unknown

NVIDIA's CUDA Core Compute Libraries (CCCL) unifies three essential CUDA C++ libraries—Thrust, CUB, and libcudacxx—into a single repository. Thrust provides high-level parallel algorithms, CUB offers low-level CUDA-specific primitives for custom kernels, and libcudacxx implements the C++ Standard Library for both host and device code. The project is header-only, supports CUDA Toolkit versions 11.1+ and 12.x, requires C++17 or C++20, and maintains backward compatibility across CTK major versions. Installation options include the CUDA Toolkit, GitHub, conda, and CMake integration. The libraries follow semantic versioning with API breaking changes only occurring with new CTK major releases, while ABI stability is not guaranteed for thrust/cub namespaces.

NVIDIA/cccl: CUDA Core Compute Libraries

Become a cool developer with Dev Source! Your ultimate dev source for resources, insights, and a thriving community to learn, grow, and stay ahead every single day!