/*
 * Vulkan Example - Hardware accelerated ray tracing intersection shader samples
 * 
 * Copyright (C) 2023 by Sascha Willems - www.saschawillems.de
 *
 * This sample uses intersection shaders for doing prodcedural ray traced geometry
 * Instead of passing actual geometry, this samples only passes bounding boxes and sphere descriptions
 * The bounding boxes are used for the ray traversal and the sphere intersections are done
 * within the intersection shader
 *
 * This code is licensed under the MIT license (MIT) (http://opensource.org/licenses/MIT)
 */

#include "VulkanRaytracingSample.h"

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

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

	struct UniformData {
		glm::mat4 viewInverse;
		glm::mat4 projInverse;
		glm::vec4 lightPos;
	} uniformData;
	vks::Buffer ubo;

	VkPipeline pipeline;
	VkPipelineLayout pipelineLayout;
	VkDescriptorSet descriptorSet;
	VkDescriptorSetLayout descriptorSetLayout;

	struct Sphere {
		glm::vec3 center;
		float radius;
		glm::vec4 color;
	};
	
	struct AABB {
		glm::vec3 min;
		glm::vec3 max;
	};

	vks::Buffer spheresBuffer;
	vks::Buffer aabbsBuffer;
	uint32_t aabbCount{ 0 };

	// This sample is derived from an extended base class that saves most of the ray tracing setup boiler plate
	VulkanExample() : VulkanRaytracingSample()
	{
		title = "Ray tracing intersection shaders";
		timerSpeed *= 0.25f;
		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, 0.0f, -60.0f));
		enableExtensions();
	}

	~VulkanExample()
	{
		vkDestroyPipeline(device, pipeline, nullptr);
		vkDestroyPipelineLayout(device, pipelineLayout, nullptr);
		vkDestroyDescriptorSetLayout(device, descriptorSetLayout, nullptr);
		deleteStorageImage();
		deleteAccelerationStructure(bottomLevelAS);
		deleteAccelerationStructure(topLevelAS);
		shaderBindingTables.raygen.destroy();
		shaderBindingTables.miss.destroy();
		shaderBindingTables.hit.destroy();
		ubo.destroy();
		spheresBuffer.destroy();
		aabbsBuffer.destroy();
	}

	void createBuffers()
	{
		// We'll be using two buffers to describe the procedural geometry

		// A buffer with randpmly generatd sphere descriptions (center, radius, material) that'll be passed to the ray tracing shaders as a shader storage buffer object
		std::vector<Sphere> spheres{};
		std::default_random_engine rndGenerator(benchmark.active ? 0 : (unsigned)time(nullptr));
		std::uniform_real_distribution<float> uniformDist(0.0, 1.0);
		std::uniform_real_distribution<float> sizeDist(1.0, 2.0);
		for (uint32_t i = 0; i < 1024; i++) {
			Sphere sphere{};
			//sphere.center = 
			sphere.radius = sizeDist(rndGenerator);
			sphere.color = glm::vec4(uniformDist(rndGenerator), uniformDist(rndGenerator), uniformDist(rndGenerator), 1.0f);
			// Get a random point in a sphere
			float x,y,z,d{ 0.0f };
			do {
				x = uniformDist(rndGenerator) * 2.0f - 1.0f;
				y = uniformDist(rndGenerator) * 2.0f - 1.0f;
				z = uniformDist(rndGenerator) * 2.0f - 1.0f;
				d = x * x + y * y + z * z;

			} while (d > 1.0);
			sphere.center = glm::vec3(x, y, z) * 25.0f;
			spheres.push_back(sphere);
		}

		// A buffer with the (axis aligned) bounding boxes of our sphere, which is used during the ray tracing traversal for hit detection
		std::vector<AABB> aabbs{};
		for (auto& sphere : spheres) {
			aabbs.push_back({ sphere.center - glm::vec3(sphere.radius), sphere.center + glm::vec3(sphere.radius) });
		}
		aabbCount = static_cast<uint32_t>(aabbs.size());

