deepseek-ai / FlashMLA

FlashMLA: Efficient Multi-head Latent Attention Kernels

wootbeer / snprecomp

Sin and Punishment N64 Recomp

Ritosa2333 / eagler-touhou

An initiative to reconstruct and port official Touhou Project games and its derivatives to every platform.

Shmargus / GSRecomp

—

martinus / unordered_dense

A fast & densely stored hashmap and hashset

thpatch / thcrap

Touhou Community Reliant Automatic Patcher

Soreepeong / XivAlexander

Double weave on high latency, and mishmash of modding tools - especially for fonts and internationalization for Final Fantasy XIV.

Flow-IPC / ipc

[Start here!] Flow-IPC - Modern C++ toolkit for high-speed inter-process communication (IPC)

gradientspace / Cotangent2

Open-Source AI-Accelerated Mesh and Texture Editing, and spiritual successor to Autodesk Meshmixer

shmimel / whale-writer

—

NVIDIA / nvshmem

NVIDIA NVSHMEM is a parallel programming interface for NVIDIA GPUs based on OpenSHMEM. NVSHMEM can significantly reduce multi-process communication and coordination overheads by allowing programmers to perform one-sided communication from within CUDA kernels and on CUDA streams.