README.adoc
May 18, 2026 · View on GitHub
image::doc/udb.svg[UDB banner]
image::https://api.reuse.software/badge/github.com/riscv/riscv-unified-db[REUSE status, link="https://api.reuse.software/info/github.com/riscv/riscv-unified-db"]
[NOTE]
A new from-scratch documentation site for this project is under construction. It is available at https://riscv.github.io/riscv-unified-db/docs-preview/[riscv.github.io/riscv-unified-db/docs-preview]. Content is incomplete and actively being developed, but is likely still helpful.
[WARNING]
This project is under rapid development. Expect schemas and APIs to change.
The data in the spec directory is also a work in progress and may be incorrect.
That said, the following data has been validated against third-party sources:
- Instruction encodings are validated to match riscv-opcodes
- Instruction names and assembly formats are validated against LLVM.
== Quick links
- https://riscv.github.io/riscv-unified-db/index.html[The latest UDB-generated UNOFFICIAL RISC-V specifications].
- UDB Documentation ** xref:doc/schemas.adoc[Schema documentation] ** xref:doc/idl.adoc[ISA Description Language (IDL)] ** xref:doc/ruby.adoc[Ruby database interface]
- How-to/recipes
** xref:CONTRIBUTING.adoc[Contribution guidelines]
** xref:doc/data-templates.adoc[Data templates]
** xref:doc/riscv-opcodes-migration.adoc[
riscv-opcodesmigration]
== Overview
The goal of the RISC-V Unified Database (UnifiedDB/UDB) is to hold all the information needed to describe RISC-V, including a list of extensions, instruction specifications, CSR specifications, and documentation prose. The vision is that anything one would need for RISC-V can be generated from the information in this repository.
In addition to storing the data, UnifiedDB also provides many tools that use the data to generate artifacts such as specification documents, toolchain inputs, and simulators.
UnifiedDB is an open-source, worldwide project. Contributors do not need to be members of https://riscv.org[RISC-V International (RVI)], though maintainers discuss strategic direction and coordinate with the RISC-V standard through the https://lf-riscv.atlassian.net/wiki/x/iwCsCw?atlOrigin=eyJpIjoiYzU3N2ZiNDViMGRkNGE3ODg0ODVlOWU5YzgzYWM2ODMiLCJwIjoiYyJ9[UnifiedDB Special Interest Group (UDB SIG)] at RVI.
== Repository structure
The following directories contain information relevant to most users of UnifiedDB
spec: Holds all data specification files, along with their schemasspec/std: The RISC-V standardspec/custom: Non-standard RISC-V extensionsspec/schemas: Data schemas for the specification contentcfgs: Configurations used by backends to customize generated outputs
Additionally, developers will be interested in the following:
backends: Tools to generate artifacts (documents, simulators, etc.) from UDB datatools/ruby-gems: Ruby gem libraries for interacting with UDB / IDLbin: Wrapper scripts to run commands in the development environment
== Getting Started
=== Prerequisites
UnifiedDB tooling is designed to run natively on your host using https://mise.jdx.dev[mise].
A container runtime (Docker or Podman) is optional. It is only required for building generated C++ ISS code when a suitable native C++ toolchain and/or RISC-V cross-compiler is unavailable natively.
All core workflows—Ruby, Python, npm, and all other UDB generators—run natively (via mise) without a container.
=== Native Setup (mise)
From the repo root, run:
[source,bash]
bin/setup
That's it. bin/setup installs https://mise.jdx.dev[mise] if needed, uses it to install all
tool versions (Ruby, Python, Node.js, CMake, etc.), installs all dependencies (Ruby gems,
Python packages, npm packages, native binaries), sets up git hooks, and walks you through
choosing a C++ toolchain. It is safe to re-run at any time.
When it finishes, it will tell you how to activate mise in your shell profile so that
mise-managed tools are available in future terminal sessions, and remind you to use the
wrappers in bin/ when invoking any tool.
=== Tasks
[source,bash]
./bin/regress -h # get help on running regression tests
./bin/generate -h # get help on generating content from UDB