Guide writing, debugging, and optimizing GPU kernels with NVIDIA CuTe DSL.
npx skills add https://github.com/Shekswess/gpu-kernel-skills --skill cutedsl
Please help me install this Agent Skill. Skill: cutedsl Source: https://github.com/Shekswess/gpu-kernel-skills/tree/main/cutedsl Command: npx skills add https://github.com/Shekswess/gpu-kernel-skills --skill cutedsl
NVIDIA CuTe DSL enables writing CUTLASS-grade GPU kernels in Python. It compiles through AST rewrite + tracing → MLIR → ptxas, delivering the same PTX/SASS quality as C++ CUTLASS with dramatically faster compilation.
Provide a CuTe DSL kernel outline or review a CuTe DSL snippet and suggest improvements.
High-intent search queries and answers about installing and using this skill.
You can port C++ CUTLASS kernels to Python CuTe DSL by translating the core logic and utilizing the Skill's API references to map CUTLASS C++ structures into equivalent Python CuTe DSL patterns.
Yes, CuTe DSL supports debugging tiled MMA and copy workflows by applying architecture-specific guidance and referencing kernel patterns to structure your approach for TMA and WGMMA operations.
CuTe DSL compiles through AST rewrite, tracing, and MLIR to ptxas, delivering the same high-quality PTX/SASS output as C++ CUTLASS but with dramatically faster compilation times.
The best way to structure a CuTe DSL kernel for tcgen05 is to read the architecture-specific reference materials and use the frontmatter-driven entry points to apply the correct kernel patterns.
Porting C++ CUTLASS kernels to Python CuTe DSL may fail if architecture-specific API details are missed, requiring a review of the api_reference.md and kernel_patterns.md to ensure correct mapping.
slowlyCDevelop, debug, and optimize CUTLASS and CuTeDSL GPU kernels for NVIDIA architectures.
LancerLabBuild, run, and profile CuTe/CUTLASS C++ GPU kernels with nvcc.
LancerLabTune and validate Python JIT GPU kernels using CuTe DSL.
yo-stevenWrite and implement GPU kernels using NVIDIA CuTe DSL for element-wise, GEMM, and reduction operations.
