Awesome GitHub RepositoriesTiled Memory Access Patterns

Triton is a parallel computing framework and high-level programming language designed for writing custom compute kernels. It functions as a deep learning compiler, translating complex mathematical operations into high-throughput instructions that maximize hardware utilization and memory efficiency on graphics processing units. The framework distinguishes itself through a hardware-agnostic compute abstraction that allows developers to define kernels without manual low-level tuning. It employs just-in-time compilation to generate optimized binary instructions at runtime, utilizing static data f

AISystem is a comprehensive AI full-stack infrastructure project covering the entire pipeline from AI chip architecture to high-level training frameworks. It encompasses the development of AI compiler frameworks, inference engines, and distributed training orchestrators designed to coordinate workloads across a heterogeneous compute stack of CPUs, GPUs, and NPUs. The project focuses on the deep integration of software and hardware, employing software-hardware co-design to align tensor layouts with physical memory structures. It provides specialized capabilities for accelerating Transformer mo

This project is a comprehensive educational resource and programming course covering C++ language semantics and features from C++03 through C++26. It provides structured tutorials and technical guides focused on modern C++ development. The material offers specialized instruction on template metaprogramming, including the use of type traits and compile-time computations. It features detailed guides on concurrency and parallelism for multi-core execution, as well as a reference for software design applying SOLID principles and RAII. Additionally, it covers build performance optimization to redu

Cutlass is a collection of C++ templates and Python interfaces for implementing high-performance linear algebra operations on NVIDIA GPUs. It provides a kernel composition framework for designing custom GPU kernels and a mixed-precision tensor library capable of executing operations across diverse data formats, ranging from 64-bit floating point to 4-bit integers. The project features a toolkit for operator fusion that integrates activation functions and bias calculations directly into matrix multiplication kernels to reduce memory passes. It also includes a Python-based domain-specific langu

LeetCUDA is a collection of high-performance GPU kernel libraries focusing on memory optimization, activation functions, and attention mechanisms. It serves as a reference library for CUDA kernel implementations, ranging from basic element-wise operations to complex neural network components, and provides Python bindings to integrate these kernels into deep learning workflows. The project is distinguished by its focus on low-level hardware optimizations. This includes the use of tensor cores for half-precision matrix multiplication, asynchronous data pipelining with double buffering, and shar

XiangShan is a high-performance RISC-V processor core and a hardware description language framework. It provides a construction-based system for designing, simulating, and verifying complex processor micro-architectures and peripheral devices. The project includes a high-performance CPU simulator used for architectural exploration and functional verification of processor execution. The project implements a superscalar out-of-order CPU architecture that uses renaming and reorder buffers to execute instructions in parallel. It generates synthesizable Verilog files from hardware descriptions to

Flashlight est une bibliothèque de machine learning et de tenseurs autonome en C++ utilisée pour construire et entraîner des réseaux de neurones. Elle fonctionne comme un framework complet de réseaux de neurones et un moteur de différenciation automatique, fournissant les outils pour construire des graphes de calcul et calculer les gradients via la rétropropagation. Le projet sert de framework d'entraînement distribué, utilisant des opérations all-reduce pour synchroniser les gradients et les paramètres sur plusieurs nœuds de calcul et appareils. Il se distingue par une intégration profonde de la manipulation de tenseurs haute performance, l'interopérabilité native de la mémoire des appareils et un système pour synchroniser les poids entre les workers distribués afin d'accélérer l'entraînement de modèles à grande échelle. Le framework couvre un large éventail de capacités de deep learning, incluant la composition modulaire de couches pour concevoir des architectures complexes comme des blocs résiduels et des cellules récurrentes. Il fournit des utilitaires étendus de gestion de données pour l'ingestion et le préchargement, ainsi que des systèmes de sérialisation pour persister les états de modèle. De plus, il inclut une suite d'outils de surveillance et d'observabilité pour suivre les métriques d'entraînement et mesurer les erreurs de séquence. La bibliothèque est implémentée en C++.

TileLang is a Python-embedded domain-specific language compiler that JIT-compiles and autotunes GPU kernels. It uses a tile-based DSL, automatic software pipelining, and parallel autotuning to generate optimized GPU kernels at runtime. It supports tensor core operations with Pythonic syntax, automatic memory management, and thread mapping. The compiler searches over tile sizes, thread counts, and scheduling policies, compiling and benchmarking candidates in parallel to find the fastest kernel. It also caches compiled binaries and tuning results to disk for reuse across sessions. TileLang inc

perf-ninja is a collection of educational resources and curricula focused on CPU architecture, memory hierarchies, SIMD programming, and low-level performance engineering. It provides instructional material and practical labs for identifying and fixing CPU bottlenecks, such as cache misses and branch mispredictions. The project differentiates itself through specialized training in hardware-level optimizations, including the use of compiler intrinsics for SIMD vectorization and the implementation of branchless predicate execution to eliminate pipeline stalls. It also covers advanced binary-lev