vulakn教程--Drawing a Triangle--Pipeline--Render passes

发表于2016-12-05
评论0 3.3k浏览
原文链接 : vulkan-tutorial

Render Pass
  在创建Pipeline 之前我们必须告诉Vulkan在渲染时要使用的FrameBuffer 附件(attachments),需要定义使用color buffer 以及 depth buffer attachments的数量,要使用多少个采样(samples)以及应该如何处理采样的内容。所有这些信息都可以填写在Render Pass里。

Attachment description
  在我们的应用里只使用了一个color buffer attachment,我们用Swap Chain里的一个image 来表示这个buffer。先来看一下需要用到的结构:

1
2
3
4
5
6
7
8
9
10
11
typedef struct VkAttachmentDescription {
    VkAttachmentDescriptionFlags flags;
    VkFormat format;
    VkSampleCountFlagBits samples;
    VkAttachmentLoadOp loadOp;
    VkAttachmentStoreOp storeOp;
    VkAttachmentLoadOp stencilLoadOp;
    VkAttachmentStoreOp stencilStoreOp;
    VkImageLayout initialLayout;
    VkImageLayout finalLayout;
} VkAttachmentDescription;

1
2
3
4
5
6
// subpass 执行前对color或depth attachment内容做何种处理
typedef enum VkAttachmentLoadOp {
    VK_ATTACHMENT_LOAD_OP_LOAD = 0, //±£´æÒÑÓеÄÄÚÈÝ
    VK_ATTACHMENT_LOAD_OP_CLEAR = 1, //Çå¿Õ
    VK_ATTACHMENT_LOAD_OP_DONT_CARE = 2, // ²»ÔÚºõ
} VkAttachmentLoadOp;

1
2
3
4
5
// subpass 执行后做何种处理
typedef enum VkAttachmentStoreOp {
    VK_ATTACHMENT_STORE_OP_STORE = 0,  //保存
    VK_ATTACHMENT_STORE_OP_DONT_CARE = 1, //
} VkAttachmentStoreOp;

  现在我们来填充这个结构 :
1
2
3
VkAttachmentDescription colorAttachment = {};
colorAttachment.format = swapChainImageFormat;
colorAttachment.samples = VK_SAMPLE_COUNT_1_BIT;
  因为我们不使用multiSampling ,所以这里samples 取值 _1_BIT 。
1
2
colorAttachment.loadOp = VK_ATTACHMENT_LOAD_OP_CLEAR;
colorAttachment.storeOp = VK_ATTACHMENT_STORE_OP_STORE;
  loadOp和storeO表示渲染前和渲染后要做的动作,在我们的例子中,写入新片原(fragment)之前先清空FrameBuffer,使FrameBuffer变为黑色。我们想让渲染后的三角形显示到屏幕上,所以这里我们将storeOp 设置为保存。
1
2
colorAttachment.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
colorAttachment.stencilStoreOp = VK_ATTACHMENT_STORE_OP_DONT_CARE;
  stencil..Op应用于stencil data ,这里我们不需要关心。
1
2
colorAttachment.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED;
colorAttachment.finalLayout = VK_IMAGE_LAYOUT_PRESENT_SRC_KHR;
  在Vulkan中,用具有特定像素格式的VkImage 表示纹理(texture)和FrameBuffer,而像素在内存中的布局(layout)随着我们使用Image 的目的不同是可以改变的。一些常见的layout有:
  VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL: Images 用作 color attachment 
  VK_IMAGE_LAYOUT_PRESENT_SRC_KHR: 表示一个要被显示的swap chain image源 
  VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL: images 作为内存拷贝操作的目的
  我们将在文理(texture)部分再详细讲解这部分,现在仅需要明确的是image 在每个后续的操作中都要转变成特定的layout。
  initial / finalLaout表示渲染前后image的layout, VK_IMAGE_LAYOUT_UNDEFINED表示我们不在乎之前的layout,而VK_IMAGE_LAYOUT_PRESENT_SRC_KHR表示在渲染后,使用Swpa Chain 时,image 处于可显示的layout。

Subpasses and attachment references
  一个Render Pass 由一系列的subPass组成,并由subpass 处理随后的渲染操作,代表渲染的一个阶段,渲染命令存储在一个Render Pass的诸多subpass中,一个subpass的动作取决于上一个subpass 的处理结果,如果我们把它们打包成一个Render Pass ,Vulkan 能够为我们重新排序它们的执行顺序,节省内存带宽,从而可能获取更好的性能。
  每一个Subpass 引用一个或多个我们在前一节用VkAttachmentDescription 描述的attachment(s) ,每一个引用用VkAttachmentReference描述:
1
2
3
VkAttachmentReference colorAttachmentRef = {};
colorAttachmentRef.attachment = 0;
colorAttachmentRef.layout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL;
  attachment 表示此VkAttachmentReference 代表哪一个attachment,它的值是一个索引(index),这里我们只有一个VkAttachmentDescription,所以索引是0, 同时这个attachment 和Frame Shader 里的: layout(location = 0) out vec4 outColor相对应。layout 表示subpass 在执行时希望此atatchment具有的layout,Vulkan将自动为我们完成attachment的转换。我们想使用color buffer attachment,VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL选项还会做些优化。
Subpass 本身用VkSubpassDescription表示:
1
2
VkSubpassDescription subPass = {};
subPass.pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS;
  pipelineBindPoint表示要绑定的Pipeline类型,这里只支持两种类型:计算(compute)和图形(graphics),这里我们选择图形Pipeline。 然后我们告诉subpass 它所引用的attachment:
1
2
subPass.colorAttachmentCount = 1;
subPass.pColorAttachments = &colorAttachmentRef;

