📚LeetCUDA: Modern CUDA Learn Notes with PyTorch for Beginners🐑, 200+ CUDA Kernels, Tensor Cores, HGEMM, FA-2 MMA.🎉

DeepEP: an efficient expert-parallel communication library

DeepGEMM: clean and efficient FP8 GEMM kernels with fine-grained scaling

[ICLR2025, ICML2025, NeurIPS2025 Spotlight] Quantized Attention achieves speedup of 2-5x compared to FlashAttention, without losing end-to-end metrics across language, image, and video models.

how to optimize some algorithm in cuda.

Mirage Persistent Kernel: Compiling LLMs into a MegaKernel

cuGraph - RAPIDS Graph Analytics Library

[ARCHIVED] Cooperative primitives for CUDA C++. See https://github.com/NVIDIA/cccl

Fast CUDA matrix multiplication from scratch

Flash Attention in ~100 lines of CUDA (forward pass only)

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 …

cuVS - a library for vector search and clustering on the GPU

Fastest kernels written from scratch

Static suckless single batch CUDA-only qwen3-0.6B mini inference engine

FlashKDA: high-performance Kimi Delta Attention kernels

[ICML 2024] Quest: Query-Aware Sparsity for Efficient Long-Context LLM Inference

[MLSys'24] Atom: Low-bit Quantization for Efficient and Accurate LLM Serving

mHC kernels implemented in CUDA

CPM.cu is a lightweight, high-performance CUDA implementation for LLMs, optimized for end-device inference and featuring cutting-edge techniques in sparse architecture, speculative sampling and qua…

Approximate nearest neighbor search with product quantization on GPU in pytorch and cuda

From Minimal GEMM to Everything

CUDA implementation of Hierarchical Navigable Small World Graph algorithm

GGNN: State of the Art Graph-based GPU Nearest Neighbor Search

FP64 equivalent GEMM by the Ozaki scheme with Int8 Tensor Cores

Benchmark code for the "Online normalizer calculation for softmax" paper

DeeperGEMM: crazy optimized version

🌈 Solutions of LeetGPU

Efficient and unified implementations for TopK-based sparse attention

High-Throughput, Cost-Effective Billion-Scale Vector Search with a Single GPU [SIGMOD'26]