Texte vergleichen

Finde den Unterschied zwischen zwei Textdateien

Live-Editor

Leerzeichen ausblenden

Gleiches ausblenden

Zeilenumbruch aus

Ansicht

Vergleichsgenauigkeit

Textstile

Syntaxhervorhebung

Diffchecker Desktop Der sicherste Weg, Diffchecker zu nutzen. Hol dir die Desktop-App: Deine Diffs verlassen nie deinen Computer!Desktop holen

diff llms-full.txt

Erstellt vor 2 WochenDiff läuft nie ab

Die beiden Texte sind identisch

Es gibt keinen Unterschied zwischen diesen beiden Texten

0 Entfernungen

Zeilen
Gesamt
Entfernt

Zeichen
Gesamt
Entfernt

Um diese Funktion weiterhin zu nutzen, aktualisiere auf Diffchecker Pro Preise anzeigen

288 Zeilen

0 Hinzufügungen

Zeilen
Gesamt
Hinzugefügt

Zeichen
Gesamt
Hinzugefügt

Um diese Funktion weiterhin zu nutzen, aktualisiere auf Diffchecker Pro Preise anzeigen

288 Zeilen

# ROCm documentation

> Note: ROCm documentation is split across multiple projects. In addition to this file, each project publishes its own `llms.txt` and `llms-full.txt` under `https://<base_url>/projects/<project_name>/en/latest/`.

## Docs

- [Release notes](https://rocm.docs.amd.com/en/latest/about/release-notes.html)

- [Transition guide](https://rocm.docs.amd.com/en/latest/about/transition-guide-TheRock.html)

- [What is ROCm?](https://rocm.docs.amd.com/en/latest/about/what-is-rocm.html): Learn what ROCm is – AMD open software stack for GPU programming, including runtimes, compilers, libraries, and tools for Linux and Windows.

- [Compatibility matrix](https://rocm.docs.amd.com/en/latest/compatibility/compatibility-matrix.html)

- [Install ROCm](https://rocm.docs.amd.com/en/latest/install/rocm.html): How to install AMD ROCm for Instinct GPUs, Radeon GPUs, and Ryzen AI APUs

- [Prerequisites](https://rocm.docs.amd.com/en/latest/install/redirect/_prerequisites.html)

- [Installation](https://rocm.docs.amd.com/en/latest/install/redirect/_install.html)

- [Post-installation](https://rocm.docs.amd.com/en/latest/install/redirect/_post-install.html)

- [Uninstalling](https://rocm.docs.amd.com/en/latest/install/redirect/_uninstall.html)

- [ROCm Runfile Installer](https://rocm.docs.amd.com/en/latest/install/rocm-runfile-installer.html): How to use the ROCm Runfile Installer

- [Build ROCm from source](https://rocm.docs.amd.com/en/latest/install/build-from-source.html): Learn how to build the ROCm Core SDK from source using TheRock. Includes references to environment setup guides for Ubuntu 24.04 and Windows 11, plus links to official instructions and compatibility guidance.

- [ROCm Core SDK](https://rocm.docs.amd.com/en/latest/components/core.html): AMD ROCm Core SDK - list of libraries and tools

- [Math and compute libraries](https://rocm.docs.amd.com/en/latest/components/math-and-compute-libs.html): AMD ROCm math and compute libraries for GPU-accelerated linear algebra, FFTs, random number generation, and deep learning.

- [Composable Kernel](https://rocm.docs.amd.com/projects/composable_kernel/en/latest)

- [hipBLAS](https://rocm.docs.amd.com/projects/hipBLAS/en/latest)

- [hipBLASLt](https://rocm.docs.amd.com/projects/hipBLASLt/en/latest)

- [hipCUB](https://rocm.docs.amd.com/projects/hipCUB/en/latest)

- [hipFFT](https://rocm.docs.amd.com/projects/hipFFT/en/latest)

- [hipRAND](https://rocm.docs.amd.com/projects/hipRAND/en/latest)

- [hipSOLVER](https://rocm.docs.amd.com/projects/hipSOLVER/en/latest)

- [hipSPARSE](https://rocm.docs.amd.com/projects/hipSPARSE/en/latest)

- [hipSPARSELt](https://rocm.docs.amd.com/projects/hipSPARSELt/en/latest)

- [MIOpen](https://rocm.docs.amd.com/projects/MIOpen/en/latest)

- [rocBLAS](https://rocm.docs.amd.com/projects/rocBLAS/en/latest)

- [rocFFT](https://rocm.docs.amd.com/projects/rocFFT/en/latest)

- [rocPRIM](https://rocm.docs.amd.com/projects/rocPRIM/en/latest)

- [rocRAND](https://rocm.docs.amd.com/projects/rocRAND/en/latest)

- [rocSOLVER](https://rocm.docs.amd.com/projects/rocSOLVER/en/latest)

- [rocSPARSE](https://rocm.docs.amd.com/projects/rocSPARSE/en/latest)

- [rocThrust](https://rocm.docs.amd.com/projects/rocThrust/en/latest)

- [rocWMMA](https://rocm.docs.amd.com/projects/rocWMMA/en/latest)

- [Communication libraries](https://rocm.docs.amd.com/en/latest/components/communication-libs.html): AMD ROCm communication libraries for multi-GPU and multi-node collective and peer-to-peer communication.

- [RCCL](https://rocm.docs.amd.com/projects/rccl/en/latest)

- [rocSHMEM](https://rocm.docs.amd.com/projects/rocSHMEM/en/latest)

- [Media libraries](https://rocm.docs.amd.com/en/latest/components/media-libs.html): AMD ROCm media libraries for GPU-accelerated video decoding and image processing.

- [rocDecode](https://rocm.docs.amd.com/projects/rocDecode/en/latest)

- [rocJPEG](https://rocm.docs.amd.com/projects/rocJPEG/en/latest)

- [Storage libraries](https://rocm.docs.amd.com/en/latest/components/storage-libs.html): AMD ROCm storage libraries

- [hipFile](https://rocm.docs.amd.com/projects/hipFile/en/latest)

- [Runtime and compilers](https://rocm.docs.amd.com/en/latest/components/runtimes-and-compilers.html): AMD ROCm runtimes and compilers for GPU application development, including HIP, HIPIFY, and LLVM.

