TheRock

Health v2 · maintenance only

Adoption not part of the score

1394 stars · 351 forks observed · 2026-10-21

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

TheRock is AMD's open-source build and release system for the ROCm software stack, replacing the legacy monolithic ROCm release process with a modular workflow. It produces the ROCm Core SDK — drivers, HIP runtime, compilers, math libraries, and profiling tools — for GPU computing on AMD Instinct, Radeon, and Ryzen AI hardware.

Use cases

• build the AMD ROCm GPU software stack from source
• install ROCm for PyTorch or JAX on AMD GPUs
• set up HIP GPU programming on Linux or Windows
• get nightly ROCm builds for unsupported or community-enabled GPUs
• deploy AI inference workloads on AMD Instinct accelerators
• package ROCm components as tarballs or DEB/RPM packages
• migrate from the legacy ROCm release stream to the new build system

When to choose

• you develop on or for AMD GPUs and need the ROCm SDK, HIP runtime, or math libraries
• you want to build ROCm components modularly or contribute to its build system
• your GPU is only supported via TheRock nightly/community builds
• you need reproducible, per-component ROCm distributions for HPC or ML environments

When to avoid

• you develop exclusively for NVIDIA GPUs (use CUDA instead)
• you need the legacy monolithic ROCm release process, which is deprecated
• you want a turnkey ML framework without touching GPU toolchains
• your hardware or OS is outside AMD's supported compatibility matrix and no community build exists

Facets

framework · maturity active

build-tool compiler gpu-computing developer-tools deployment gpu-computing machine-learning deep-learning developer-tools compilers windows python cpp amd rocm hip therock build-and-release-system open-source-gpu-stack pytorch hpc linux gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="ROCm/TheRock")