/*
* Vulkan Example - Using inline uniform blocks for passing data to shader stages

* Note: Requires a device that supports the VK_EXT_inline_uniform_block extension
*
* Relevant code parts are marked with [POI]
*
* Copyright (C) 2018 by Sascha Willems - www.saschawillems.de
*
* This code is licensed under the MIT license (MIT) (http://opensource.org/licenses/MIT)
*/

#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <assert.h>
#include <vector>

#define GLM_FORCE_RADIANS
#define GLM_FORCE_DEPTH_ZERO_TO_ONE
#include <glm/glm.hpp>
#include <glm/gtc/matrix_transform.hpp>
#include <glm/gtc/type_ptr.hpp>

#include <vulkan/vulkan.h>
#include "vulkanexamplebase.h"
#include "VulkanTexture.hpp"
#include "VulkanModel.hpp"

#define ENABLE_VALIDATION false

class VulkanExample : public VulkanExampleBase
{
public:
	bool animate = true;

	vks::VertexLayout vertexLayout = vks::VertexLayout({
		vks::VERTEX_COMPONENT_POSITION,
		vks::VERTEX_COMPONENT_NORMAL,
		vks::VERTEX_COMPONENT_UV,
		vks::VERTEX_COMPONENT_COLOR,
	});

	/*
		[POI] This is the data structure that'll be passed using inline uniform blocks
	*/
	struct InlineBlockData {
		glm::vec4 color;
	};

	struct Cube {
		struct Matrices {
			glm::mat4 projection;
			glm::mat4 view;
			glm::mat4 model;
		} matrices;
		InlineBlockData inlineBlockData;
		VkDescriptorSet descriptorSet;
		vks::Texture2D texture;
		vks::Buffer uniformBuffer;
		glm::vec3 rotation;
	};
	std::array<Cube, 2> cubes;

	struct Models {
		vks::Model cube;
	} models;
	
	VkPipeline pipeline;
	VkPipelineLayout pipelineLayout;

	VkDescriptorSetLayout descriptorSetLayout;

	VulkanExample() : VulkanExampleBase(ENABLE_VALIDATION)
	{
		title = "Inline uniform blocks";
		settings.overlay = true;
		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, -5.0f));
		/*
			[POI] Enable extension required for conditional rendering
		*/
		enabledDeviceExtensions.push_back(VK_EXT_INLINE_UNIFORM_BLOCK_EXTENSION_NAME);
	}

	~VulkanExample()
	{
		vkDestroyPipeline(device, pipeline, nullptr);
		vkDestroyPipelineLayout(device, pipelineLayout, nullptr);
		vkDestroyDescriptorSetLayout(device, descriptorSetLayout, nullptr);
		models.cube.destroy();
		for (auto cube : cubes) {
			cube.uniformBuffer.destroy();
			cube.texture.destroy();
		}
	}

	void buildCommandBuffers()
	{
		VkCommandBufferBeginInfo cmdBufInfo = vks::initializers::commandBufferBeginInfo();

		VkClearValue clearValues[2];
		clearValues[0].color = defaultClearColor;
		clearValues[1].depthStencil = { 1.0f, 0 };

		VkRenderPassBeginInfo renderPassBeginInfo = vks::initializers::renderPassBeginInfo();
		renderPassBeginInfo.renderPass = renderPass;
		renderPassBeginInfo.renderArea.offset.x = 0;
		renderPassBeginInfo.renderArea.offset.y = 0;
		renderPassBeginInfo.renderArea.extent.width = width;
		renderPassBeginInfo.renderArea.extent.height = height;
		renderPassBeginInfo.clearValueCount = 2;
		renderPassBeginInfo.pClearValues = clearValues;

		for (int32_t i = 0; i < drawCmdBuffers.size(); ++i) {
			renderPassBeginInfo.framebuffer = frameBuffers[i];

			VK_CHECK_RESULT(vkBeginCommandBuffer(drawCmdBuffers[i], &cmdBufInfo));

			vkCmdBeginRenderPass(drawCmdBuffers[i], &renderPassBeginInfo, VK_SUBPASS_CONTENTS_INLINE);

			vkCmdBindPipeline(drawCmdBuffers[i], VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline);

			VkViewport viewport = vks::initializers::viewport((float)width, (float)height, 0.0f, 1.0f);
			vkCmdSetViewport(drawCmdBuffers[i], 0, 1, &viewport);

