Introduction 简介

What is Unreal Engine 5.6?什么是 Unreal Engine 5.6？

Unreal Engine 5.6 is Epic Games' flagship real-time rendering and game engine, used in AAA games, film-quality cinematics, architectural visualization, and simulation. At its core, UE 5.6 ships three revolutionary rendering technologies — Nanite, Lumen, and Virtual Shadow Maps — that fundamentally change how real-time content is created and rendered. Understanding how these systems work at the source code level gives you a decisive advantage: you can debug rendering issues no tool can explain, extend the engine for custom effects, and make architectural decisions grounded in first principles rather than guesswork.

Unreal Engine 5.6 是 Epic Games 的旗舰实时渲染与游戏引擎，被用于 AAA 游戏、电影级过场动画、建筑可视化和仿真领域。 其核心搭载了三项革命性渲染技术——Nanite、Lumen 和Virtual Shadow Maps——从根本上改变了实时内容的创作和渲染方式。 理解这些系统在源码层面的工作原理能带来决定性优势：你可以调试任何工具都无法解释的渲染问题、为自定义效果扩展引擎， 并基于第一性原理而非猜测做出架构决策。

The Three Renderers — One Engine, Multiple Pipelines三大渲染器 — 一个引擎，多条管线

UE 5.6 does not have a single monolithic rendering path. Instead it has three distinct renderer classes that share common infrastructure but implement fundamentally different GPU strategies:

UE 5.6 没有单一的整体渲染路径。它有三个不同的渲染器类，共享公共基础设施，但实现根本不同的 GPU 策略：

The deferred renderer is the focus of most articles in this series. It is by far the most complex (its main Render() function is ~2300 lines and orchestrates 12 distinct render stages), but also the most powerful — it is the path used for all PC and console shipping titles in UE5.

延迟渲染器是本系列大多数文章的关注焦点。它是目前最复杂的（其主 Render() 函数约 2300 行， 编排 12 个不同的渲染阶段），但也是最强大的——它是 UE5 中所有 PC 和主机发行项目使用的路径。

Next-Gen Technologies — The Three Pillars次世代核心技术 — 三大支柱

UE 5.x's landmark technologies all solve the same fundamental problem: how do you render film-quality visuals in real time without crushing artists under technical constraints?

UE 5.x 的标志性技术都在解决同一个根本问题：如何在不给美术师施加技术约束的情况下实时渲染电影级画质？

The Infrastructure Layer — What Everything Runs On基础设施层 — 所有技术的运行基础

Beneath Nanite, Lumen, and VSM lies an equally important infrastructure that most engineers first encounter as an obstacle before understanding it as a powerful tool:

在 Nanite、Lumen 和 VSM 之下，有一套同样重要的基础设施，大多数工程师最初将其视为障碍， 后来才理解它是一个强大的工具：

• RDG (Render Dependency Graph): UE's frame-level GPU task graph. Every render pass is registered declaratively with its resource reads/writes. RDG automatically inserts barriers, culls unused passes, merges adjacent render passes, and schedules async compute. Understanding RDG is the single most important prerequisite for reading any UE rendering code.
• RDG（渲染依赖图）：UE 的帧级 GPU 任务图。每个渲染 Pass 以声明方式注册其资源读写。RDG 自动插入屏障、剔除未使用的 Pass、合并相邻渲染 Pass，并调度异步计算。理解 RDG 是阅读任何 UE 渲染代码最重要的前提条件。
• Three-Thread Model: Game Thread → Render Thread → RHI Thread. UE's rendering is always one frame behind the game logic. Understanding the thread boundaries is essential to avoid race conditions when extending the engine.
• 三线程模型：Game Thread → Render Thread → RHI Thread。UE 的渲染始终落后游戏逻辑一帧。理解线程边界对于在扩展引擎时避免竞争条件至关重要。
• Shader System: FShader type hierarchy, permutation domains, FHLSLMaterialTranslator that converts Material Expression graphs to HLSL, and the ShaderMap caching system (DDC).
• Shader 系统：FShader 类型层级、排列组合域、将材质表达式图转换为 HLSL 的 FHLSLMaterialTranslator，以及 ShaderMap 缓存系统（DDC）。
• GPUScene: A GPU-side structured buffer holding all primitive data, enabling GPU-driven instance culling without CPU readback.
• GPUScene：保存所有图元数据的 GPU 侧结构化缓冲区，无需 CPU 回读即可实现 GPU 驱动的实例裁剪。

How to Read This Series如何阅读本系列

Each article in this series traces a specific area of the rendering codebase from the top down: starting with the high-level function call that triggers the system, then diving into the data structures and GPU shaders that do the actual work. Every claim is tied to a specific source file and line number in UE 5.6.

本系列每篇文章都自上而下追踪渲染代码库的特定区域：从触发系统的高级函数调用开始，然后深入到完成实际工作的数据结构和 GPU 着色器。 每个论断都与 UE 5.6 中的特定源文件和行号相关联。

Prerequisites前置知识

• C++ fluency: UE uses heavy template metaprogramming, macros, and custom containers. You don't need to know UE's C++ conventions upfront — this series explains them as they appear — but basic C++ is essential.
• C++ 熟练度：UE 大量使用模板元编程、宏和自定义容器。你不需要事先了解 UE 的 C++ 惯例——本系列在出现时会解释它们——但基本 C++ 是必需的。
• Real-time rendering fundamentals: Understanding what a GBuffer is, what a render pass does, what vertex/pixel shaders are, and what "deferred shading" means at a conceptual level.
• 实时渲染基础：理解 GBuffer 是什么、渲染 Pass 做什么、顶点/像素着色器是什么，以及"延迟着色"在概念层面意味着什么。
• GPU architecture awareness: Knowing roughly how a GPU executes draw calls, what a compute shader is, and what VRAM bandwidth means. Knowledge of DirectX 12 or Vulkan is helpful but not required.
• GPU 架构认知：大致了解 GPU 如何执行绘制调用、计算着色器是什么，以及 VRAM 带宽意味着什么。了解 DirectX 12 或 Vulkan 有帮助但非必需。
• UE installed with source access: Engine source at Engine/Source/ and shader source at Engine/Shaders/. All file path references in this series assume UE 5.6.
• 安装了带源码访问的 UE：引擎源码位于 Engine/Source/，着色器源码位于 Engine/Shaders/。本系列所有文件路径引用均假设 UE 5.6。

Series Map系列地图