  Subpass 还可以引用的attachment有:
  pInputAttachments: 从着色器中读取的 attachment 
  pResolveAttachments: multiple color attachment 
  pDepthStencilAttachment: depth and stencil data 的attachment 
  pPreserveAttachments: 不被Subpass 使用,但出于某种原因需要保存

Create Render pass
  Attachment 和 Subpass 都已经声明好了,现在开始创建Render Pass:

1
2
3
4
5
6
7
8
9
10
11
VDeleter renderPass {device, vkDestroyRenderPass};
VkRenderPassCreateInfo renderPassInfo = {};
renderPassInfo.sType = VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO;
renderPassInfo.attachmentCount = 1;
renderPassInfo.pAttachments = &colorAttachment;
renderPassInfo.subpassCount = 1;
renderPassInfo.pSubpasses = &subPass;
//需要一个VkDevice 参数 即:Logical Device
if (vkCreateRenderPass(device, &renderPassInfo, nullptr, &renderPass) != VK_SUCCESS) {
    throw std::runtime_error("failed to create render pass!");
}

  不知道你的思绪是否有点混乱,让我们来理清一下。创建Render Pass 需要Subpass 和Attachment, 现在就简单的理解为Render Pass 包含Subpass 数组和 Attachment数组吧。Render Pass的工作需要Subpass 来完成, 每一个Subpass 可以操作多个Attachment ,怎么从Attachment数组中表示哪些attachment会被某个Subpass处理呢,所以我们需要一个VkAttachmentReference来描述attachment在Attachment数组中的小标和处理时的layout。

Create Pipeline
  先回顾一下这一章我们都做了什么:
  着色器阶段: 我们定义了 Shader module, 这是可编程部分。
  配置固定功能状态(fixed-function state): Pipeline 所需要的input assembly, rasterizer, viewport and color blending 配置, 这是不可编程部分。
  Pipeline layout: 主要是Uniform 变量,这些变量可以在绘画阶段动态改变。
  Render pass : Attachment 的定义以及他们的使用。
  现在已经到了本章的最终阶段 ->创建Pipeline :

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
VkGraphicsPipelineCreateInfo pipelineInfo = {};
pipelineInfo.sType = VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO;
pipelineInfo.stageCount = 2;
pipelineInfo.pStages = shaderStages;
 
pipelineInfo.pVertexInputState = &vertexInputInfo;
pipelineInfo.pInputAssemblyState = &inputAssembly;
pipelineInfo.pViewportState = &viewportState;
pipelineInfo.pRasterizationState = &rasterizer;
pipelineInfo.pMultisampleState = &multisampling;
pipelineInfo.pDepthStencilState = nullptr; // Optional
pipelineInfo.pColorBlendState = &colorBlending;
pipelineInfo.pDynamicState = nullptr; // Optional
 
pipelineInfo.layout = pipelineLayout;
pipelineInfo.renderPass = renderPass;
pipelineInfo.subpass = 0; //subpass索引,表示pipeline 将会使用renderPass 里的哪个subpass.
 
pipelineInfo.basePipelineHandle = VK_NULL_HANDLE;
pipelineInfo.basePipelineIndex = -1; // Optional

  Vulkan允许我们从已有的Pipeline派生新的Pipeline, 它们有很多相似点会使新Pipeline的创建花费更少。这里我们没有已存在的Pipeline。

1
2
3
4
5
6
// 创建Pipeline
VDeleter graphicsPipeline {device, vkDestroyPipeline};
if (vkCreateGraphicsPipelines(device, VK_NULL_HANDLE, 1, &pipelineInfo,
                              nullptr, &graphicsPipeline) != VK_SUCCESS) {
    throw std::runtime_error("failed to create graphics pipeline!");
}

  让我们来看看这个函数:

1
2
3
4
5
6
7
VkResult vkCreateGraphicsPipelines(
    VkDevice device,
    VkPipelineCache pipelineCache,
    uint32_t createInfoCount,
    const VkGraphicsPipelineCreateInfo* pCreateInfos,
    const VkAllocationCallbacks* pAllocator,
    VkPipeline* pPipelines);

  它允许我们传入多个 createInfo 来创建多个Pipeline, 第二个参数pipelineCache允许存储和重用数据,加速pipeine的创建过程。
  原文源码 : source code

如社区发表内容存在侵权行为,您可以点击这里查看侵权投诉指引