- [HIP](https://rocm.docs.amd.com/projects/HIP/en/latest)

- [HIPIFY](https://rocm.docs.amd.com/projects/HIPIFY/en/latest)

- [LLVM](https://rocm.docs.amd.com/projects/llvm-project/en/latest)

- [Profiling and debugging tools](https://rocm.docs.amd.com/en/latest/components/profilers-and-debuggers.html): AMD ROCm profiling and debugging tools for GPU application performance analysis and fault diagnosis.

- [ROCm Compute Profiler (rocprofiler-compute)](https://rocm.docs.amd.com/projects/rocprofiler-compute/en/latest)

- [ROCm Systems Profiler (rocprofiler-systems)](https://rocm.docs.amd.com/projects/rocprofiler-systems/en/latest)

- [ROCprofiler-SDK](https://rocm.docs.amd.com/projects/rocprofiler-sdk/en/latest)

- [ROCdbgapi](https://rocm.docs.amd.com/projects/ROCdbgapi/en/latest)

- [ROCm Debugger (ROCgdb)](https://rocm.docs.amd.com/projects/ROCgdb/en/latest)

- [ROCR Debug Agent](https://rocm.docs.amd.com/projects/rocr_debug_agent/en/latest)

- [Control and monitoring tools](https://rocm.docs.amd.com/en/latest/components/control-and-monitoring-tools.html): AMD ROCm control and monitoring tools for inspecting and managing AMD GPU hardware state.

- [AMD SMI](https://rocm.docs.amd.com/projects/amdsmi/en/latest)

- [ROCm Data Center Tool (RDC)](https://rocm.docs.amd.com/projects/rdc/en/latest)

- [rocminfo](https://rocm.docs.amd.com/projects/rocminfo/en/latest)

- [ROCm HPC SDK](https://rocm.docs.amd.com/en/latest/components/hpc-sdk/index.html): Learn how to use AMD ROCm for high-performance computing (HPC).

- [Install HPC SDK](https://rocm.docs.amd.com/en/latest/components/hpc-sdk/install.html): Install ROCm to run high-performance computing (HPC) workloads.

- [ROCm Extras](https://rocm.docs.amd.com/en/latest/components/extras.html): AMD ROCm Extras - list of tools and test suites

- [ROCm Validation Suite (RVS)](https://rocm.docs.amd.com/projects/ROCmValidationSuite/en/docs-1.5.122/index.html)

- [TransferBench](https://rocm.docs.amd.com/projects/TransferBench/en/docs-1.66.02/index.html)

- [AI ecosystem](https://rocm.docs.amd.com/en/latest/ai-ecosystem.html)

- [PyTorch](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/frameworks/pytorch/install.html)

- [JAX](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/frameworks/jax/install.html)

- [vLLM](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/inference/vllm.html)

- [SGLang](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/inference/sglang.html)

- [AI Developer Hub](https://rocm.docs.amd.com/projects/ai-developer-hub/en/latest/)

- [ROCm examples](https://github.com/ROCm/rocm-examples/tree/release/therock-7.14)

- [AMD GPU specifications](https://rocm.docs.amd.com/en/latest/reference/gpu-specs.html): AMD Instinct™ GPU, AMD Radeon PRO™, and AMD Radeon™ GPU architecture information

- [AMD GPU architectures](https://rocm.docs.amd.com/en/latest/reference/gpu-arch/index.html)

- [MI350 microarchitecture](https://rocm.docs.amd.com/en/latest/reference/gpu-arch/mi350.html): AMD Instinct MI350 Series microarchitecture reference.

- [CDNA4 ISA reference](https://www.amd.com/content/dam/amd/en/documents/instinct-tech-docs/instruction-set-architectures/amd-instinct-cdna4-instruction-set-architecture.pdf)

- [CDNA4 white paper](https://www.amd.com/content/dam/amd/en/documents/instinct-tech-docs/white-papers/amd-cdna-4-architecture-whitepaper.pdf)

- [Instinct MI350 Series performance counters](https://rocm.docs.amd.com/en/latest/reference/gpu-arch/mi350-performance-counters.html): MI355 Series performance counters and metrics

- [MI300 microarchitecture](https://rocm.docs.amd.com/en/latest/reference/gpu-arch/mi300.html): Learn about the AMD Instinct MI300 Series architecture.

- [CDNA3 ISA reference](https://www.amd.com/content/dam/amd/en/documents/instinct-tech-docs/instruction-set-architectures/amd-instinct-mi300-cdna3-instruction-set-architecture.pdf)

- [CDNA3 white paper](https://www.amd.com/content/dam/amd/en/documents/instinct-tech-docs/white-papers/amd-cdna-3-white-paper.pdf)

- [Instinct MI300 and MI200 Series performance counters](https://rocm.docs.amd.com/en/latest/reference/gpu-arch/mi300-mi200-performance-counters.html): MI300 and MI200 Series performance counters and metrics

- [MI250 microarchitecture](https://rocm.docs.amd.com/en/latest/reference/gpu-arch/mi250.html): Learn about the AMD Instinct MI250 Series architecture.

- [CDNA2 ISA reference](https://www.amd.com/content/dam/amd/en/documents/instinct-tech-docs/instruction-set-architectures/instinct-mi200-cdna2-instruction-set-architecture.pdf)

- [CDNA2 white paper](https://www.amd.com/content/dam/amd/en/documents/instinct-business-docs/white-papers/amd-cdna2-white-paper.pdf)

- [MI100 microarchitecture](https://rocm.docs.amd.com/en/latest/reference/gpu-arch/mi100.html): Learn about the AMD Instinct MI100 Series architecture.

- [CDNA1 ISA reference](https://www.amd.com/content/dam/amd/en/documents/instinct-tech-docs/instruction-set-architectures/instinct-mi100-cdna1-shader-instruction-set-architecture.pdf)

- [CDNA1 white paper](https://www.amd.com/content/dam/amd/en/documents/instinct-business-docs/white-papers/amd-cdna-white-paper.pdf)

- [AMD GPU system optimization](https://rocm.docs.amd.com/en/latest/reference/system-optimization/index.html): Learn about AMD hardware optimization for HPC-specific and workstation workloads.

