https://yiwen-cai.github.io/tags/%E6%80%A7%E8%83%BD%E4%BC%98%E5%8C%96/Recent content in 性能优化 on Yiwen CaiHugo -- gohugo.iozh-cncaiyiwen.cs@foxmail.com (Yiwen Cai)caiyiwen.cs@foxmail.com (Yiwen Cai)© 2026 Yiwen CaiFri, 10 Jul 2026 21:12:36 +0800https://yiwen-cai.github.io/notes/triton/cs336-kernels-triton/Sun, 14 Jun 2026 00:00:00 +0000caiyiwen.cs@foxmail.com (Yiwen Cai)https://yiwen-cai.github.io/notes/triton/cs336-kernels-triton/从硬件抽象进入 kernel 编程实践：warp/occupancy/bank conflict/coalescing 如何映射到性能，benchmarking 与 profiling 方法论，以及用 Triton 实现 GeLU、softmax、row sum、matmul+ReLU 四个 kernel。https://yiwen-cai.github.io/notes/cuda/cs336-gpu-attention/Sat, 13 Jun 2026 00:00:00 +0000caiyiwen.cs@foxmail.com (Yiwen Cai)https://yiwen-cai.github.io/notes/cuda/cs336-gpu-attention/从模型结构转入系统视角：为什么 LLM 必须依赖 GPU、GPU 的执行模型与内存层次、arithmetic intensity 与 roofline model，以及如何用这套性能模型解释 FlashAttention 这类 IO-aware 算法。