		// Copy the buffer to the device for performance reasons
		vks::Buffer stagingBuffer{};
		VkBufferUsageFlags usageFlags = VK_BUFFER_USAGE_SHADER_DEVICE_ADDRESS_BIT | VK_BUFFER_USAGE_ACCELERATION_STRUCTURE_BUILD_INPUT_READ_ONLY_BIT_KHR | VK_BUFFER_USAGE_STORAGE_BUFFER_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT;

		// Spheres
		VK_CHECK_RESULT(vulkanDevice->createBuffer(VK_BUFFER_USAGE_TRANSFER_SRC_BIT, VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT | VK_MEMORY_PROPERTY_HOST_COHERENT_BIT, &stagingBuffer, sizeof(Sphere)* spheres.size(), spheres.data()));
		VK_CHECK_RESULT(vulkanDevice->createBuffer(usageFlags, VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT, &spheresBuffer, sizeof(Sphere)* spheres.size()));
		vulkanDevice->copyBuffer(&stagingBuffer, &spheresBuffer, queue);
		stagingBuffer.destroy();

		// AABBs
		VK_CHECK_RESULT(vulkanDevice->createBuffer(VK_BUFFER_USAGE_TRANSFER_SRC_BIT, VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT | VK_MEMORY_PROPERTY_HOST_COHERENT_BIT, &stagingBuffer, sizeof(AABB)* aabbs.size(), aabbs.data()));
		VK_CHECK_RESULT(vulkanDevice->createBuffer(usageFlags, VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT, &aabbsBuffer, sizeof(AABB)* aabbs.size()));
		vulkanDevice->copyBuffer(&stagingBuffer, &aabbsBuffer, queue);
		stagingBuffer.destroy();
	}

	/*
		Create the bottom level acceleration structure only containing axis aligned bounding boxes for our procedural geometry
	*/
	void createBottomLevelAccelerationStructure()
	{
		// Build
		VkAccelerationStructureGeometryKHR accelerationStructureGeometry = vks::initializers::accelerationStructureGeometryKHR();
		accelerationStructureGeometry.flags = VK_GEOMETRY_OPAQUE_BIT_KHR;
		// Instead of providing actual geometry (e.g. triangles), we only provide the axis aligned bounding boxes (AABBs) of the spheres
		// The data for the actual spheres is passed elsewhere as a shader storage buffer object
		accelerationStructureGeometry.geometryType = VK_GEOMETRY_TYPE_AABBS_KHR;
		accelerationStructureGeometry.geometry.aabbs.sType = VK_STRUCTURE_TYPE_ACCELERATION_STRUCTURE_GEOMETRY_AABBS_DATA_KHR;
		accelerationStructureGeometry.geometry.aabbs.data.deviceAddress = getBufferDeviceAddress(aabbsBuffer.buffer);
		accelerationStructureGeometry.geometry.aabbs.stride = sizeof(AABB);

		// Get size info
		VkAccelerationStructureBuildGeometryInfoKHR accelerationStructureBuildGeometryInfo = vks::initializers::accelerationStructureBuildGeometryInfoKHR();
		accelerationStructureBuildGeometryInfo.type = VK_ACCELERATION_STRUCTURE_TYPE_BOTTOM_LEVEL_KHR;
		accelerationStructureBuildGeometryInfo.flags = VK_BUILD_ACCELERATION_STRUCTURE_PREFER_FAST_TRACE_BIT_KHR;
		accelerationStructureBuildGeometryInfo.geometryCount = 1;
		accelerationStructureBuildGeometryInfo.pGeometries = &accelerationStructureGeometry;

		VkAccelerationStructureBuildSizesInfoKHR accelerationStructureBuildSizesInfo = vks::initializers::accelerationStructureBuildSizesInfoKHR();
		vkGetAccelerationStructureBuildSizesKHR(
			device,
			VK_ACCELERATION_STRUCTURE_BUILD_TYPE_DEVICE_KHR,
			&accelerationStructureBuildGeometryInfo,
			&aabbCount,
			&accelerationStructureBuildSizesInfo);

		createAccelerationStructure(bottomLevelAS, VK_ACCELERATION_STRUCTURE_TYPE_BOTTOM_LEVEL_KHR, accelerationStructureBuildSizesInfo);

