Blogs

I plan to write blogs during my process of learning CUDA/Cutlass/Cute/Triton programming. Stay tuned for more!

• [New] Blackwell MMA SMEM Descriptor

• 4 Ways to Do CuTe Copy

• GPU Memory Consistency Model

• Tensor Core MMA Swizzle Layout [中文版]

• Using Programmatic Dependent Launch (PDL) to Reduce End-to-End Latency [中文版]

• CuTe Layout and Tensor

• How to Make a Compute-bound Problem Actually Compute-bound

• Using TMA Load, Prefetch, and Multicast Load in CuTe