- [AMD Instinct GPUs](https://rocm.docs.amd.com/en/latest/reference/system-optimization/cdna.html): System optimization guides for AMD Instinct CDNA GPUs.

- [AMD MI355X](https://instinct.docs.amd.com/projects/system-acceptance/en/latest/gpus/mi355x.html)

- [AMD MI350X](https://instinct.docs.amd.com/projects/system-acceptance/en/latest/gpus/mi350x.html)

- [AMD MI325X](https://instinct.docs.amd.com/projects/system-acceptance/en/latest/gpus/mi325x.html)

- [AMD MI300X](https://instinct.docs.amd.com/projects/system-acceptance/en/latest/gpus/mi300x.html)

- [AMD MI300A](https://instinct.docs.amd.com/projects/system-acceptance/en/latest/gpus/mi300a.html)

- [AMD MI250 / MI250X](https://instinct.docs.amd.com/projects/system-acceptance/en/latest/gpus/mi250.html)

- [AMD MI210](https://instinct.docs.amd.com/projects/system-acceptance/en/latest/gpus/mi210.html)

- [AMD MI100](https://instinct.docs.amd.com/projects/system-acceptance/en/latest/gpus/mi100.html)

- [AMD Radeon and Ryzen GPUs](https://rocm.docs.amd.com/en/latest/reference/system-optimization/rdna.html): System optimization guides for AMD Radeon and Ryzen RDNA GPUs.

- [RDNA3.5](https://rocm.docs.amd.com/en/latest/reference/system-optimization/rdna3-5.html): System optimization of AMD RDNA3.5 Ryzen APUs (gfx1150/gfx1151/gfx1152) systems. Learn about VRAM, GTT, TTM tuning, shared memory configuration, and required Linux kernel support.

- [RDNA2](https://rocm.docs.amd.com/en/latest/reference/system-optimization/rdna2.html): Learn about system settings and performance tuning for RDNA2-based GPUs.

- [Common system settings](https://rocm.docs.amd.com/en/latest/reference/system-optimization/common.html): Common system settings for AMD GPUs including GPU isolation and BAR access configuration.

- [GPU isolation techniques](https://rocm.docs.amd.com/en/latest/reference/system-optimization/gpu-isolation.html)

- [BAR access limits](https://rocm.docs.amd.com/en/latest/reference/system-optimization/bar-access-limits.html): Learn about BAR configuration in AMD GPUs and ways to troubleshoot physical addressing limit

- [AMD GPU and ROCm components data types and precision support](https://rocm.docs.amd.com/en/latest/reference/precision-support.html): Supported data types of AMD GPUs and libraries in ROCm.

- [ROCm libraries graph-safe support](https://rocm.docs.amd.com/en/latest/reference/graph-safe-support.html): This page lists supported graph safe ROCm libraries.

- [AMD GPU atomics operation support](https://rocm.docs.amd.com/en/latest/reference/gpu-atomics-operation.html): AMD Instinct GPU, AMD Radeon PRO, and AMD Radeon GPU atomics operations information

- [ROCm environment variables](https://rocm.docs.amd.com/en/latest/reference/environment-variables/index.html): Environment variables reference

- [Set the number of CUs](https://rocm.docs.amd.com/en/latest/reference/environment-variables/setting-cus.html): Setting the number of CUs

- [AMD GPU programming on ROCm](https://rocm.docs.amd.com/en/latest/reference/hip-programming.html): HIP programming

- [Contributing to the ROCm documentation](https://rocm.docs.amd.com/en/latest/contribute/contributing.html)

- [ROCm documentation toolchain](https://rocm.docs.amd.com/en/latest/contribute/toolchain.html)

- [Building documentation](https://rocm.docs.amd.com/en/latest/contribute/building.html)

- [Providing feedback about the ROCm documentation](https://rocm.docs.amd.com/en/latest/contribute/feedback.html)

- [ROCm licenses](https://rocm.docs.amd.com/en/latest/about/license.html)

---

Source: https://rocm.docs.amd.com/en/latest/index.html

# AMD ROCm

ROCm is AMD’s open-source GPU computing platform: an end‑to‑end ecosystem of

compilers, runtimes, and libraries for AI, HPC, and domain‑specific workloads.

It is open source, cross‑platform (Linux and Windows), and optimized for AMD

Instinct™, AMD Radeon™, and AMD Ryzen™ AI devices.

The foundational libraries, runtimes, and tools for GPU computing on AMD

hardware — math and compute libraries, communication primitives, HIP runtime,

profiling and debugging tools, and more.

- [Install the ROCm Core SDK](https://rocm.docs.amd.com/en/latest/install/rocm.html)

- [ROCm 7.14.0 compatibility](https://rocm.docs.amd.com/en/latest/compatibility/compatibility-matrix.html)

[Go to Core SDK docs](https://rocm.docs.amd.com/en/latest/about/release-notes.html)

Full-stack documentation and recipes to deploy AI workloads on AMD

GPUs using popular ROCm-enabled frameworks.

- Deep learning frameworks

- [PyTorch](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/frameworks/pytorch/install.html)

- [JAX](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/frameworks/jax/install.html)

- Inference

- [vLLM](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/inference/vllm.html)

- [SGLang](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/inference/sglang.html)

[Go to AI Ecosystem docs](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/index.html)

Deployment and operations guidance for AMD Instinct GPUs at scale, including

AMD GPU Driver installation, cluster management, GPU partitioning, monitoring,

virtualization, cloud deployments, and containers.

- [AMD GPU Driver (amdgpu)](https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/)

- [Network Operator (Kubernetes)](https://instinct.docs.amd.com/projects/network-operator/en/latest/overview.html)

[Go to GPU Systems docs](https://instinct.docs.amd.com)

Open-source collections of ROCm-accelerated libraries for building

high-performance domain-specific applications.

