LLM training in simple, raw C/CUDA

A massively parallel, optimal functional runtime in Rust

Tile primitives for speedy kernels

Mirage Persistent Kernel: Compiling LLMs into a MegaKernel

[MICRO'23, MLSys'22] TorchSparse: Efficient Training and Inference Framework for Sparse Convolution on GPUs.

Efficient GPU kernels for block-sparse matrix multiplication and convolution

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

High-speed GEMV kernels, at most 2.7x speedup compared to pytorch baseline.

PyTorch half precision gemm lib w/ fused optional bias + optional relu/gelu

FlashInfer: Kernel Library for LLM Serving