Jingkun Zhang

Jingkun Zhang

Posts

2026

Understanding GEMM on Blackwell with CuTeDSL

10 minute read

Published:

In this post I want to walk through how a high-performance GEMM is structured on Blackwell, using the official CUTLASS dense GEMM example as a case study. This kernel is written in CuTeDSL — NVIDIA’s Python DSL for authoring CUTLASS kernels — and it packs in nearly every Blackwell-specific optimization available.

Optimizing an NVFP4 Group GEMM Kernel on Blackwell

6 minute read

Published:

Happy Chinese New Year! Besides celebrating CNY this weekend, I also spent some time working on NVFP4 block-scaled group GEMM kernel optimizations on NVIDIA B200. This is from a GPU Mode competition where participants optimize CUDA kernels on Blackwell GPUs. The code is written in CuTeDSL, CUTLASS’s Python DSL for Blackwell kernels.