heavydb

HeavyDB, formerly known as MapD Core or OmniSciDB, is an in-memory, column-oriented SQL relational database designed from the ground up to run on GPUs. This repository contains the database engine itself along with build tooling and developer documentation. What the project provides: - A SQL relational database with an in-memory columnar storage model. - GPU-oriented execution as a core design goal, with GPU builds requiring a compatible NVIDIA driver and CUDA toolkit. - Build scripts and dependency tooling targeting Ubuntu 22.04 and Rocky Linux 8.x, installing dependencies under /usr/local/mapd-deps and generating an environment script that must be sourced before configuring the database. - Support for prebuilt dependency archives (static on both platforms, shared additionally on Ubuntu) as well as building dependency archives from source. - Documentation links for installation, configuration, release notes and a dependency quickstart. Development and contribution: - Licensed under Apache License 2.0; third-party packages are covered by separate licenses documented in the repository. - Contribution instructions are provided in CONTRIBUTING.md. - Code style guidance references the C++ Core Guidelines, with clang-format (Chromium-based style) and clang-tidy configurations at the repository root; a make target runs clang-tidy. - Source files are expected to carry SPDX license headers, managed by a pre-commit hook and enforced in CI for added or modified files; vendored third-party code is exempt. - Building from source is documented as producing the broader HeavyAI platform, including Immerse, WebServer, GEOS DSOs and HeavyDB, into a usable Docker image. Security and support: - Security vulnerabilities should not be reported through public GitHub issues; private reporting channels are described in SECURITY.md. - Community questions and feedback are directed to HeavyAI GitHub Discussions, with maintainers reviewing issues and pull requests on a best-effort basis without guaranteed response timelines. This overview reflects only what the repository README states; it makes no claims about performance, benchmarks or comparisons with other systems.