Open-source book with Modern CUDA Learn Notes for Beginners, includes FP16/BF16, FP8, HGEMM, FlashAttention, CuTe, etc.
Kernl lets you run PyTorch transformer models several times faster on GPU with a single line of code, and is designed to be easily hackable.
Fast, differentiable sorting and ranking in PyTorch
row-major matmul optimization
GPU 性能与 AI Infra 学习项目：CUDA/Triton 算子、NCU/NSYS、vLLM/SGLang/TRT-LLM/ms-swift、PyTorch/DeepSpeed/ms-swift 训练、并行架构
Direct NVIDIA CUDA access for PHP. GPU-accelerated tensors, JIT-compiled kernels, and async execution—no Python required
A performance comparison of standard matrix functions between CPU and GPU using Nvidia CUDA on Visual Studio using C++
A beginner's guide to CUDA programming
frontier.fast challenge repo for agents and submissions to push the frontier of AI!
Rust KV-cache compression for LLM inference. Implements TurboQuant (Zandieh et al., ICLR 2026) plus PQO — our variant that drops QJL, adds a fused CUDA kernel, and shrinks the cache to ~20% of FP16 (49% total VRAM at 32K). mistral.rs integration.
A collection of scripts on MLSys-related stuff
SNU CSE Scalable High Performance Computing (M1522.006700) - 2023 Autumn
a custom CUDA kernel for windowed matrix multiplication
An unofficial re-implementation of Cursor's warp decode MoE inference technique. Achieved nearly 2.0x speedup on MOE layer on Nvidia H20 at batch size 1.
Winning submission for StartHack 2024: HPC optimized multi-GPU/CPU inference
原生C++/CUDA/OpenBLAS实现的张量加速计算库。性能碾压torch（原生CUDA算子实现+0拷贝优化），且更加方便部署。
A production-ready, high-performance CUDA Runtime and Driver API library for Zig.
To associate your repository with the cuda-kernel topic, visit your repo's landing page and select "manage topics."