- [ROCm Data Science](https://rocm.docs.amd.com/projects/rocm-ds/en/latest/index.html)

- [ROCm Finance](https://rocm.docs.amd.com/projects/rocm-finance/en/latest/index.html)

- [ROCm Life Science](https://rocm.docs.amd.com/projects/rocm-ls/en/latest/index.html)

- [ROCm LLMExt](https://rocm.docs.amd.com/projects/rocm-llmext/en/latest/index.html)

- [ROCm Simulation](https://rocm.docs.amd.com/projects/rocm-simulation/en/latest/index.html)

---

Source: https://rocm.docs.amd.com/en/latest/about/release-notes.html

# ROCm Core SDK 7.14.0 release notes

ROCm Core SDK 7.14.0 transitions ROCm to [TheRock](https://github.com/ROCm/TheRock), a build and release system that introduces a modular architecture to improve flexibility, maintainability, and alignment with community use cases:

* **Leaner core**: The Core SDK focuses on essential runtime and development components.

* **Use case-specific expansions**: Optional domain-specific SDKs for AI, data science, and HPC.

* **Modular installation**: Install only the components required for your workflow.

This approach streamlines installation, reduces footprint, and accelerates innovation through independently released packages. To learn more, see the [transition guide](https://rocm.docs.amd.com/en/latest/about/transition-guide-TheRock.html).

#### NOTE

ROCm 7.14.0 follows the [versioning discontinuity that began with the 7.9.0 preview](https://rocm.docs.amd.com/en/7.9.0-preview/about/release-notes.html#preview-stream-note) release.

## Release highlights

This release focuses on AI inference, distributed workloads, and profiling across AMD Instinct™, Radeon™, and Ryzen™ AI platforms. Highlights include inference-ready vLLM images and packages, ROCprofiler-SDK adoption across AI profiling workflows, expanded system telemetry and validation coverage, and updates to math, sparse, and communication libraries.

### Platform and hardware support

This release expands GPU, operating system, virtualization, and partitioning support.

#### Expanded AMD GPU support

ROCm 7.14.0 adds support for the following AMD APUs:

* [AMD Ryzen AI Max+ PRO 495 (gfx1151)](https://www.amd.com/en/products/processors/laptop/ryzen-pro/ai-max-pro-400-series/amd-ryzen-ai-max-plus-pro-495.html)

* [AMD Ryzen AI Max PRO 490 (gfx1151)](https://www.amd.com/en/products/processors/laptop/ryzen-pro/ai-max-pro-400-series/amd-ryzen-ai-max-pro-490.html)

* [AMD Ryzen AI Max PRO 485 (gfx1151)](https://www.amd.com/en/products/processors/laptop/ryzen-pro/ai-max-pro-400-series/amd-ryzen-ai-max-pro-485.html)

* [AMD Ryzen AI 5 435 (gfx1153)](https://www.amd.com/en/products/processors/laptop/ryzen/ai-400-series/amd-ryzen-ai-5-435.html)

* [AMD Ryzen AI 5 430 (gfx1153)](https://www.amd.com/en/products/processors/laptop/ryzen/ai-400-series/amd-ryzen-ai-5-430.html)

* [AMD Ryzen AI 5 PRO 435 (gfx1153)](https://www.amd.com/en/products/processors/laptop/ryzen-pro/ai-400-series/amd-ryzen-ai-5-pro-435.html)

* [AMD Ryzen AI 7 445 (gfx1153)](https://www.amd.com/en/products/processors/laptop/ryzen/ai-400-series/amd-ryzen-ai-7-445.html)

For the complete list of supported AMD hardware, see [AMD hardware support]().

#### Expanded operating system support

ROCm 7.14.0 adds support for RHEL 10.2 and RHEL 9.8 on AMD Instinct and Radeon GPUs. RHEL 10.2 replaces RHEL 10.1 as the validated RHEL 10 release; RHEL 9.8 replaces RHEL 9.7 as the validated RHEL 9 release.

SUSE Linux Enterprise Server (SLES) 15 SP7, SLES 16, and Debian 13 are now supported on AMD Instinct MI350P.

For the full list of supported Linux distributions, see [Operating system support]().

#### Expanded GPU virtualization support for Instinct and Radeon GPUs

ROCm 7.14.0 adds support for the following virtualization configurations on AMD Instinct and Radeon GPUs:

* On MI350X: VMware ESXi 9.1 with Ubuntu 24.04 guest OS.

* On Radeon AI PRO R9700S: KVM Passthrough with Ubuntu 24.04 host OS and Ubuntu 24.04 guest OS.

* On Radeon PRO V710: KVM SR-IOV with Ubuntu 24.04 host OS and Ubuntu 24.04 guest OS.

Supported Single Root I/O Virtualization (SR-IOV) configurations require the [AMD GPU Virtualization Driver (GIM) 9.1.0.K](https://github.com/amd/MxGPU-Virtualization/releases/tag/9.1.0.K). For details, see [GPU virtualization support]().

#### Expanded Instinct GPU partitioning support

ROCm 7.14.0 has enabled and optimized multi-VF partition modes for the following GPU partitioning configurations in SR-IOV deployments:

On MI355X and MI350X:

DPX compute partition mode with NPS2 memory partitioning.

CPX compute partition mode with NPS2 memory partitioning.

For details, see [GPU partitioning support]().

### AI inference and frameworks

This release enables support for the following frameworks:

* PyTorch 2.12.0

* JAX 0.10.0

* vLLM 0.23.0 [\*]

* SGLang 0.5.13

* TensorFlow 2.21

The updated framework support replaces the previous PyTorch 2.9.1, JAX 0.8.2, vLLM 0.19.1, and SGLang 0.5.9 support.

For details, see [AI ecosystem support]().

### Developer tools, profiling, and validation

This release improves ROCm developer workflows with new HIP APIs, expanded profiling and tracing capabilities, and broader telemetry coverage.

#### HIP feature highlights

The following are notable enhancements to HIP:

* **HIP API additions for CUDA parity**:

* **Batch memory management**: New batch asynchronous memory management APIs let applications discard (`hipMemDiscardBatchAsync`), prefetch (`hipMemPrefetchBatchAsync`), or combine both operations (`hipMemDiscardAndPrefetchBatchAsync`) across multiple memory ranges in a single call, reducing API call overhead. Both HIP runtime and HIP driver variants are available.

