sgl-project/DeepGEMM

DeepGEMM: clean and efficient BLAS kernel library on GPU

CUDA Templates and Python DSLs for High-Performance Linear Algebra

A Quirky Assortment of CuTe Kernels

CODA: Rewriting Transformer Blocks as GEMM-Epilogue Programs

An agent harness that compiles a model into one provably-correct, self-retargeting CUDA megakernel and self-tunes it past cuBLAS at batch-1 LLM decode, paper: https://arxiv.org/abs/2606.09682

High Performance LLM Inference Operator Library

A library for accelerating Transformer models on NVIDIA GPUs, including using 8-bit and 4-bit floating point (FP8 and FP4) precision on Hopper, Ada and Blackwell GPUs, to provide better performance with lower memory utilization in both training and inference.

DeepGEMM-Ascend: clean and efficient matrix multiplication kernel library for Huawei Ascend NPUs

Humming is a high-performance, lightweight, and highly flexible JIT (Just-In-Time) compiled GEMM kernel library specifically designed for quantized inference.