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

		VkAccelerationStructureBuildGeometryInfoKHR accelerationBuildGeometryInfo = vks::initializers::accelerationStructureBuildGeometryInfoKHR();
		accelerationBuildGeometryInfo.type = VK_ACCELERATION_STRUCTURE_TYPE_BOTTOM_LEVEL_KHR;
		accelerationBuildGeometryInfo.flags = VK_BUILD_ACCELERATION_STRUCTURE_PREFER_FAST_TRACE_BIT_KHR;
		accelerationBuildGeometryInfo.mode = VK_BUILD_ACCELERATION_STRUCTURE_MODE_BUILD_KHR;
		accelerationBuildGeometryInfo.dstAccelerationStructure = bottomLevelAS.handle;
		accelerationBuildGeometryInfo.geometryCount = 1;
		accelerationBuildGeometryInfo.pGeometries = &accelerationStructureGeometry;
		accelerationBuildGeometryInfo.scratchData.deviceAddress = scratchBuffer.deviceAddress;

		VkAccelerationStructureBuildRangeInfoKHR accelerationStructureBuildRangeInfo{};
		accelerationStructureBuildRangeInfo.primitiveCount = aabbCount;
		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);

		deleteScratchBuffer(scratchBuffer);
	}

	/*
		The top level acceleration structure contains the scene's object instances
	*/
	void createTopLevelAccelerationStructure()
	{
		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 = vks::initializers::accelerationStructureGeometryKHR();
		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
		VkAccelerationStructureBuildGeometryInfoKHR accelerationStructureBuildGeometryInfo = vks::initializers::accelerationStructureBuildGeometryInfoKHR();
		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 = vks::initializers::accelerationStructureBuildSizesInfoKHR();
		vkGetAccelerationStructureBuildSizesKHR(
			device,
			VK_ACCELERATION_STRUCTURE_BUILD_TYPE_DEVICE_KHR,
			&accelerationStructureBuildGeometryInfo,
			&primitive_count,
			&accelerationStructureBuildSizesInfo);

		createAccelerationStructure(topLevelAS, VK_ACCELERATION_STRUCTURE_TYPE_TOP_LEVEL_KHR, accelerationStructureBuildSizesInfo);

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

		VkAccelerationStructureBuildGeometryInfoKHR accelerationBuildGeometryInfo = vks::initializers::accelerationStructureBuildGeometryInfoKHR();
		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);

		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      |
			|-----------|
			| hit + int |
			\-----------/

	*/
	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, 1);
		createShaderBindingTable(shaderBindingTables.hit, 1);

		// Copy handles
		memcpy(shaderBindingTables.raygen.mapped, shaderHandleStorage.data(), handleSize);
		memcpy(shaderBindingTables.miss.mapped, shaderHandleStorage.data() + handleSizeAligned, handleSize);
		memcpy(shaderBindingTables.hit.mapped, shaderHandleStorage.data() + handleSizeAligned * 2, handleSize);
	}

	/*
		Create the descriptor sets used for the ray tracing dispatch
	*/
	void createDescriptorSets()
	{
		std::vector<VkDescriptorPoolSize> poolSizes = {
			{ VK_DESCRIPTOR_TYPE_ACCELERATION_STRUCTURE_KHR, 1 },
			{ VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, 1 },
			{ VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER, 1 },
			{ VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, 2 }
		};
		VkDescriptorPoolCreateInfo descriptorPoolCreateInfo = vks::initializers::descriptorPoolCreateInfo(poolSizes, 1);
		VK_CHECK_RESULT(vkCreateDescriptorPool(device, &descriptorPoolCreateInfo, nullptr, &descriptorPool));