* **Library management**: New library management APIs return the device pointer and size of a device global (`hipLibraryGetGlobal`) and the host pointer and size of a managed variable (`hipLibraryGetManaged`) defined in a `hipLibrary_t`, improving parity with CUDA library APIs.

* **Faster HIP graph replay for asynchronous memory allocations**: HIP graph replay now reduces overhead for graphs that interleave asynchronous memory allocations with compute. Allocation nodes no longer block during replay. Physical memory is reused across nodes instead of being mapped and unmapped on each launch, eliminating the gaps between kernels this pattern previously caused. For background on HIP graphs, see [Graph Management](https://rocm.docs.amd.com/projects/HIP/en/latest/reference/hip_runtime_api/modules/graph_management.html).

* **New HIP documentation topic on device properties**: The [HIP Device Properties and Topology on CDNA Architectures](https://rocm.docs.amd.com/projects/HIP/en/latest/how-to/hipDeviceProperties.html) provides an overview of the CDNA architecture topology and its key hardware characteristics, helping developers better understand the underlying architecture and optimize the performance of their HIP applications.

For more information, see the [HIP section]() in the ROCm component changelogs.

#### ROCprofiler-SDK feature highlights

The following are notable enhancements to ROCprofiler-SDK:

##### ROCprofiler-SDK integration with PyTorch Profiler

Starting with PyTorch 2.12, `rocprofiler-sdk` is used as the ROCm profiling backend for PyTorch Profiler on supported ROCm configurations, replacing the legacy `roctracer`-based profiling path. This enables PyTorch users to collect GPU activity traces through the `rocprofiler-sdk` infrastructure and provides a stronger foundation for correctness, stability, and future profiling capabilities. The integration also positions PyTorch Profiler to benefit from additional `rocprofiler-sdk` capabilities as framework-level support continues to evolve.

##### ROCprofiler-SDK beta support for Streaming Performance Monitors

`rocprofiler-sdk` and `rocprofv3` add beta support for Streaming Performance Monitors (SPM), enabling selected hardware counters to be sampled over time while workloads execute. Unlike traditional counter collection, which captures a single aggregated value per kernel dispatch, SPM provides time-resolved hardware counter data. This is useful for analyzing long-running workloads and training jobs where temporal behavior matters as much as aggregate metrics. ROCpd support is planned for a future release.

In ROCm 7.14.0, SPM support is available through the `rocprofiler-sdk` API and `rocprofv3`. To enable SPM in `rocprofv3`, use the `--spm-beta-enabled` flag or set the `ROCPROFILER_SPM_BETA_ENABLED` environment variable. For API-based usage, set `ROCPROFILER_SPM_BETA_ENABLED`.

Supported hardware: AMD Instinct MI300X, MI325X, MI350X, and MI355X GPUs.

For more information, see the [SPM API reference guide](https://rocm.docs.amd.com/projects/rocprofiler-sdk/en/latest/api-reference/spm.html) and the [SPM usage guide](https://rocm.docs.amd.com/projects/rocprofiler-sdk/en/latest/how-to/using-spm.html) for `rocprofv3`.

