[Docs] PyPTO 性能优化案例：DeepSeek-V4 MoE 预处理算子融合与同步开销复盘 - #825

Open

Little-oil wants to merge 3 commits into

hw-native-sys:mainfrom

Little-oil:merge-moe

Commits on Jul 27, 2026

Perf: fuse DeepSeek V4 MoE preprocessing stages
Youhezhen
committed
4ab10f4
Perf: use 8-AIV soft barriers in fused pre-norm
Youhezhen
committed
3192c95

Commits on Jul 29, 2026

feat(deepseek): migrate fused pre norm to atomic soft sync
Youhezhen
committed
90ff155