		VkDescriptorSetAllocateInfo descriptorSetAllocateInfo = vks::initializers::descriptorSetAllocateInfo(descriptorPool, &descriptorSetLayout, 1);
		VK_CHECK_RESULT(vkAllocateDescriptorSets(device, &descriptorSetAllocateInfo, &descriptorSet));

		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 = descriptorSet;
		accelerationStructureWrite.dstBinding = 0;
		accelerationStructureWrite.descriptorCount = 1;
		accelerationStructureWrite.descriptorType = VK_DESCRIPTOR_TYPE_ACCELERATION_STRUCTURE_KHR;

		// We pass the sphere descriptions as shader storage buffer, so the ray tracing shaders can source properties from it
		VkDescriptorImageInfo storageImageDescriptor{ VK_NULL_HANDLE, storageImage.view, VK_IMAGE_LAYOUT_GENERAL };
		VkDescriptorBufferInfo spheresBufferDescriptor{ spheresBuffer.buffer, 0, VK_WHOLE_SIZE };

		std::vector<VkWriteDescriptorSet> writeDescriptorSets = {
			// Binding 0: Top level acceleration structure
			accelerationStructureWrite,
			// Binding 1: Ray tracing result image
			vks::initializers::writeDescriptorSet(descriptorSet, VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, 1, &storageImageDescriptor),
			// Binding 2: Uniform data
			vks::initializers::writeDescriptorSet(descriptorSet, VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER, 2, &ubo.descriptor),
			// Binding 3: Spheres buffer
			vks::initializers::writeDescriptorSet(descriptorSet, VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, 3, &spheresBufferDescriptor),
		};
		vkUpdateDescriptorSets(device, static_cast<uint32_t>(writeDescriptorSets.size()), writeDescriptorSets.data(), 0, VK_NULL_HANDLE);
	}

	/*
		Create our ray tracing pipeline
	*/
	void createRayTracingPipeline()
	{
		std::vector<VkDescriptorSetLayoutBinding> setLayoutBindings = {
			// Binding 0: 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: Storage 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 | VK_SHADER_STAGE_INTERSECTION_BIT_KHR, 2),
			// Binding 3: Spheres buffer 
			vks::initializers::descriptorSetLayoutBinding(VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, VK_SHADER_STAGE_CLOSEST_HIT_BIT_KHR | VK_SHADER_STAGE_INTERSECTION_BIT_KHR, 3),
		};

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

		VkPipelineLayoutCreateInfo pPipelineLayoutCI = vks::initializers::pipelineLayoutCreateInfo(&descriptorSetLayout, 1);
		VK_CHECK_RESULT(vkCreatePipelineLayout(device, &pPipelineLayoutCI, nullptr, &pipelineLayout));
	
		/*
			Setup ray tracing shader groups
		*/
		std::vector<VkPipelineShaderStageCreateInfo> shaderStages;

		// Ray generation group
		{
			shaderStages.push_back(loadShader(getShadersPath() + "raytracingintersection/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() + "raytracingintersection/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);
		}

		// Closest hit group (procedural)
		{
			shaderStages.push_back(loadShader(getShadersPath() + "raytracingintersection/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_PROCEDURAL_HIT_GROUP_KHR;
			shaderGroup.generalShader = VK_SHADER_UNUSED_KHR;
			shaderGroup.closestHitShader = static_cast<uint32_t>(shaderStages.size()) - 1;
			shaderGroup.anyHitShader = VK_SHADER_UNUSED_KHR;
			// This group als uses an intersection shader for proedural geometry (see interseciton.rint for details)
			shaderStages.push_back(loadShader(getShadersPath() + "raytracingintersection/intersection.rint.spv", VK_SHADER_STAGE_INTERSECTION_BIT_KHR));
			shaderGroup.intersectionShader = static_cast<uint32_t>(shaderStages.size()) - 1;
			shaderGroups.push_back(shaderGroup);
		}