#### WARNING

SPM is a beta capab

Gespeicherte Diffs

Originaltext

Datei öffnen

# ROCm documentation

## Docs

- [Release notes](https://rocm.docs.amd.com/en/latest/about/release-notes.html)
- [Transition guide](https://rocm.docs.amd.com/en/latest/about/transition-guide-TheRock.html)
- [What is ROCm?](https://rocm.docs.amd.com/en/latest/about/what-is-rocm.html): Learn what ROCm is – AMD open software stack for GPU programming, including runtimes, compilers, libraries, and tools for Linux and Windows.
- [Compatibility matrix](https://rocm.docs.amd.com/en/latest/compatibility/compatibility-matrix.html)
- [Install ROCm](https://rocm.docs.amd.com/en/latest/install/rocm.html): How to install AMD ROCm for Instinct GPUs, Radeon GPUs, and Ryzen AI APUs
  - [Prerequisites](https://rocm.docs.amd.com/en/latest/install/redirect/_prerequisites.html)
  - [Installation](https://rocm.docs.amd.com/en/latest/install/redirect/_install.html)
  - [Post-installation](https://rocm.docs.amd.com/en/latest/install/redirect/_post-install.html)
  - [Uninstalling](https://rocm.docs.amd.com/en/latest/install/redirect/_uninstall.html)
- [ROCm Runfile Installer](https://rocm.docs.amd.com/en/latest/install/rocm-runfile-installer.html): How to use the ROCm Runfile Installer
- [Build ROCm from source](https://rocm.docs.amd.com/en/latest/install/build-from-source.html): Learn how to build the ROCm Core SDK from source using TheRock. Includes references to environment setup guides for Ubuntu 24.04 and Windows 11, plus links to official instructions and compatibility guidance.
- [ROCm Core SDK](https://rocm.docs.amd.com/en/latest/components/core.html): AMD ROCm Core SDK - list of libraries and tools
  - [Math and compute libraries](https://rocm.docs.amd.com/en/latest/components/math-and-compute-libs.html): AMD ROCm math and compute libraries for GPU-accelerated linear algebra, FFTs, random number generation, and deep learning.
    - [Composable Kernel](https://rocm.docs.amd.com/projects/composable_kernel/en/latest)
    - [hipBLAS](https://rocm.docs.amd.com/projects/hipBLAS/en/latest)
    - [hipBLASLt](https://rocm.docs.amd.com/projects/hipBLASLt/en/latest)
    - [hipCUB](https://rocm.docs.amd.com/projects/hipCUB/en/latest)
    - [hipFFT](https://rocm.docs.amd.com/projects/hipFFT/en/latest)
    - [hipRAND](https://rocm.docs.amd.com/projects/hipRAND/en/latest)
    - [hipSOLVER](https://rocm.docs.amd.com/projects/hipSOLVER/en/latest)
    - [hipSPARSE](https://rocm.docs.amd.com/projects/hipSPARSE/en/latest)
    - [hipSPARSELt](https://rocm.docs.amd.com/projects/hipSPARSELt/en/latest)
    - [MIOpen](https://rocm.docs.amd.com/projects/MIOpen/en/latest)
    - [rocBLAS](https://rocm.docs.amd.com/projects/rocBLAS/en/latest)
    - [rocFFT](https://rocm.docs.amd.com/projects/rocFFT/en/latest)
    - [rocPRIM](https://rocm.docs.amd.com/projects/rocPRIM/en/latest)
    - [rocRAND](https://rocm.docs.amd.com/projects/rocRAND/en/latest)
    - [rocSOLVER](https://rocm.docs.amd.com/projects/rocSOLVER/en/latest)
    - [rocSPARSE](https://rocm.docs.amd.com/projects/rocSPARSE/en/latest)
    - [rocThrust](https://rocm.docs.amd.com/projects/rocThrust/en/latest)
    - [rocWMMA](https://rocm.docs.amd.com/projects/rocWMMA/en/latest)
  - [Communication libraries](https://rocm.docs.amd.com/en/latest/components/communication-libs.html): AMD ROCm communication libraries for multi-GPU and multi-node collective and peer-to-peer communication.
    - [RCCL](https://rocm.docs.amd.com/projects/rccl/en/latest)
    - [rocSHMEM](https://rocm.docs.amd.com/projects/rocSHMEM/en/latest)
  - [Media libraries](https://rocm.docs.amd.com/en/latest/components/media-libs.html): AMD ROCm media libraries for GPU-accelerated video decoding and image processing.
    - [rocDecode](https://rocm.docs.amd.com/projects/rocDecode/en/latest)
    - [rocJPEG](https://rocm.docs.amd.com/projects/rocJPEG/en/latest)
  - [Storage libraries](https://rocm.docs.amd.com/en/latest/components/storage-libs.html): AMD ROCm storage libraries
    - [hipFile](https://rocm.docs.amd.com/projects/hipFile/en/latest)
  - [Runtime and compilers](https://rocm.docs.amd.com/en/latest/components/runtimes-and-compilers.html): AMD ROCm runtimes and compilers for GPU application development, including HIP, HIPIFY, and LLVM.
    - [HIP](https://rocm.docs.amd.com/projects/HIP/en/latest)
    - [HIPIFY](https://rocm.docs.amd.com/projects/HIPIFY/en/latest)
    - [LLVM](https://rocm.docs.amd.com/projects/llvm-project/en/latest)
  - [Profiling and debugging tools](https://rocm.docs.amd.com/en/latest/components/profilers-and-debuggers.html): AMD ROCm profiling and debugging tools for GPU application performance analysis and fault diagnosis.
    - [ROCm Compute Profiler (rocprofiler-compute)](https://rocm.docs.amd.com/projects/rocprofiler-compute/en/latest)
    - [ROCm Systems Profiler (rocprofiler-systems)](https://rocm.docs.amd.com/projects/rocprofiler-systems/en/latest)
    - [ROCprofiler-SDK](https://rocm.docs.amd.com/projects/rocprofiler-sdk/en/latest)
    - [ROCdbgapi](https://rocm.docs.amd.com/projects/ROCdbgapi/en/latest)
    - [ROCm Debugger (ROCgdb)](https://rocm.docs.amd.com/projects/ROCgdb/en/latest)
    - [ROCR Debug Agent](https://rocm.docs.amd.com/projects/rocr_debug_agent/en/latest)
  - [Control and monitoring tools](https://rocm.docs.amd.com/en/latest/components/control-and-monitoring-tools.html): AMD ROCm control and monitoring tools for inspecting and managing AMD GPU hardware state.
    - [AMD SMI](https://rocm.docs.amd.com/projects/amdsmi/en/latest)
    - [ROCm Data Center Tool (RDC)](https://rocm.docs.amd.com/projects/rdc/en/latest)
    - [rocminfo](https://rocm.docs.amd.com/projects/rocminfo/en/latest)
- [ROCm HPC SDK](https://rocm.docs.amd.com/en/latest/components/hpc-sdk/index.html): Learn how to use AMD ROCm for high-performance computing (HPC).
  - [Install HPC SDK](https://rocm.docs.amd.com/en/latest/components/hpc-sdk/install.html): Install ROCm to run high-performance computing (HPC) workloads.
- [ROCm Extras](https://rocm.docs.amd.com/en/latest/components/extras.html): AMD ROCm Extras - list of tools and test suites
  - [ROCm Validation Suite (RVS)](https://rocm.docs.amd.com/projects/ROCmValidationSuite/en/docs-1.5.122/index.html)
  - [TransferBench](https://rocm.docs.amd.com/projects/TransferBench/en/docs-1.66.02/index.html)
- [AI ecosystem](https://rocm.docs.amd.com/en/latest/ai-ecosystem.html)
  - [PyTorch](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/frameworks/pytorch/install.html)
  - [JAX](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/frameworks/jax/install.html)
