Write a CuTe DSL GPU Kernel

Goal

Use the bundled CuTe DSL API snapshots in this skill and the workspace CUTLASS checkout to design, implement, debug, and integrate CuTe DSL GPU kernels in a way that is reusable across projects, including cache-dit.

When to Use

Use this skill when you need to:

• write or modify a CuTe DSL GPU kernel in Python
• study CuTe DSL types, runtime helpers, architecture APIs, or pipeline abstractions
• port or rewrite an existing CUDA or C++ operator into CuTe DSL
• use CuTe DSL examples from the workspace CUTLASS checkout as reference material
• debug CuTe DSL compilation, runtime behavior, layout issues, or integration problems

Do not use this skill for: