generated: '2026-08-30' method: searched source: >- https://rocm.docs.amd.com/llms.txt (AMD's own machine-readable documentation index) and the per-project documentation sites it links. description: >- AMD ships an unusually large first-party command-line surface. These are the control, monitoring, validation and profiling tools in the ROCm Core SDK — they are the practical operator interface to AMD GPUs, and several of them are the human front end for the same gRPC contracts captured in grpc/. Distribution is via the ROCm distro repositories (repo.radeon.com) and OCI containers, not a language package registry; see packages/advanced-micro-devices-packages.yml. install: - method: distribution package url: https://rocm.docs.amd.com/en/latest/install/rocm.html note: apt/dnf/zypper install of the ROCm Core SDK brings the control and monitoring tools. - method: runfile installer url: https://rocm.docs.amd.com/en/latest/install/rocm-runfile-installer.html - method: build from source url: https://rocm.docs.amd.com/en/latest/install/build-from-source.html - method: container url: https://www.amd.com/en/developer/resources/infinity-hub.html note: AMD Infinity Hub publishes pre-built ROCm and framework containers. tools: - name: AMD SMI group: control and monitoring docs: https://rocm.docs.amd.com/projects/amdsmi/en/latest repository: https://github.com/ROCm/amdsmi python_binding: amdsmi description: >- AMD System Management Interface — query and manage AMD GPUs: device enumeration, telemetry, clocks, power, memory and compute partitioning, RAS/bad-page state, resets. The supported successor to rocm-smi. - name: ROCm Data Center Tool (RDC) group: control and monitoring docs: https://rocm.docs.amd.com/projects/rdc/en/latest/ repository: https://github.com/ROCm/rocm-systems version: 1.3.1 description: >- Cluster and data-center administration for AMD GPUs — GPU telemetry, per-job GPU statistics, diagnostics, health, policy and topology. Runs as a daemon exposing the RdcAPI gRPC service; the CLI is the client for that same contract. contract: grpc/advanced-micro-devices-rdc.proto - name: rocminfo group: control and monitoring docs: https://rocm.docs.amd.com/projects/rocminfo/en/latest description: Reports the HSA agents (CPU and GPU) visible to the ROCm runtime. - name: ROCm Compute Profiler (rocprofiler-compute) group: profiling and debugging docs: https://rocm.docs.amd.com/projects/rocprofiler-compute/en/latest description: Kernel-level performance analysis and roofline profiling for GPU compute workloads. - name: ROCm Systems Profiler (rocprofiler-systems) group: profiling and debugging docs: https://rocm.docs.amd.com/projects/rocprofiler-systems/en/latest description: Whole-application, multi-node system tracing and profiling. - name: ROCm Debugger (ROCgdb) group: profiling and debugging docs: https://rocm.docs.amd.com/projects/ROCgdb/en/latest description: GDB-based source-level debugger for GPU kernels. - name: HIPIFY group: porting docs: https://rocm.docs.amd.com/projects/HIPIFY/en/latest description: Translates CUDA source to portable HIP source. - name: ROCm Validation Suite (RVS) group: validation docs: https://rocm.docs.amd.com/projects/ROCmValidationSuite/en/latest/index.html description: Hardware and stack validation tests for deployed AMD GPU systems. - name: TransferBench group: validation docs: https://rocm.docs.amd.com/projects/TransferBench/en/latest/index.html description: Measures host-to-device, device-to-device and multi-GPU transfer bandwidth. tool_count: 9 note: >- Tool names and documentation URLs are transcribed verbatim from AMD's published llms.txt. Exact binary names, subcommand trees and flags were NOT transcribed — reading them reliably requires the per-tool reference pages, and guessing an invocation is worse than recording the gap. evidence: - url: https://rocm.docs.amd.com/llms.txt status: 200 - url: https://rocm.docs.amd.com/projects/amdsmi/en/latest/ status: 200 - url: https://rocm.docs.amd.com/projects/rdc/en/latest/ status: 200