- 神奇妙妙通信小问题 10-09-2026
- 从 NCCL Ring LL 到 PCIe IPC AllReduce 10-03-2026
- KV Size on sm10+ 09-12-2026
- 对 Looped Transformer 的简要学习笔记 09-06-2026
- 尝试对Kimi KDA的数学推导&算子实现分析 08-13-2026
- Flash Attention CUDA Kernel 优化: 从 56 到 986 TFLOPS on B300 — G1: 一切的开始 (56→130) 07-30-2026
- Flash Attention CUDA Kernel 优化: 从 56 到 986 TFLOPS on B300 — G2: 从算术指令到 Occupancy 的挣扎 (130→144) 07-30-2026
- Flash Attention CUDA Kernel 优化: 从 56 到 986 TFLOPS on B300 — G3: Fragment Reshape 与架构级突破 (144→227) 07-30-2026
- Flash Attention CUDA Kernel 优化: 从 56 到 986 TFLOPS on B300 — G4: SM103 + TMA + TCGen05 (227→221) 07-30-2026
- Flash Attention CUDA Kernel 优化: 从 56 到 986 TFLOPS on B300 — G5: Pipeline 重构与双缓冲 (221→727) 07-30-2026
- Flash Attention CUDA Kernel 优化: 从 56 到 986 TFLOPS on B300 — G7: 回归 1CTA (880→987) 07-30-2026
- Flash Attention CUDA Kernel 优化: 从 56 到 986 TFLOPS on B300 — G6: Cluster 与 2CTA 探索 (727→880) 07-30-2026
- TopK kernel 优化: from 20GB/s to 2024GB/s on B300 07-06-2026
- Paged attention kernel optimization(I) 05-22-2026
- MoE vs Dense Models in Inference 05-21-2026
- Streams and Concurrency on CUDA 05-21-2026
- Foundation of Reinforcement learning(V) 05-18-2026
- Foundation of Reinforcement learning(IV) 05-18-2026
- Foundation of Reinforcement learning(III) 05-17-2026
- Foundation of Reinforcement learning(II) 05-17-2026
- Foundation of Reinforcement learning(I) 05-16-2026
- nanoPD:一个 LLM P/D 分离推理引擎的实现笔记 04-11-2026
- 3D Reconstruction Series 03-15-2026
- 学习笔记：Tensor Parallelism（TP） 03-13-2026
- HPCGames 题解 D E 题 02-05-2026
- HPCGames 题解 A B C 题 02-04-2026
- SF3D 论文阅读记录 11-29-2025
- ViT Transformer 的阅读?(应该算是阅读吧) 11-25-2025
- 回顾一下Transformer 11-24-2025
- SLAM Former 阅读 11-01-2025
- 重返vggt 10-31-2025
- 论文阅读记录：reloc3r 09-06-2025
- 论文阅读记录：Fast3R 09-04-2025
- 论文阅读记录：MAst3R 09-02-2025
- VGGT读后有感 08-14-2025
- 为SLAM3R补充实时处理函数方法 08-12-2025
- SLAM3R读后有感 08-02-2025
- Celebrate and Introduce My First Page 08-01-2025