			VkRect2D scissor = vks::initializers::rect2D(width, height, 0, 0);
			vkCmdSetScissor(drawCmdBuffers[i], 0, 1, &scissor);

			VkDeviceSize offsets[1] = { 0 };
			vkCmdBindVertexBuffers(drawCmdBuffers[i], 0, 1, &models.cube.vertices.buffer, offsets);
			vkCmdBindIndexBuffer(drawCmdBuffers[i], models.cube.indices.buffer, 0, VK_INDEX_TYPE_UINT32);

			for (auto cube : cubes) {
				vkCmdBindDescriptorSets(drawCmdBuffers[i], VK_PIPELINE_BIND_POINT_GRAPHICS, pipelineLayout, 0, 1, &cube.descriptorSet, 0, nullptr);
				vkCmdDrawIndexed(drawCmdBuffers[i], models.cube.indexCount, 1, 0, 0, 0);
			}

			drawUI(drawCmdBuffers[i]);

			vkCmdEndRenderPass(drawCmdBuffers[i]);

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

	void loadAssets()
	{
		models.cube.loadFromFile(getAssetPath() + "models/cube.dae", vertexLayout, 1.0f, vulkanDevice, queue);
		cubes[0].texture.loadFromFile(getAssetPath() + "textures/crate01_color_height_rgba.ktx", VK_FORMAT_R8G8B8A8_UNORM, vulkanDevice, queue);
		cubes[1].texture.loadFromFile(getAssetPath() + "textures/crate02_color_height_rgba.ktx", VK_FORMAT_R8G8B8A8_UNORM, vulkanDevice, queue);
		cubes[0].inlineBlockData.color = glm::vec4(1.0f, 0.0f, 0.0f, 1.0f);
		cubes[1].inlineBlockData.color = glm::vec4(0.0f, 0.0f, 1.0f, 1.0f);
	}

	/*
		[POI] Set up descriptor sets and set layout
	*/
	void setupDescriptors()
	{
		const uint32_t cubeCount = static_cast<uint32_t>(cubes.size());

		/*
			Descriptor pool
		*/

		std::array<VkDescriptorPoolSize, 3> descriptorPoolSizes{};

		// One uniform buffer descriptor per cube
		descriptorPoolSizes[0].type = VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER;
		descriptorPoolSizes[0].descriptorCount = cubeCount;

		/*
			[POI] One inline uniform block descriptor per cube
		*/
		descriptorPoolSizes[1].type = VK_DESCRIPTOR_TYPE_INLINE_UNIFORM_BLOCK_EXT;
		// Descriptor count for inline uniform blocks contains the combined data sizes of all inline uniform blocks used from this pool
		descriptorPoolSizes[1].descriptorCount = cubeCount * sizeof(InlineBlockData);

		// One combined image samples per cube
		descriptorPoolSizes[2].type = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER;
		descriptorPoolSizes[2].descriptorCount = static_cast<uint32_t>(cubes.size());

		// Create the global descriptor pool

		VkDescriptorPoolCreateInfo descriptorPoolCI = {};
		descriptorPoolCI.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_POOL_CREATE_INFO;
		descriptorPoolCI.poolSizeCount = static_cast<uint32_t>(descriptorPoolSizes.size());
		descriptorPoolCI.pPoolSizes = descriptorPoolSizes.data();
		descriptorPoolCI.maxSets = static_cast<uint32_t>(descriptorPoolSizes.size());
#
		/*
			[POI] New structure that has to be chained into the descriptor pool create info if you want to allocate inline uniform blocks
		*/
		VkDescriptorPoolInlineUniformBlockCreateInfoEXT descriptorPoolInlineUniformBlockCreateInfo{};
		descriptorPoolInlineUniformBlockCreateInfo.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_POOL_INLINE_UNIFORM_BLOCK_CREATE_INFO_EXT;
		descriptorPoolInlineUniformBlockCreateInfo.maxInlineUniformBlockBindings = 1;
		// Chain into descriptor pool create info
		descriptorPoolCI.pNext = &descriptorPoolInlineUniformBlockCreateInfo;
		
		VK_CHECK_RESULT(vkCreateDescriptorPool(device, &descriptorPoolCI, nullptr, &descriptorPool));

