CUDA® is a parallel computing platform and programming model developed by NVIDIA for general computing on graphical processing units (GPUs). With CUDA, developers are able to dramatically speed up computing applications by harnessing the power of GPUs.
Instant neural graphics primitives: lightning fast NeRF and more
cuGraph - RAPIDS Graph Analytics Library
GPU Accelerated t-SNE for CUDA with Python bindings
FlashInfer: Kernel Library for LLM Serving
Quantized Attention that achieves speedups of 2.1-3.1x and 2.7-5.1x compared to FlashAttention2 and xformers, respectively, without lossing end-to-end metrics across various models.
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.
Fuse multiple depth frames into a TSDF voxel volume.
A throughput-oriented high-performance serving framework for LLMs
CUDA Kernel Benchmarking Library
Graphics Processing Units Molecular Dynamics
MegBA: A GPU-Based Distributed Library for Large-Scale Bundle Adjustment
PopSift is an implementation of the SIFT algorithm in CUDA.
Neighborhood Attention Extension. Bringing attention to a neighborhood near you!
A simple GPU hash table implemented in CUDA using lock free techniques
Several optimization methods of half-precision general matrix multiplication (HGEMM) using tensor core with WMMA API and MMA PTX instruction.
Created by Nvidia
Released June 23, 2007