ROCm/TheRock

The HIP Environment and ROCm Kit - A lightweight open source build system for HIP and ROCm

What it solves

TheRock provides a unified, modular build and release system for HIP and ROCm, replacing a previously monolithic release process. It simplifies the integration, building, and distribution of ROCm components, making it easier for contributors and advanced users to avoid the complexity of traditional package-based installations.

How it works

It operates as a CMake super-project that integrates various ROCm components into a single buildable and testable product. The system supports multiple Linux distributions and native Windows, providing tools for developing individual components and comprehensive CI/CD pipelines for building and releasing artifacts, including nightly releases of ROCm, PyTorch, and JAX.

Who it’s for

It is designed for ROCm contributors, developers, researchers, and advanced users who need the latest ROCm capabilities and the flexibility to build from source.

Highlights

  • Modular Workflow: Replaces monolithic releases with a modular system for easier component integration.
  • C-Suite Support: Includes support for building PyTorch and JAX with ROCm from source.
  • Cross-Platform: Supports multiple Linux distributions and native Windows.
  • Build Optimization: Provides scripts for ccache integration to speed up frequent rebuilds.
  • Comprehensive Tooling: Includes a super-project for source builds and tools for developing individual ROCm components.

Related

  • Project
    ROCm/legacy-rocm-build

    ROCm / legacy‑rocm‑build is the (now‑deprecated) repository that assembles AMD’s open‑source GPU compute stack—drivers, HIP runtime, math libraries, profiling tools, and more—so developers can run AI/ML workloads on AMD GPUs. It points users to install guides and to the newer TheRock monorepo.

  • Project
    ROCm/rocm-libraries

    A consolidated super-repository for ROCm libraries that streamlines the development, build, and integration of high-performance math and AI libraries for AMD GPUs.

  • Project
    ROCm/rocm-systems

    ROCm Systems is a monorepo that unifies many ROCm low‑level GPU libraries (HIP, profiling, tracing, system‑management, etc.) to simplify building, testing, and contributing to the AMD GPU stack that underlies AI/ML frameworks.

  • Dispatch
    Easily Build and Share ROCm Kernels with Hugging Face

    Hugging Face introduces a streamlined workflow using the kernels library and kernel-builder to build, package, and share ROCm-compatible GPU kernels on the Hugging Face Hub.

  • Project
    alibaba/ROCK

    A scalable sandbox environment management framework designed for agentic reinforcement learning, providing tools for building, deploying, and scheduling isolated runtimes.