		/*
			Descriptor set layout		
		*/

		std::array<VkDescriptorSetLayoutBinding,3> setLayoutBindings{};

		// Binding 0: Uniform buffers (used to pass matrices)
		setLayoutBindings[0].descriptorType = VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER;
		setLayoutBindings[0].binding = 0;
		setLayoutBindings[0].stageFlags = VK_SHADER_STAGE_VERTEX_BIT;
		setLayoutBindings[0].descriptorCount = 1;

		/*
			[POI] Binding 1: Inline uniform block
		*/
		setLayoutBindings[1].descriptorType = VK_DESCRIPTOR_TYPE_INLINE_UNIFORM_BLOCK_EXT;
		setLayoutBindings[1].binding = 1;
		setLayoutBindings[1].stageFlags = VK_SHADER_STAGE_VERTEX_BIT;
		// Descriptor count for an inline uniform block contains data sizes of the block
		setLayoutBindings[1].descriptorCount = sizeof(InlineBlockData);

		// Binding 2: Combined image sampler (used to pass per object texture information) 
		setLayoutBindings[2].descriptorType = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER;
		setLayoutBindings[2].binding = 2;
		setLayoutBindings[2].stageFlags = VK_SHADER_STAGE_FRAGMENT_BIT;
		setLayoutBindings[2].descriptorCount = 1;

		// Create the descriptor set layout
		VkDescriptorSetLayoutCreateInfo descriptorLayoutCI{};
		descriptorLayoutCI.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_LAYOUT_CREATE_INFO;
		descriptorLayoutCI.bindingCount = static_cast<uint32_t>(setLayoutBindings.size());
		descriptorLayoutCI.pBindings = setLayoutBindings.data();
		
		VK_CHECK_RESULT(vkCreateDescriptorSetLayout(device, &descriptorLayoutCI, nullptr, &descriptorSetLayout));

		/*
			Descriptor sets
		*/

		for (auto &cube: cubes) {

			// Allocates an empty descriptor set without actual descriptors from the pool using the set layout
			VkDescriptorSetAllocateInfo allocateInfo{};
			allocateInfo.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_ALLOCATE_INFO;
			allocateInfo.descriptorPool = descriptorPool;
			allocateInfo.descriptorSetCount = 1;
			allocateInfo.pSetLayouts = &descriptorSetLayout;
			VK_CHECK_RESULT(vkAllocateDescriptorSets(device, &allocateInfo, &cube.descriptorSet));

			// Update the descriptor set with the actual descriptors matching shader bindings set in the layout

			std::array<VkWriteDescriptorSet, 3> writeDescriptorSets{};

			// Binding 0: Object matrices Uniform buffer
			writeDescriptorSets[0].sType = VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET;
			writeDescriptorSets[0].dstSet = cube.descriptorSet;
			writeDescriptorSets[0].dstBinding = 0;
			writeDescriptorSets[0].descriptorType = VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER;
			writeDescriptorSets[0].pBufferInfo = &cube.uniformBuffer.descriptor;
			writeDescriptorSets[0].descriptorCount = 1;
			
			/*
				[POI] Binding 1: Inline uniform block
			*/
			writeDescriptorSets[1].sType = VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET;
			writeDescriptorSets[1].dstSet = cube.descriptorSet;
			writeDescriptorSets[1].dstBinding = 1;
			writeDescriptorSets[1].descriptorType = VK_DESCRIPTOR_TYPE_INLINE_UNIFORM_BLOCK_EXT;
			// The dstArrayElement member can be used to define an offset for inline uniform blocks
			writeDescriptorSets[1].dstArrayElement = 0;
			// TODO: API-Design from hell
			writeDescriptorSets[1].descriptorCount = sizeof(glm::vec4);

			/*
				[POI] New structure that defines size and data  of the inline uniform block 
			*/
			VkWriteDescriptorSetInlineUniformBlockEXT writeDescriptorSetInlineUniformBlock{};
			writeDescriptorSetInlineUniformBlock.sType = VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET_INLINE_UNIFORM_BLOCK_EXT;
			writeDescriptorSetInlineUniformBlock.dataSize = sizeof(InlineBlockData);
			writeDescriptorSetInlineUniformBlock.pData = &cube.inlineBlockData;
			// Needs to be chained to an existing write descriptor set structure
			writeDescriptorSets[1].pNext = &writeDescriptorSetInlineUniformBlock;

