Ifrit-v2

GPU/CPU-Parallelized tile-based software rasterizer & raytracer.

Check gallery.md for More Example Pictures

Successor to following repos: Ifrit, Aria , Iris (Tiny Renderer CPP) & Iris (Tiny Renderer C#)

1. Structure

The project is organized into following parts.

• softrenderer: CUDA and CPU multithreaded SIMD software rasterizer & ray-tracer.

  • Covers culling, mesh shading pipeline (mesh shader), MSAA (up to 8x), mipmap, anisotropic filtering, SPIR-V JIT execution and shader derivatives.
  • CPU rasterizer is organized in TBR-like manner, utilize AVX2 instructions and tag buffer (with early-z) for performance gain.
  • For implementation details and performance, check here

• rhi/vkrenderer: Vulkan renderer, intended to be the refactored version for my original implementation.

  • Get rid of messy pass management, synchronization primitives and descriptor bindings in the original repo.

• meshproclib: Mesh algorithms.

  • Nanite-like mesh level of details

• ircompile: LLVM JIT compilation for shader codes.

• display: Presentation and window surface supporting utilities.

2. Setup / Run

WARN: Compiling files inside dev branch might yield UNEXPECTED result. Only x86-64 architecture Windows and Ubuntu are tested. It mainly covers Windows and some Linux systems. Other operation systems like MacOS are NOT supported.

2.1 Clone the Repository

git clone https://github.com/Aeroraven/Ifrit-v2.git --recursive 

2.2 Install Dependencies

Following dependencies should be manually configured. Other dependencies will be configured via submodule.

• OpenGL >=3.3

• CMake >=3.24

• Vulkan SDK 1.3, or at least include:

  • Core Features 1.3
  • shaderc combined 1.3
  • with mesh_shader extension

• LLVM 10 or LLVM 11 (Maybe higher version is OK, but LLVM 18 or higher might not work properly)

2.3 Quick Start

Note:

1. CUDA support is temporarily not included in repo's CMake.
2. Your compiler should support C++20 standards.

Under Refactoring, Linux GCC compilation MIGHT be unavailable now

cmake -S . -B ./build
cmake --build ./build

To run the demo

./bin/ifrit.demo

3. References & Acknowledgements

This project relies on following open-source projects. Corresponding licenses are in licenses folder.

Repository	Usage	License
stb	Image Parsing	MIT / Unlicensed
glfw3	Window and Display support	Zlib
spirv-headers	SPIR-V Standard Reference	MIT
glad	OpenGL Header Generation.	Generated files used
llvm-project	JIT Runtime	Apache 2.0
meshoptimizer	Mesh Algorithm	MIT
METIS	Mesh Algorithm / Graph Partition	Apache 2.0
Vulkan Memory Allocator	Memory Allocation	MIT
cereal	Serialization	BSD-3
tinygltf	Model Loading	MIT
stduuid	UUID	MIT
spirv-reflect	Shader Reflection	Apache 2.0
shaderc	Shader Compilation	Apache 2.0

And for debugging

• RenderDoc, for debugging and inspecting.

And some references that give inspirations. Thanks for their selfless dedications.

Software Rasterization：

1. https://tayfunkayhan.wordpress.com/2019/07/26/chasing-triangles-in-a-tile-based-rasterizer/
   1. https://github.com/NotCamelCase/Tyler
2. https://www.slideshare.net/slideshow/optimizing-the-graphics-pipeline-with-compute-gdc-2016/59747720
3. https://docs.nvidia.com/cuda/cuda-c-programming-guide/
4. https://llvm.org/docs/LangRef.html
5. https://www.mesa3d.org/
6. https://agner.org/optimize/

Modern Graphics Pipeline

1. https://advances.realtimerendering.com/s2021/Karis_Nanite_SIGGRAPH_Advances_2021_final.pdf
2. https://qiutang98.github.io/post/%E5%AE%9E%E6%97%B6%E6%B8%B2%E6%9F%93%E5%BC%80%E5%8F%91/mynanite01_mesh_processor/
   1. https://github.com/qiutang98/chord/tree/master
3. https://jglrxavpok.github.io/2024/01/19/recreating-nanite-lod-generation.html
4. https://lesleylai.info/en/vk-khr-dynamic-rendering/
5. https://vulkan-tutorial.com/
6. https://poniesandlight.co.uk/reflect/island_rendergraph_1/
   1. https://github.com/tgfrerer/island
7. https://dev.to/gasim/implementing-bindless-design-in-vulkan-34no
8. https://github.com/KhronosGroup/Vulkan-Samples