Back to homepage

Blog

Engineering, reflections, and reviews.

RSS
Writing activity
5 posts across 3 active months · last 12 months
  1. September 2025: 0 posts
  2. October 2025: 0 posts
  3. November 2025: 0 posts
  4. December 2025: 0 posts
  5. January 2026: 0 posts
  6. February 2026: 0 posts
  7. March 2026: 0 posts
  8. April 2026: 0 posts
  9. May 2026: 0 posts
  10. June 2026: 1 post
  11. July 2026: 3 posts
  12. August 2026: 1 post
Less More
Engineering 14 min read
GPU 如何执行 CUDA 程序?

沿一次 CUDA 任务从 CPU 提交到 GPU 完成的路径,理解 Stream、Grid、Block、Warp、SM、内存层级与同步机制。

GPUCUDAGPU ArchitectureCUDA Programming
Engineering 10 min read
DeepSeek Sparse Attention:从 SWA 到稀疏 Attention 计算

从 SWA 的固定窗口出发,解释 Lightning Indexer 如何为完整前缀打分并选出 top-k、融合式 Sparse Attention 算子如何按索引读取 KV,以及 DSA 如何降低 Attention 的计算量与显存读取量。

DeepSeek Sparse AttentionSWAAttentionTransformer