  - [vLLM](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/inference/vllm.html)
  - [SGLang](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/inference/sglang.html)
- [AI Developer Hub](https://rocm.docs.amd.com/projects/ai-developer-hub/en/latest/)
- [ROCm examples](https://github.com/ROCm/rocm-examples/tree/release/therock-7.14)
- [AMD GPU specifications](https://rocm.docs.amd.com/en/latest/reference/gpu-specs.html): AMD Instinct™ GPU, AMD Radeon PRO™, and AMD Radeon™ GPU architecture information
- [AMD GPU architectures](https://rocm.docs.amd.com/en/latest/reference/gpu-arch/index.html)
  - [MI350 microarchitecture](https://rocm.docs.amd.com/en/latest/reference/gpu-arch/mi350.html): AMD Instinct MI350 Series microarchitecture reference.
    - [CDNA4 ISA reference](https://www.amd.com/content/dam/amd/en/documents/instinct-tech-docs/instruction-set-architectures/amd-instinct-cdna4-instruction-set-architecture.pdf)
    - [CDNA4 white paper](https://www.amd.com/content/dam/amd/en/documents/instinct-tech-docs/white-papers/amd-cdna-4-architecture-whitepaper.pdf)
    - [Instinct MI350 Series performance counters](https://rocm.docs.amd.com/en/latest/reference/gpu-arch/mi350-performance-counters.html): MI355 Series performance counters and metrics
  - [MI300 microarchitecture](https://rocm.docs.amd.com/en/latest/reference/gpu-arch/mi300.html): Learn about the AMD Instinct MI300 Series architecture.
    - [CDNA3 ISA reference](https://www.amd.com/content/dam/amd/en/documents/instinct-tech-docs/instruction-set-architectures/amd-instinct-mi300-cdna3-instruction-set-architecture.pdf)
    - [CDNA3 white paper](https://www.amd.com/content/dam/amd/en/documents/instinct-tech-docs/white-papers/amd-cdna-3-white-paper.pdf)
    - [Instinct MI300 and MI200 Series performance counters](https://rocm.docs.amd.com/en/latest/reference/gpu-arch/mi300-mi200-performance-counters.html): MI300 and MI200 Series performance counters and metrics
  - [MI250 microarchitecture](https://rocm.docs.amd.com/en/latest/reference/gpu-arch/mi250.html): Learn about the AMD Instinct MI250 Series architecture.
    - [CDNA2 ISA reference](https://www.amd.com/content/dam/amd/en/documents/instinct-tech-docs/instruction-set-architectures/instinct-mi200-cdna2-instruction-set-architecture.pdf)
    - [CDNA2 white paper](https://www.amd.com/content/dam/amd/en/documents/instinct-business-docs/white-papers/amd-cdna2-white-paper.pdf)
  - [MI100 microarchitecture](https://rocm.docs.amd.com/en/latest/reference/gpu-arch/mi100.html): Learn about the AMD Instinct MI100 Series architecture.
    - [CDNA1 ISA reference](https://www.amd.com/content/dam/amd/en/documents/instinct-tech-docs/instruction-set-architectures/instinct-mi100-cdna1-shader-instruction-set-architecture.pdf)
    - [CDNA1 white paper](https://www.amd.com/content/dam/amd/en/documents/instinct-business-docs/white-papers/amd-cdna-white-paper.pdf)
- [AMD GPU system optimization](https://rocm.docs.amd.com/en/latest/reference/system-optimization/index.html): Learn about AMD hardware optimization for HPC-specific and workstation workloads.
  - [AMD Instinct GPUs](https://rocm.docs.amd.com/en/latest/reference/system-optimization/cdna.html): System optimization guides for AMD Instinct CDNA GPUs.
    - [AMD MI355X](https://instinct.docs.amd.com/projects/system-acceptance/en/latest/gpus/mi355x.html)
    - [AMD MI350X](https://instinct.docs.amd.com/projects/system-acceptance/en/latest/gpus/mi350x.html)
    - [AMD MI325X](https://instinct.docs.amd.com/projects/system-acceptance/en/latest/gpus/mi325x.html)
    - [AMD MI300X](https://instinct.docs.amd.com/projects/system-acceptance/en/latest/gpus/mi300x.html)
    - [AMD MI300A](https://instinct.docs.amd.com/projects/system-acceptance/en/latest/gpus/mi300a.html)
    - [AMD MI250 / MI250X](https://instinct.docs.amd.com/projects/system-acceptance/en/latest/gpus/mi250.html)
    - [AMD MI210](https://instinct.docs.amd.com/projects/system-acceptance/en/latest/gpus/mi210.html)
    - [AMD MI100](https://instinct.docs.amd.com/projects/system-acceptance/en/latest/gpus/mi100.html)
  - [AMD Radeon and Ryzen GPUs](https://rocm.docs.amd.com/en/latest/reference/system-optimization/rdna.html): System optimization guides for AMD Radeon and Ryzen RDNA GPUs.
    - [RDNA3.5](https://rocm.docs.amd.com/en/latest/reference/system-optimization/rdna3-5.html): System optimization of AMD RDNA3.5 Ryzen APUs (gfx1150/gfx1151/gfx1152) systems. Learn about VRAM, GTT, TTM tuning, shared memory configuration, and required Linux kernel support.
    - [RDNA2](https://rocm.docs.amd.com/en/latest/reference/system-optimization/rdna2.html): Learn about system settings and performance tuning for RDNA2-based GPUs.
  - [Common system settings](https://rocm.docs.amd.com/en/latest/reference/system-optimization/common.html): Common system settings for AMD GPUs including GPU isolation and BAR access configuration.
    - [GPU isolation techniques](https://rocm.docs.amd.com/en/latest/reference/system-optimization/gpu-isolation.html)
    - [BAR access limits](https://rocm.docs.amd.com/en/latest/reference/system-optimization/bar-access-limits.html): Learn about BAR configuration in AMD GPUs and ways to troubleshoot physical addressing limit
- [AMD GPU and ROCm components data types and precision support](https://rocm.docs.amd.com/en/latest/reference/precision-support.html): Supported data types of AMD GPUs and libraries in ROCm.
- [ROCm libraries graph-safe support](https://rocm.docs.amd.com/en/latest/reference/graph-safe-support.html): This page lists supported graph safe ROCm libraries.
- [AMD GPU atomics operation support](https://rocm.docs.amd.com/en/latest/reference/gpu-atomics-operation.html): AMD Instinct GPU, AMD Radeon PRO, and AMD Radeon GPU atomics operations information
- [ROCm environment variables](https://rocm.docs.amd.com/en/latest/reference/environment-variables/index.html): Environment variables reference
  - [Set the number of CUs](https://rocm.docs.amd.com/en/latest/reference/environment-variables/setting-cus.html): Setting the number of CUs
- [AMD GPU programming on ROCm](https://rocm.docs.amd.com/en/latest/reference/hip-programming.html): HIP programming
- [Contributing to the ROCm documentation](https://rocm.docs.amd.com/en/latest/contribute/contributing.html)
  - [ROCm documentation toolchain](https://rocm.docs.amd.com/en/latest/contribute/toolchain.html)
  - [Building documentation](https://rocm.docs.amd.com/en/latest/contribute/building.html)
- [Providing feedback about the ROCm documentation](https://rocm.docs.amd.com/en/latest/contribute/feedback.html)
- [ROCm licenses](https://rocm.docs.amd.com/en/latest/about/license.html)

