURE_KHR;

			VkDescriptorImageInfo storageImageDescriptor{ VK_NULL_HANDLE, storageImage.view, VK_IMAGE_LAYOUT_GENERAL };

			std::vector<VkWriteDescriptorSet> writeDescriptorSets = {
				// Binding 0: Top level acceleration structure
				accelerationStructureWrite,
				// Binding 1: Ray tracing result image
				vks::initializers::writeDescriptorSet(descriptorSets[i], VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, 1, &storageImageDescriptor),
				// Binding 2: Uniform data
				vks::initializers::writeDescriptorSet(descriptorSets[i], VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER, 2, &uniformBuffers[i].descriptor),
				// Binding 4: Geometry node information SSBO
				vks::initializers::writeDescriptorSet(descriptorSets[i], VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, 4, &geometryNodesBuffer.descriptor),
			};

			// Image descriptors for the variable no. of images of the glTF model
			std::vector<VkDescriptorImageInfo> textureDescriptors{};
			for (auto texture : model.textures) {
				VkDescriptorImageInfo descriptor{};
				descriptor.imageLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL;
				descriptor.sampler = texture.sampler;
				descriptor.imageView = texture.view;
				textureDescriptors.push_back(descriptor);
			}

			VkWriteDescriptorSet writeDescriptorImgArray{};
			writeDescriptorImgArray.sType = VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET;
			writeDescriptorImgArray.dstBinding = 5;
			writeDescriptorImgArray.descriptorType = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER;
			writeDescriptorImgArray.descriptorCount = imageCount;
			writeDescriptorImgArray.dstSet = descriptorSets[i];
			writeDescriptorImgArray.pImageInfo = textureDescriptors.data();
			writeDescriptorSets.push_back(writeDescriptorImgArray);

			vkUpdateDescriptorSets(device, static_cast<uint32_t>(writeDescriptorSets.size()), writeDescriptorSets.data(), 0, VK_NULL_HANDLE);
		}
	}

	/*
		Create the uniform buffer used to pass matrices to the ray tracing ray generation shader
	*/
	void createUniformBuffer()
	{
		for (auto& buffer : uniformBuffers) {
			VK_CHECK_RESULT(vulkanDevice->createBuffer(VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT, VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT | VK_MEMORY_PROPERTY_HOST_COHERENT_BIT, &buffer, sizeof(UniformData), &uniformData));
			VK_CHECK_RESULT(buffer.map());
		}
	}

	/*
		If the window has been resized, we need to recreate the storage image and it's descriptor
	*/
	void handleResize()
	{
		// Recreate image
		createStorageImage(swapChain.colorFormat, { width, height, 1 });
		// Update descriptors
		VkDescriptorImageInfo storageImageDescriptor{ VK_NULL_HANDLE, storageImage.view, VK_IMAGE_LAYOUT_GENERAL };
		for (auto i = 0; i < maxConcurrentFrames; i++) {
			VkWriteDescriptorSet resultImageWrite = vks::initializers::writeDescriptorSet(descriptorSets[i], VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, 1, &storageImageDescriptor);
			vkUpdateDescriptorSets(device, 1, &resultImageWrite, 0, VK_NULL_HANDLE);
		}
		resized = false;
	}

	void updateUniformBuffers()
	{
		uniformData.projInverse = glm::inverse(camera.matrices.perspective);
		uniformData.viewInverse = glm::inverse(camera.matrices.view);
		// This value is used to accumulate multiple frames into the finale picture
		// It's required as ray tracing needs to do multiple passes for transparency
		// In this sample we use noise offset by this frame index to shoot rays for transparency into different directions
		// Once enough frames with random ray directions have been accumulated, it looks like proper transparency
		uniformData.frame++;
		memcpy(uniformBuffers[currentBuffer].mapped, &uniformData, sizeof(UniformData));
	}

	void getEnabledFeatures()
	{
		// Enable features required for ray tracing using feature chaining via pNext		
		enabledBufferDeviceAddresFeatures.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_BUFFER_DEVICE_ADDRESS_FEATURES;
		enabledBufferDeviceAddresFeatures.bufferDeviceAddress = VK_TRUE;

		enabledRayTracingPipelineFeatures.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_RAY_TRACING_PIPELINE_FEATURES_KHR;
		enabledRayTracingPipelineFeatures.rayTracingPipeline = VK_TRUE;
		enabledRayTracingPipelineFeatures.pNext = &enabledBufferDeviceAddresFeatures;

		enabledAccelerationStructureFeatures.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_ACCELERATION_STRUCTURE_FEATURES_KHR;
		enabledAccelerationStructureFeatures.accelerationStructure = VK_TRUE;
		enabledAccelerationStructureFeatures.pNext = &enabledRayTracingPipelineFeatures;

		physicalDeviceDescriptorIndexingFeatures.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_DESCRIPTOR_INDEXING_FEATURES_EXT;
		physicalDeviceDescriptorIndexingFeatures.shaderSampledImageArrayNonUniformIndexing = VK_TRUE;
		physicalDeviceDescriptorIndexingFeatures.runtimeDescriptorArray = VK_TRUE;
		physicalDeviceDescriptorIndexingFeatures.descriptorBindingVariableDescriptorCount = VK_TRUE;
		physicalDeviceDescriptorIndexingFeatures.pNext = &enabledAccelerationStructureFeatures;