			// Binding 2: Object texture
			writeDescriptorSets[2].sType = VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET;
			writeDescriptorSets[2].dstSet = cube.descriptorSet;
			writeDescriptorSets[2].dstBinding = 2;
			writeDescriptorSets[2].descriptorType = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER;
			writeDescriptorSets[2].pImageInfo = &cube.texture.descriptor;
			writeDescriptorSets[2].descriptorCount = 1;

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

	}

	void preparePipelines()
	{
		VkPipelineLayoutCreateInfo pipelineLayoutCI{};
		pipelineLayoutCI.sType = VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO;
		pipelineLayoutCI.setLayoutCount = 1;
		pipelineLayoutCI.pSetLayouts = &descriptorSetLayout;
		VK_CHECK_RESULT(vkCreatePipelineLayout(device, &pipelineLayoutCI, nullptr, &pipelineLayout));

		const std::vector<VkDynamicState> dynamicStateEnables = { VK_DYNAMIC_STATE_VIEWPORT, VK_DYNAMIC_STATE_SCISSOR };

		VkPipelineInputAssemblyStateCreateInfo inputAssemblyStateCI = vks::initializers::pipelineInputAssemblyStateCreateInfo(VK_PRIMITIVE_TOPOLOGY_TRIANGLE_LIST, 0, VK_FALSE);
		VkPipelineRasterizationStateCreateInfo rasterizationStateCI = vks::initializers::pipelineRasterizationStateCreateInfo(VK_POLYGON_MODE_FILL, VK_CULL_MODE_BACK_BIT, VK_FRONT_FACE_CLOCKWISE, 0);
		VkPipelineColorBlendAttachmentState blendAttachmentState = vks::initializers::pipelineColorBlendAttachmentState(0xf, VK_FALSE);
		VkPipelineColorBlendStateCreateInfo colorBlendStateCI = vks::initializers::pipelineColorBlendStateCreateInfo(1, &blendAttachmentState);
		VkPipelineDepthStencilStateCreateInfo depthStencilStateCI = vks::initializers::pipelineDepthStencilStateCreateInfo(VK_TRUE, VK_TRUE, VK_COMPARE_OP_LESS_OR_EQUAL);
		VkPipelineViewportStateCreateInfo viewportStateCI = vks::initializers::pipelineViewportStateCreateInfo(1, 1, 0);
		VkPipelineMultisampleStateCreateInfo multisampleStateCI = vks::initializers::pipelineMultisampleStateCreateInfo(VK_SAMPLE_COUNT_1_BIT, 0);
		VkPipelineDynamicStateCreateInfo dynamicStateCI = vks::initializers::pipelineDynamicStateCreateInfo(dynamicStateEnables.data(), static_cast<uint32_t>(dynamicStateEnables.size()),0);

		// Vertex bindings and attributes
		const std::vector<VkVertexInputBindingDescription> vertexInputBindings = {
			vks::initializers::vertexInputBindingDescription(0, vertexLayout.stride(), VK_VERTEX_INPUT_RATE_VERTEX),
		};
		const std::vector<VkVertexInputAttributeDescription> vertexInputAttributes = {
			vks::initializers::vertexInputAttributeDescription(0, 0, VK_FORMAT_R32G32B32_SFLOAT, 0),					// Location 0: Position			
			vks::initializers::vertexInputAttributeDescription(0, 1, VK_FORMAT_R32G32B32_SFLOAT, sizeof(float) * 3),	// Location 1: Normal
			vks::initializers::vertexInputAttributeDescription(0, 2, VK_FORMAT_R32G32_SFLOAT, sizeof(float) * 6),		// Location 2: UV
			vks::initializers::vertexInputAttributeDescription(0, 3, VK_FORMAT_R32G32B32_SFLOAT, sizeof(float) * 8),	// Location 3: Color
		};
		VkPipelineVertexInputStateCreateInfo vertexInputState = vks::initializers::pipelineVertexInputStateCreateInfo();
		vertexInputState.vertexBindingDescriptionCount = static_cast<uint32_t>(vertexInputBindings.size());
		vertexInputState.pVertexBindingDescriptions = vertexInputBindings.data();
		vertexInputState.vertexAttributeDescriptionCount = static_cast<uint32_t>(vertexInputAttributes.size());
		vertexInputState.pVertexAttributeDescriptions = vertexInputAttributes.data();

