Awesome GitHub RepositoriesZero-Copy Memory Mappings

Meilisearch is a Rust-based search engine providing typo-tolerant full-text and vector-based semantic search with real-time conversational capabilities.

simdjson is a high-performance, header-only C++ library designed for parsing, querying, and serializing JSON data with minimal memory overhead. It functions as a hardware-aware data processing engine that leverages vector instructions to achieve gigabyte-per-second parsing speeds. By detecting host processor capabilities at runtime, the library automatically selects the most efficient instruction sets to accelerate structural analysis and validation. The library distinguishes itself through a focus on extreme efficiency and resource management. It utilizes memory mapping and padded buffer ali

Candle is a minimalist machine learning framework and deep learning inference engine designed for the Rust programming language. It functions as a low-level tensor computation library, providing the necessary primitives for multi-dimensional array operations and mathematical transformations required to execute pre-trained neural network models. The framework distinguishes itself through a focus on memory efficiency and hardware utilization. It employs static-typed tensor operations to enforce shape validation and memory safety at compile time, while utilizing a lazy-loaded computational graph

This project is a cross-platform graphics and compute framework that provides a unified, hardware-agnostic abstraction layer for rendering and parallel processing. It enables developers to build high-performance applications that execute consistently across diverse operating systems and hardware backends, including Vulkan, Metal, and DirectX. By mapping high-level graphics commands to native APIs, it serves as a portable foundation for both real-time 3D rendering and general-purpose GPU computing. The framework distinguishes itself through a robust architecture that supports both native deskt

Realm-Cocoa is a NoSQL mobile database engine and reactive object database designed for local data storage on mobile devices. It serves as a non-relational alternative to Core Data and SQLite, storing data as objects rather than tables. The system functions as an encrypted local store that protects sensitive application data using encryption. It provides reactive data synchronization, allowing application objects and user interfaces to update automatically when the underlying database changes.

This is a mobile object database and NoSQL local data store that replaces relational tables with a schema-based model. It functions as a reactive data store, using live object observations and change notifications to trigger automatic user interface refreshes. The system provides built-in mobile cloud data synchronization to keep local datasets consistent with a remote server across multiple devices. It also includes security features for encrypted local storage, protecting sensitive on-disk data using at-rest encryption keys and fine-grained access control. Broad capabilities include object

Mold is a high-performance linker designed to replace standard system tools for the creation of executable binaries and shared libraries. It functions as a drop-in replacement for existing linkers, focusing on accelerating the final build phase of large software projects to improve developer productivity. The tool achieves its performance by utilizing multi-threaded processing to distribute the linking of object files across multiple CPU cores. It supports cross-architecture binary linking, allowing it to process compiled files for diverse platforms efficiently. By intercepting standard linke

Redpanda is a distributed event streaming engine designed to serve as a high-performance, drop-in replacement for existing event-driven architectures. It provides a foundation for building and scaling applications that require reliable data movement, analytical querying, and strict operational compliance across both cloud and self-managed environments. The platform distinguishes itself through a shared-nothing architecture that utilizes thread-per-core execution and a non-blocking asynchronous input/output engine to maximize throughput. It maintains data consistency through a consensus-based

OpenVINO is an AI inference engine and model serving platform designed to execute optimized deep learning models across CPUs, GPUs, and NPUs through a unified API. It includes a model optimization toolkit for converting, quantizing, and compressing models from various frameworks, alongside a specialized generative AI runtime for large language models. The project distinguishes itself through a plugin-based hardware acceleration layer that maps neural network operations to vendor-specific drivers. It features advanced execution mechanisms such as continuous batching, speculative decoding, and

Seastar is a C++ server application framework and asynchronous programming library designed for building high-performance, shared-nothing server applications. It functions as a high-performance I/O engine providing direct disk and network access through a shared-nothing framework that partitions data and execution across CPU cores. The framework distinguishes itself through a thread-per-core architecture that eliminates locking and resource contention by assigning one execution thread to each physical CPU core. It implements a userspace TCP/IP stack and kernel-bypass techniques, integrating w

Sled is an embedded key-value store and ACID-compliant database designed for high-performance data persistence. It functions as a log-structured storage engine that organizes data using B+ trees to support efficient range queries and prefix scans. The engine implements a zero-copy data store model, utilizing epoch-based reclamation to provide direct references to cached values without memory allocations. It distinguishes itself through a combination of write-ahead logging, page cache optimizations to reduce write amplification on flash storage, and serializable transactions for atomic multi-k

Crossbeam is a concurrency toolkit for Rust providing low-level primitives for writing multi-threaded programs. It focuses on lock-free data structures and memory management primitives designed for shared-memory concurrent environments. The project includes a work-stealing scheduler that uses double-ended queues to balance workloads across multiple processor cores. This system enables the implementation of work-stealing deques to distribute tasks and prevent bottlenecks. The toolkit covers broader capabilities for parallel algorithm development, multi-threaded task scheduling, and general co

Kitex is a high-performance remote procedure call framework for building distributed microservices in Go. It provides a communication system compatible with gRPC and Protobuf standards and includes a suite of tools for microservices governance. The framework features proxyless service mesh integration, which allows it to manage network traffic and governance directly without adding sidecar proxy latency. It optimizes data throughput through a high-performance communication library that minimizes memory copies. The system covers a broad surface of distributed capabilities, including service d

Hertz is a high-performance Go HTTP framework designed for building scalable microservices, RESTful APIs, and AI applications. It functions as a high-performance web server and a communication framework for microservices, utilizing non-blocking I/O and zero-copy memory management to handle high-concurrency traffic. The project distinguishes itself through a microservices communication toolkit that supports high-efficiency remote procedure calls via gRPC and Thrift protocols. It implements an asynchronous middleware engine based on an onion model, allowing for a pluggable request-response pipe

MessagePack-CSharp is a high-performance binary serialization library for .NET applications that converts object graphs into the MessagePack format. It functions as a C# data serialization toolkit and a polymorphic binary encoder capable of handling abstract classes and interfaces using union keys to identify concrete derived types. The library provides a binary format transcoder to transform binary data into human-readable JSON for debugging. It supports ahead-of-time formatter generation to avoid runtime overhead and implements LZ4 binary compression to reduce the size of serialized data.