		deviceCreatepNextChain = &physicalDeviceDescriptorIndexingFeatures;

		enabledFeatures.samplerAnisotropy = VK_TRUE;
	}

	void loadAssets()
	{
		vkglTF::memoryPropertyFlags = VK_BUFFER_USAGE_ACCELERATION_STRUCTURE_BUILD_INPUT_READ_ONLY_BIT_KHR | VK_BUFFER_USAGE_SHADER_DEVICE_ADDRESS_BIT | VK_BUFFER_USAGE_STORAGE_BUFFER_BIT;
		model.loadFromFile(getAssetPath() + "models/FlightHelmet/glTF/FlightHelmet.gltf", vulkanDevice, queue);
	}

	void prepare()
	{
		VulkanRaytracingSample::prepare();

		loadAssets();

		// Create the acceleration structures used to render the ray traced scene
		createBottomLevelAccelerationStructure();
		createTopLevelAccelerationStructure();

		createStorageImage(swapChain.colorFormat, { width, height, 1 });
		createUniformBuffer();
		createRayTracingPipeline();
		createShaderBindingTables();
		createDescriptorSets();
		prepared = true;
	}

	void buildCommandBuffer()
	{
		if (resized)
		{
			handleResize();
		}

		VkCommandBuffer cmdBuffer = drawCmdBuffers[currentBuffer];
		
		VkCommandBufferBeginInfo cmdBufInfo = vks::initializers::commandBufferBeginInfo();

		VkImageSubresourceRange subresourceRange = { VK_IMAGE_ASPECT_COLOR_BIT, 0, 1, 0, 1 };

		VK_CHECK_RESULT(vkBeginCommandBuffer(cmdBuffer, &cmdBufInfo));

		/*
			Dispatch the ray tracing commands
		*/
		vkCmdBindPipeline(cmdBuffer, VK_PIPELINE_BIND_POINT_RAY_TRACING_KHR, pipeline);
		vkCmdBindDescriptorSets(cmdBuffer, VK_PIPELINE_BIND_POINT_RAY_TRACING_KHR, pipelineLayout, 0, 1, &descriptorSets[currentBuffer], 0, 0);

		VkStridedDeviceAddressRegionKHR emptySbtEntry = {};
		vkCmdTraceRaysKHR(
			cmdBuffer,
			&shaderBindingTables.raygen.stridedDeviceAddressRegion,
			&shaderBindingTables.miss.stridedDeviceAddressRegion,
			&shaderBindingTables.hit.stridedDeviceAddressRegion,
			&emptySbtEntry,
			width,
			height,
			1);

		/*
			Copy ray tracing output to swap chain image
		*/

		// Prepare current swap chain image as transfer destination
		vks::tools::setImageLayout(
			cmdBuffer,
			swapChain.images[currentImageIndex],
			VK_IMAGE_LAYOUT_UNDEFINED,
			VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,
			subresourceRange);

		// Prepare ray tracing output image as transfer source
		vks::tools::setImageLayout(
			cmdBuffer,
			storageImage.image,
			VK_IMAGE_LAYOUT_GENERAL,
			VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL,
			subresourceRange);

		VkImageCopy copyRegion{};
		copyRegion.srcSubresource = { VK_IMAGE_ASPECT_COLOR_BIT, 0, 0, 1 };
		copyRegion.srcOffset = { 0, 0, 0 };
		copyRegion.dstSubresource = { VK_IMAGE_ASPECT_COLOR_BIT, 0, 0, 1 };
		copyRegion.dstOffset = { 0, 0, 0 };
		copyRegion.extent = { width, height, 1 };
		vkCmdCopyImage(cmdBuffer, storageImage.image, VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, swapChain.images[currentImageIndex], VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, 1, &copyRegion);

		// Transition swap chain image back for presentation
		vks::tools::setImageLayout(
			cmdBuffer,
			swapChain.images[currentImageIndex],
			VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,
			VK_IMAGE_LAYOUT_PRESENT_SRC_KHR,
			subresourceRange);

		// Transition ray tracing output image back to general layout
		vks::tools::setImageLayout(
			cmdBuffer,
			storageImage.image,
			VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL,
			VK_IMAGE_LAYOUT_GENERAL,
			subresourceRange);

		drawUI(cmdBuffer, frameBuffers[currentImageIndex]);

		VK_CHECK_RESULT(vkEndCommandBuffer(cmdBuffer));
	}


	virtual void render()
	{
		if (!prepared)
			return;
		VulkanExampleBase::prepareFrame();
		if (camera.updated) {
			// If the camera's view has been updated we need to  reset the frame accumulation (which is used for transparent surfaces and anti-aliasing)
			uniformData.frame = -1;
		}
		updateUniformBuffers();
		buildCommandBuffer();
		VulkanExampleBase::submitFrame();
	}
};

VULKAN_EXAMPLE_MAIN()