rapidsai/rmm

stdgpu: Efficient STL-like Data Structures on the GPU

CUDA Core Compute Libraries

🦀 GPU memory allocator for Vulkan, DirectX 12 and Metal. Written in pure Rust

CUDA Templates and Python DSLs for High-Performance Linear Algebra

cuDNN Frontend is NVIDIA's modern, open-source entry point to the cuDNN library and a growing collection of high-performance open-source kernels.

💥 Cython memory pool for RAII-style memory management

Thin, unified, C++-flavored wrappers for the CUDA APIs

RAFT contains fundamental widely-used algorithms and primitives for machine learning and information retrieval. The algorithms are CUDA-accelerated and form building blocks for more easily writing high performance applications.