{"library":"nvidia-cutlass-dsl","type":"library","category":null,"description":"NVIDIA CUTLASS Python DSL (version 4.4.2) is a Python-based domain-specific language (DSL) for writing high-performance CUDA kernels. It provides a Pythonic interface to CUTLASS's CuTe library, enabling kernel development with automatic JIT compilation to optimized PTX/SASS for NVIDIA GPUs (Ampere, Hopper, Blackwell architectures). It aims for zero-cost abstraction, performance comparable to C++ kernels, and seamless integration with deep learning frameworks like PyTorch and JAX. The library maintains an active development pace with frequent updates and minor version releases.","language":"python","status":"active","version":"4.4.2","tags":["cuda","gpu","deep learning","high performance","compiler","dsl","nvidia","kernel development"],"install":[{"cmd":"pip install nvidia-cutlass-dsl","imports":["import cutlass.cute as cute","from cutlass.cute import kernel","from cutlass.cute import jit","from cutlass.cute.runtime import from_dlpack"]},{"cmd":"pip install nvidia-cutlass-dsl[cu13]","imports":[]}],"homepage":null,"github":"https://github.com/NVIDIA/cutlass","docs":"https://github.com/NVIDIA/cutlass","changelog":null,"pypi":"https://pypi.org/project/nvidia-cutlass-dsl/","npm":null,"openapi_spec":null,"status_page":null,"smithery":null,"compatibility":{"summary":{"python_range":"3.10–3.9","success_rate":40,"avg_install_s":7.3,"avg_import_s":1.82,"wheel_type":"wheel"},"url":"https://checklist.day/v1/registry/nvidia-cutlass-dsl/compatibility"},"provenance":{"verified_status":"passing","verified_at":"Sun Jun 28","last_verified":"Sun Jun 28","next_check":"Tue Jul 28","install_tag":null}}