		VkRayTracingPipelineCreateInfoKHR rayTracingPipelineCI = vks::initializers::rayTracingPipelineCreateInfoKHR();
		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 = 2;
		rayTracingPipelineCI.layout = pipelineLayout;
		VK_CHECK_RESULT(vkCreateRayTracingPipelinesKHR(device, VK_NULL_HANDLE, VK_NULL_HANDLE, 1, &rayTracingPipelineCI, nullptr, &pipeline));
	}

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

		updateUniformBuffers();
	}

	/*
		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 descriptor
		VkDescriptorImageInfo storageImageDescriptor{ VK_NULL_HANDLE, storageImage.view, VK_IMAGE_LAYOUT_GENERAL };
		VkWriteDescriptorSet resultImageWrite = vks::initializers::writeDescriptorSet(descriptorSet, VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, 1, &storageImageDescriptor);
		vkUpdateDescriptorSets(device, 1, &resultImageWrite, 0, VK_NULL_HANDLE);
		resized = false;
	}

	/*
		Command buffer generation
	*/
	void buildCommandBuffers()
	{
		if (resized)
		{
			handleResize();
		}

		VkCommandBufferBeginInfo cmdBufInfo = vks::initializers::commandBufferBeginInfo();

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

		for (int32_t i = 0; i < drawCmdBuffers.size(); ++i)
		{
			VK_CHECK_RESULT(vkBeginCommandBuffer(drawCmdBuffers[i], &cmdBufInfo));

			/*
				Dispatch the ray tracing commands
			*/
			vkCmdBindPipeline(drawCmdBuffers[i], VK_PIPELINE_BIND_POINT_RAY_TRACING_KHR, pipeline);
			vkCmdBindDescriptorSets(drawCmdBuffers[i], VK_PIPELINE_BIND_POINT_RAY_TRACING_KHR, pipelineLayout, 0, 1, &descriptorSet, 0, 0);

			VkStridedDeviceAddressRegionKHR emptySbtEntry = {};
			vkCmdTraceRaysKHR(
				drawCmdBuffers[i],
				&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(
				drawCmdBuffers[i],
				swapChain.images[i],
				VK_IMAGE_LAYOUT_UNDEFINED,
				VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,
				subresourceRange);

			// Prepare ray tracing output image as transfer source
			vks::tools::setImageLayout(
				drawCmdBuffers[i],
				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(drawCmdBuffers[i], storageImage.image, VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, swapChain.images[i], VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, 1, &copyRegion);

			// Transition swap chain image back for presentation
			vks::tools::setImageLayout(
				drawCmdBuffers[i],
				swapChain.images[i],
				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(
				drawCmdBuffers[i],
				storageImage.image,
				VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL,
				VK_IMAGE_LAYOUT_GENERAL,
				subresourceRange);

			drawUI(drawCmdBuffers[i], frameBuffers[i]);

			VK_CHECK_RESULT(vkEndCommandBuffer(drawCmdBuffers[i]));
		}
	}

	void updateUniformBuffers()
	{
		uniformData.projInverse = glm::inverse(camera.matrices.perspective);
		uniformData.viewInverse = glm::inverse(camera.matrices.view);
		uniformData.lightPos = glm::vec4(cos(glm::radians(timer * 360.0f)) * 60.0f, 0.0f, 25.0f + sin(glm::radians(timer * 360.0f)) * 60.0f, 0.0f);
		memcpy(ubo.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;

		deviceCreatepNextChain = &enabledAccelerationStructureFeatures;
	}

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

		createBuffers();

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

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

	void draw()
	{
		VulkanExampleBase::prepareFrame();
		submitInfo.commandBufferCount = 1;
		submitInfo.pCommandBuffers = &drawCmdBuffers[currentBuffer];
		VK_CHECK_RESULT(vkQueueSubmit(queue, 1, &submitInfo, VK_NULL_HANDLE));
		VulkanExampleBase::submitFrame();
	}

	virtual void render()
	{
		if (!prepared)
			return;
		draw();
		if (!paused || camera.updated)
			updateUniformBuffers();
	}
};

VULKAN_EXAMPLE_MAIN()