Dao-AILab/quack

DeepGEMM: clean and efficient BLAS kernel library on GPU

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

CUDA Templates and Python DSLs for High-Performance Linear Algebra

Tile primitives for speedy kernels

Fast and Furious AMD Kernels

Helpful kernel tutorials, examples and SKILLs for tile-based GPU programming

KernelBench: Can LLMs Write GPU Kernels? - Benchmark + Toolkit with Torch -> CUDA (+ more DSLs)

cuTile is a programming model for writing parallel kernels for NVIDIA GPUs

A kernel library written in tilelang