		VkGraphicsPipelineCreateInfo pipelineCreateInfoCI = vks::initializers::pipelineCreateInfo(pipelineLayout, renderPass, 0);
		pipelineCreateInfoCI.pVertexInputState = &vertexInputState;
		pipelineCreateInfoCI.pInputAssemblyState = &inputAssemblyStateCI;
		pipelineCreateInfoCI.pRasterizationState = &rasterizationStateCI;
		pipelineCreateInfoCI.pColorBlendState = &colorBlendStateCI;
		pipelineCreateInfoCI.pMultisampleState = &multisampleStateCI;
		pipelineCreateInfoCI.pViewportState = &viewportStateCI;
		pipelineCreateInfoCI.pDepthStencilState = &depthStencilStateCI;
		pipelineCreateInfoCI.pDynamicState = &dynamicStateCI;

		const std::array<VkPipelineShaderStageCreateInfo, 2> shaderStages = {
			loadShader(getAssetPath() + "shaders/inlineuniformblocks/cube.vert.spv", VK_SHADER_STAGE_VERTEX_BIT),
			loadShader(getAssetPath() + "shaders/inlineuniformblocks/cube.frag.spv", VK_SHADER_STAGE_FRAGMENT_BIT)
		};

		pipelineCreateInfoCI.stageCount = static_cast<uint32_t>(shaderStages.size());
		pipelineCreateInfoCI.pStages = shaderStages.data();

		VK_CHECK_RESULT(vkCreateGraphicsPipelines(device, pipelineCache, 1, &pipelineCreateInfoCI, nullptr, &pipeline));
	}

	void prepareUniformBuffers()
	{
		// Vertex shader matrix uniform buffer block
		for (auto& cube : cubes) {
			VK_CHECK_RESULT(vulkanDevice->createBuffer(
				VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT,
				VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT | VK_MEMORY_PROPERTY_HOST_COHERENT_BIT,
				&cube.uniformBuffer,
				sizeof(Cube::Matrices)));
			VK_CHECK_RESULT(cube.uniformBuffer.map());
		}

		updateUniformBuffers();
	}

	void updateUniformBuffers()
	{
		cubes[0].matrices.model = glm::translate(glm::mat4(1.0f), glm::vec3(-2.0f, 0.0f, 0.0f));
		cubes[1].matrices.model = glm::translate(glm::mat4(1.0f), glm::vec3( 1.5f, 0.5f, 0.0f));

		for (auto& cube : cubes) {
			cube.matrices.projection = camera.matrices.perspective;
			cube.matrices.view = camera.matrices.view;
			cube.matrices.model = glm::rotate(cube.matrices.model, glm::radians(cube.rotation.x), glm::vec3(1.0f, 0.0f, 0.0f));
			cube.matrices.model = glm::rotate(cube.matrices.model, glm::radians(cube.rotation.y), glm::vec3(0.0f, 1.0f, 0.0f));
			cube.matrices.model = glm::rotate(cube.matrices.model, glm::radians(cube.rotation.z), glm::vec3(0.0f, 0.0f, 1.0f));
			memcpy(cube.uniformBuffer.mapped, &cube.matrices, sizeof(cube.matrices));
		}
	}

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

	void prepare()
	{
		VulkanExampleBase::prepare();
		loadAssets();
		prepareUniformBuffers();
		setupDescriptors();
		preparePipelines();
		buildCommandBuffers();
		prepared = true;
	}

	virtual void render()
	{
		if (!prepared)
			return;
		draw();
		if (animate) {
			cubes[0].rotation.x += 2.5f * frameTimer;
			if (cubes[0].rotation.x > 360.0f)
				cubes[0].rotation.x -= 360.0f;
			cubes[1].rotation.y += 2.0f * frameTimer;
			if (cubes[1].rotation.x > 360.0f)
				cubes[1].rotation.x -= 360.0f;
		}
		if ((camera.updated) || (animate)) {
			updateUniformBuffers();
		}
	}

	virtual void OnUpdateUIOverlay(vks::UIOverlay *overlay)
	{
		if (overlay->header("Settings")) {
			overlay->checkBox("Animate", &animate);
		}
	}
};

VULKAN_EXAMPLE_MAIN()