←Back to projectsTechnical project
Optimized Software Rasterizer
CPU rasterizer optimization using pipeline improvements, eight-pixel SIMD shading, tile scheduling, and a custom thread pool.
- C++
- AVX SIMD
- Multithreading
Rasterizer benchmark media
Project overview
- Focus
- CPU Rendering Performance
- Language
- C++
- Optimization
- AVX SIMD / Multithreading
- Result
- Roughly 2-4x over baseline
Key work
- Added vertex cache, pre-normalized lighting, Early-Z, backface and frustum culling.
- Vectorized eight-pixel shading and lighting with AVX SIMD.
- Implemented tile-based rendering through a custom thread pool and compared FPS, P99 latency, and speedup.
关键工作
- 加入顶点缓存、预归一化光照、Early-Z、背面剔除与视锥剔除。
- 使用 AVX SIMD 对八像素着色与光照进行向量化。
- 通过自研线程池实现分块渲染,并对比 FPS、P99 延迟与加速比。
View all projects→