我们在使用vulkan时,往往被繁琐的概念和大量的参数设置困扰,这个库的主要目的是创建一套能快速构建Vulkan对象的构造器,使用这些builder可以快速创建常用的,带有良好设计默认参数的,带有常用配置方案的对象。从而可以快速创建可用的渲染流水线。
我们对Instance,Device等对象都进行了包装,将其和其他常用对象打包,使得很多API调用时可以减少参数的使用,尤其是减少参数的传入错误
设计思路:
- 先提供常用的参数,允许默认构建性能一般但可用的对象,提供接口对构造时进行特化指定
- 大量使用流式API
- 通过打包其他对象,尽量减少接口调用时的参数
- 明确哪些对象和其他对象的约束关系,在接口中体现这种约束关系,尽量使得不符合约束关系的调用无法被编译
- 使用模板减少代码量
使用VKBuilder可以快速创建实例,设备,渲染界面,交换链等等Vulkan对象。
#define VKB_IMPL
#include "vkbuilder.hpp"
class Render {
public:
void init(void *window) {
// 首先创建构造器,选择API版本,需要的layer,配置debug messenger
vkb::InstanceBuilder builder;
inst = builder.require_api_version(1, 0)
.request_validation_layers() // validate correctness
.use_default_debug_messenger() // use DebugUtilsMessage
.build();
// 创建surface用于渲染
auto surface = (VkSurfaceKHR)create_surface_glfw(inst.instance, window);
// 创建一个PhysicalDeviceSelector用来选择渲染设备
vkb::PhysicalDeviceSelector selector{inst};
auto phys = selector.set_surface(surface)
.set_minimum_version(1, 0) // require a vulkan 1.0 capable device
.select();
vkb::DeviceBuilder device_builder{phys};
// 构造设备
// automatically propagate needed data from instance & physical device
device = device_builder.build();
create_swapchain();
create_pipeline();
}
void create_swapchain() {
// 创建默认交换链
vkb::SwapchainBuilder swapchain_builder{device};
auto swap_ret = swapchain_builder.set_old_swapchain(swapchain).build();
swapchain.destroy();
swapchain = swap_ret;
}
void create_pipeline() {
// 创建渲染Pass
vkb::RenderPassBuilder renderpass_builder{device};
renderpass = renderpass_builder
.addPresentAttachment(swapchain.image_format, vk::AttachmentLoadOp::eClear)
.addSubpass(vkb::SubpassBuilder()
.addAttachmentRef(0, vk::ImageLayout::eColorAttachmentOptimal))
.addDependency(VK_SUBPASS_EXTERNAL, 0)
.build();
auto vert_code = readFile("vert.spv");
auto frag_code = readFile("frag.spv");
// 创建渲染流水线
vkb::PipelineBuilder pipeline_builder{device, swapchain};
pipeline = pipeline_builder
.useClassicPipeline(vert_code, frag_code)
.setVertexInputState(
vkb::VertexInputStateBuilder() // VertexInputStateBuilder 可以根据Vertex上的函数进行创建
.addInputBinding<Vertex>()
.addAttributeDescription<Vertex>()
)
.build(renderpass);
vkb::PresentBuilder present_builder{device, swapchain};
present = present_builder.build(renderpass);
initVertex();
}
vkb::Instance inst;
vkb::Device device;
vkb::Swapchain swapchain;
vk::RenderPass renderpass;
vk::Pipeline pipeline;
vkb::Present present;
std::vector<Vertex> v;
vkb::HostVertexBuffer buffer;
};
Render render;
void onRender() {
render.render();
}
void initVulkan(void *window) {
render.init(window);
}
int main() {
GLFW_Window window;
window.createWindow();
initVulkan(window.getWindow());
window.mainLoop();
return 0;
}
对于顶点数据结构,只需要创建getBindingDescription
函数与getAttributeDescription
函数,就可以将该结构体注册到我们的系统中来。
struct Vertex {
glm::vec2 pos;
glm::vec3 color;
static vk::VertexInputBindingDescription
getBindingDescription(uint32_t binding) {
vk::VertexInputBindingDescription bindingDescription{};
bindingDescription.binding = binding;
bindingDescription.stride = sizeof(Vertex);
bindingDescription.inputRate = vk::VertexInputRate::eVertex;
return bindingDescription;
}
static std::vector<vk::VertexInputAttributeDescription>
getAttributeDescription(uint32_t binding) {
std::vector<vk::VertexInputAttributeDescription> attrs = {
vk::VertexInputAttributeDescription(0, binding, vk::Format::eR32G32Sfloat, offsetof(Vertex, pos)),
vk::VertexInputAttributeDescription(1, binding, vk::Format::eR32G32B32Sfloat, offsetof(Vertex, color))
};
return attrs;
}
};
void initVertex() {
v.resize(3);
v[0].pos = {0, 0.5}; v[0].color = {1,0,0};
v[1].pos = {-0.5, -0.5}; v[1].color = {0,1,0};
v[2].pos = {0.5, -0.5}; v[2].color = {0,0,1};
// create Vertex buffer
buffer.allocate<Vertex>(device, v);
}
渲染控制也非常轻松:
void render() {
present.begin();
present.beginRenderPass(renderpass);
vk::DeviceSize offset(0);
auto& cb = present.getCurrentCommandBuffer();
cb.bindPipeline(vk::PipelineBindPoint::eGraphics, pipeline);
cb.bindVertexBuffers(0, 1, buffer, &offset);
cb.draw(3, 1, 0, 0);
present.endRenderPass();
present.end();
present.drawFrame();
}