当前位置：首页 > news >正文

Vulkan笔记（十）-图形管道的七个配置项

news 2025/8/20 10:59:06

在Vulkan图形管道中需要设置七道工序：

1.输入汇编器并且进行装配

 VkPipelineVertexInputStateCreateInfo vertexInputInfo{};vertexInputInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO;vertexInputInfo.vertexBindingDescriptionCount = 1;vertexInputInfo.pVertexBindingDescriptions = &bindingDescription;;vertexInputInfo.vertexAttributeDescriptionCount = static_cast<uint32_t>(attributeDescriptions.size());vertexInputInfo.pVertexAttributeDescriptions = attributeDescriptions.data();//输入装配VkPipelineInputAssemblyStateCreateInfo assemblyInfo{};assemblyInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_INPUT_ASSEMBLY_STATE_CREATE_INFO;assemblyInfo.topology = VK_PRIMITIVE_TOPOLOGY_TRIANGLE_STRIP;assemblyInfo.primitiveRestartEnable = VK_FALSE;

这里对顶点属性比如：位置、颜色、法线、纹理等描述信息进行创建，装配按照什么方式是点、线还是三角形等方式把顶点数据组织起来。

2.视口与裁剪矩形

VkPipelineViewportStateCreateInfo viewPortInfo{};
viewPortInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_VIEWPORT_STATE_CREATE_INFO;
viewPortInfo.pNext = nullptr;
viewPortInfo.flags = 0;
viewPortInfo.viewportCount = 1;
viewPortInfo.pViewports = nullptr;
viewPortInfo.scissorCount = 1;
viewPortInfo.pScissors = nullptr;

3.光栅化离散器

 VkPipelineRasterizationStateCreateInfo rasterizationInfo{};rasterizationInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_RASTERIZATION_STATE_CREATE_INFO;rasterizationInfo.depthClampEnable = VK_FALSE;rasterizationInfo.rasterizerDiscardEnable = VK_FALSE;rasterizationInfo.polygonMode = VK_POLYGON_MODE_FILL;rasterizationInfo.lineWidth = 1.f;rasterizationInfo.cullMode = VK_CULL_MODE_NONE;rasterizationInfo.frontFace = VK_FRONT_FACE_CLOCKWISE;rasterizationInfo.depthBiasEnable = VK_FALSE;rasterizationInfo.depthBiasConstantFactor = 0.f;rasterizationInfo.depthBiasSlopeFactor = 0.f;rasterizationInfo.depthBiasClamp = 0.f;

将几何图元（如三角形）离散化为屏幕上的像素（称为片元/Fragment），确定哪些像素被图元覆盖，并进行属性插值计算（如颜色、深度）

4.多重采样

VkPipelineMultisampleStateCreateInfo multisampleInfo{};
multisampleInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_MULTISAMPLE_STATE_CREATE_INFO;
multisampleInfo.rasterizationSamples = VK_SAMPLE_COUNT_1_BIT;
multisampleInfo.sampleShadingEnable = VK_FALSE;
multisampleInfo.minSampleShading = 1.0f;
multisampleInfo.pSampleMask = nullptr;
multisampleInfo.alphaToCoverageEnable = VK_FALSE;
multisampleInfo.alphaToOneEnable = VK_FALSE;

对每个像素进行多次采样（而不仅仅是中心点），然后混合这些采样结果，使边缘看起来更平滑。

5.深度与模板测试

VkPipelineDepthStencilStateCreateInfo depthStencilInfo{};
depthStencilInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_DEPTH_STENCIL_STATE_CREATE_INFO;
depthStencilInfo.depthTestEnable = VK_TRUE;
depthStencilInfo.depthWriteEnable = VK_TRUE;
depthStencilInfo.depthCompareOp = VK_COMPARE_OP_LESS;
depthStencilInfo.depthBoundsTestEnable = VK_FALSE;
depthStencilInfo.minDepthBounds = 0.0f;
depthStencilInfo.maxDepthBounds = 1.0f;
depthStencilInfo.stencilTestEnable = VK_FALSE;
depthStencilInfo.front = {};
depthStencilInfo.back = {};

深度测试和模板测试是两种关键的像素筛选机制，用于决定哪些像素最终被绘制到屏幕上。它们通常在片元处理阶段（Fragment Processing）之后执行，确保正确的遮挡关系和特殊效果（如阴影、轮廓、镜面反射等）。深度测试：解决“谁在前谁在后”的问题，是 3D 渲染的基础。模板测试：实现复杂特效（如镜子、X 光效果），灵活性高。结合使用：可高效实现 阴影体积（Shadow Volumes）、动态门户等高级渲染技术。

6.颜色混合

VkPipelineColorBlendAttachmentState colorBlendAttachmentInfo{};colorBlendAttachmentInfo.colorWriteMask = VK_COLOR_COMPONENT_R_BIT | VK_COLOR_COMPONENT_G_BIT | VK_COLOR_COMPONENT_B_BIT | VK_COLOR_COMPONENT_A_BIT;colorBlendAttachmentInfo.blendEnable = VK_TRUE;colorBlendAttachmentInfo.colorBlendOp = VK_BLEND_OP_ADD;colorBlendAttachmentInfo.alphaBlendOp = VK_BLEND_OP_ADD;colorBlendAttachmentInfo.srcColorBlendFactor = VK_BLEND_FACTOR_SRC_ALPHA;colorBlendAttachmentInfo.dstColorBlendFactor = VK_BLEND_FACTOR_ONE_MINUS_SRC_ALPHA;colorBlendAttachmentInfo.srcAlphaBlendFactor = VK_BLEND_FACTOR_ONE;colorBlendAttachmentInfo.dstAlphaBlendFactor = VK_BLEND_FACTOR_ZERO;VkPipelineColorBlendStateCreateInfo colorCreateInfo{};colorCreateInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_COLOR_BLEND_STATE_CREATE_INFO;colorCreateInfo.logicOpEnable = VK_FALSE;colorCreateInfo.logicOp = VK_LOGIC_OP_COPY;colorCreateInfo.attachmentCount = 1;colorCreateInfo.pAttachments = &colorBlendAttachmentInfo;colorCreateInfo.blendConstants[0] = 0.f;colorCreateInfo.blendConstants[1] = 0.f;colorCreateInfo.blendConstants[2] = 0.f;colorCreateInfo.blendConstants[3] = 0.f;

用于处理 半透明物体 或 多渲染目标（MRT） 的颜色叠加。它决定了当前绘制的像素（源颜色，Source Color）如何与帧缓冲区（目标颜色，Destination Color）中已有的像素进行混合。

7.管道布局

VkPipelineLayoutCreateInfo pipelineLayoutInfo{};pipelineLayoutInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO;pipelineLayoutInfo.setLayoutCount = 1;pipelineLayoutInfo.pSetLayouts = &uboSetLayout_;pipelineLayoutInfo.pushConstantRangeCount = 0;pipelineLayoutInfo.pPushConstantRanges = nullptr;if (vkCreatePipelineLayout(logicDevice_, &pipelineLayoutInfo, nullptr, &pipelineLayout_) != VK_SUCCESS){throw std::runtime_error("Failed to create pipeline layout!");}

着色器如何访问外部资源（如 Uniform Buffer、纹理、采样器等）的核心对象。它相当于一个“资源访问协议”，告诉 GPU 着色器可以读取哪些数据以及这些数据的组织方式。