Pony is an open-source, actor-model, capabilities-secure, high performance programming language

cuml es una librería y framework de aprendizaje automático acelerado por GPU que utiliza CUDA para acelerar el preprocesamiento de datos tabulares y la ejecución de modelos. Proporciona un conjunto de herramientas para entrenar y desplegar modelos de clasificación, regresión y agrupamiento en GPUs de NVIDIA y clústeres de GPU. La librería está diseñada para la escalabilidad, ofreciendo un entorno de aprendizaje automático de GPU distribuido que puede repartir la computación y los datos a través de múltiples aceleradores de hardware y nodos para manejar conjuntos de datos que exceden la memoria de un solo dispositivo. Refleja las interfaces de estimador estándar para permitir el reemplazo de modelos basados en CPU con versiones aceleradas por GPU dentro de los flujos de trabajo existentes. El proyecto cubre una amplia gama de capacidades de aprendizaje automático, incluyendo aprendizaje supervisado, agrupamiento no supervisado, búsqueda de vecinos más cercanos y reducción de dimensionalidad de alta dimensión. También incluye preprocesamiento de datos tabulares acelerado por hardware para escalado y codificación de características, extracción de características de texto, análisis de series temporales y explicabilidad de predicción de modelos. Las utilidades de soporte incluyen herramientas para la generación de conjuntos de datos sintéticos, serialización del estado del modelo y el cálculo de métricas de rendimiento del modelo.

NCCL es una biblioteca de comunicación de alto rendimiento y un framework de computación distribuida en GPU diseñado para ejecutar intercambios de datos colectivos y punto a punto a través de múltiples GPUs en sistemas de uno o varios nodos. Sirve como capa de transporte RDMA para GPU y orquestador de memoria, facilitando la sincronización de gran ancho de banda de datos y gradientes de modelos para el entrenamiento e inferencia distribuida en GPU. La biblioteca se distingue por su capacidad para ejecutar primitivas de comunicación directamente desde kernels de GPU, eliminando la CPU anfitriona del camino crítico. Utiliza la selección de rutas consciente de la topología para optimizar el movimiento de datos y emplea transporte de red basado en RDMA, incluyendo InfiniBand y NVLink, para permitir el acceso a memoria de copia cero entre dispositivos a través de diferentes nodos físicos. El proyecto cubre una amplia gama de patrones de comunicación colectiva, incluyendo reducciones, broadcasts, gathers e intercambios all-to-all, junto con acceso remoto a memoria punto a punto. Proporciona una gestión integral de comunicadores para inicializar, particionar y redimensionar grupos de GPU, así como una gestión de memoria especializada para registrar buffers y coordinar memoria compartida de dispositivo. El sistema incluye un conjunto de herramientas de monitoreo y observabilidad para el seguimiento de la salud, registro de diagnósticos y monitoreo de eventos en tiempo real, así como interfaces de integración para frameworks de aprendizaje automático, CUDA graphs, MPI y Python.

MemoryPack es una librería de serialización binaria de alto rendimiento para C# y Unity. Proporciona un pipeline de datos de cero asignaciones y un framework de evolución de esquemas diseñado para minimizar las asignaciones de memoria y la sobrecarga de codificación. El proyecto utiliza generadores de código fuente en tiempo de compilación para evitar la reflexión en tiempo de ejecución e implementa un formato binario de codificación cero para un rendimiento máximo. Se distingue por un enfoque de cero asignaciones que reutiliza instancias de objetos para reducir la presión del recolector de basura y copia el diseño de memoria de structs no gestionados directamente a flujos binarios. La librería cubre el versionado de esquemas binarios para compatibilidad hacia atrás y admite streaming de colecciones asíncronas para grandes datasets. Incluye un generador de tipos de TypeScript para sincronizar estructuras de datos entre C# y TypeScript para el intercambio de datos entre lenguajes. Las capacidades adicionales incluyen compresión de carga útil, implementaciones de formateadores personalizados y devoluciones de llamada de serialización.

rkyv es un framework de deserialización de copia cero para Rust que proporciona un formato de serialización binaria para archivos de datos mapeables en memoria. Permite que estructuras de datos complejas se mapeen a bytes y se accedan directamente desde un búfer sin asignar nueva memoria o copiar datos. El proyecto permite la serialización de tipos polimórficos y objetos de rasgo (trait objects), manteniendo su comportamiento dinámico y estructura dentro de la forma binaria. Utiliza direccionamiento de puntero relativo y empaquetado de estructura alineado por bytes para garantizar que los datos sigan siendo válidos independientemente de dónde se carguen en la memoria. El framework cubre la persistencia de datos de alto rendimiento y la gestión de estado de baja latencia a través del mapeo directo de memoria. Proporciona mecanismos tanto para el acceso a datos de copia cero como para la reconstrucción completa del tipo de datos cuando lo requiere la aplicación.