---

Source: https://rocm.docs.amd.com/en/latest/index.html

# AMD ROCm

ROCm is AMD’s open-source GPU computing platform: an end‑to‑end ecosystem of
compilers, runtimes, and libraries for AI, HPC, and domain‑specific workloads.
It is open source, cross‑platform (Linux and Windows), and optimized for AMD
Instinct™, AMD Radeon™, and AMD Ryzen™ AI devices.

The foundational libraries, runtimes, and tools for GPU computing on AMD
hardware — math and compute libraries, communication primitives, HIP runtime,
profiling and debugging tools, and more.

- [Install the ROCm Core SDK](https://rocm.docs.amd.com/en/latest/install/rocm.html)
- [ROCm 7.14.0 compatibility](https://rocm.docs.amd.com/en/latest/compatibility/compatibility-matrix.html)

[Go to Core SDK docs](https://rocm.docs.amd.com/en/latest/about/release-notes.html)

Full-stack documentation and recipes to deploy AI workloads on AMD
GPUs using popular ROCm-enabled frameworks.

- Deep learning frameworks
  - [PyTorch](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/frameworks/pytorch/install.html)
  - [JAX](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/frameworks/jax/install.html)
- Inference
  - [vLLM](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/inference/vllm.html)
  - [SGLang](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/inference/sglang.html)

[Go to AI Ecosystem docs](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/index.html)

Deployment and operations guidance for AMD Instinct GPUs at scale, including
AMD GPU Driver installation, cluster management, GPU partitioning, monitoring,
virtualization, cloud deployments, and containers.

- [AMD GPU Driver (amdgpu)](https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/)
- [Network Operator (Kubernetes)](https://instinct.docs.amd.com/projects/network-operator/en/latest/overview.html)

[Go to GPU Systems docs](https://instinct.docs.amd.com)

Open-source collections of ROCm-accelerated libraries for building
high-performance domain-specific applications.

- [ROCm Data Science](https://rocm.docs.amd.com/projects/rocm-ds/en/latest/index.html)
- [ROCm Finance](https://rocm.docs.amd.com/projects/rocm-finance/en/latest/index.html)
- [ROCm Life Science](https://rocm.docs.amd.com/projects/rocm-ls/en/latest/index.html)
- [ROCm LLMExt](https://rocm.docs.amd.com/projects/rocm-llmext/en/latest/index.html)
- [ROCm Simulation](https://rocm.docs.amd.com/projects/rocm-simulation/en/latest/index.html)

---

Source: https://rocm.docs.amd.com/en/latest/about/release-notes.html

# ROCm Core SDK 7.14.0 release notes

* **Leaner core**: The Core SDK focuses on essential runtime and development components.
* **Use case-specific expansions**: Optional domain-specific SDKs for AI, data science, and HPC.
* **Modular installation**: Install only the components required for your workflow.

#### NOTE
ROCm 7.14.0 follows the [versioning discontinuity that began with the 7.9.0 preview](https://rocm.docs.amd.com/en/7.9.0-preview/about/release-notes.html#preview-stream-note) release.

## Release highlights

### Platform and hardware support

This release expands GPU, operating system, virtualization, and partitioning support.

#### Expanded AMD GPU support

ROCm 7.14.0 adds support for the following AMD APUs:

* [AMD Ryzen AI Max+ PRO 495 (gfx1151)](https://www.amd.com/en/products/processors/laptop/ryzen-pro/ai-max-pro-400-series/amd-ryzen-ai-max-plus-pro-495.html)
* [AMD Ryzen AI Max PRO 490 (gfx1151)](https://www.amd.com/en/products/processors/laptop/ryzen-pro/ai-max-pro-400-series/amd-ryzen-ai-max-pro-490.html)
* [AMD Ryzen AI Max PRO 485 (gfx1151)](https://www.amd.com/en/products/processors/laptop/ryzen-pro/ai-max-pro-400-series/amd-ryzen-ai-max-pro-485.html)
* [AMD Ryzen AI 5 435 (gfx1153)](https://www.amd.com/en/products/processors/laptop/ryzen/ai-400-series/amd-ryzen-ai-5-435.html)
* [AMD Ryzen AI 5 430 (gfx1153)](https://www.amd.com/en/products/processors/laptop/ryzen/ai-400-series/amd-ryzen-ai-5-430.html)
* [AMD Ryzen AI 5 PRO 435 (gfx1153)](https://www.amd.com/en/products/processors/laptop/ryzen-pro/ai-400-series/amd-ryzen-ai-5-pro-435.html)
* [AMD Ryzen AI 7 445 (gfx1153)](https://www.amd.com/en/products/processors/laptop/ryzen/ai-400-series/amd-ryzen-ai-7-445.html)

For the complete list of supported AMD hardware, see [AMD hardware support]().

#### Expanded operating system support

SUSE Linux Enterprise Server (SLES) 15 SP7, SLES 16, and Debian 13 are now supported on AMD Instinct MI350P.

For the full list of supported Linux distributions, see [Operating system support]().

#### Expanded GPU virtualization support for Instinct and Radeon GPUs

ROCm 7.14.0 adds support for the following virtualization configurations on AMD Instinct and Radeon GPUs:

* On MI350X: VMware ESXi 9.1 with Ubuntu 24.04 guest OS.
* On Radeon AI PRO R9700S: KVM Passthrough with Ubuntu 24.04 host OS and Ubuntu 24.04 guest OS.
* On Radeon PRO V710: KVM SR-IOV with Ubuntu 24.04 host OS and Ubuntu 24.04 guest OS.

#### Expanded Instinct GPU partitioning support

ROCm 7.14.0 has enabled and optimized multi-VF partition modes for the following GPU partitioning configurations in SR-IOV deployments:

On MI355X and MI350X:

DPX compute partition mode with NPS2 memory partitioning.

CPX compute partition mode with NPS2 memory partitioning.

For details, see [GPU partitioning support]().

### AI inference and frameworks

This release enables support for the following frameworks:

* PyTorch 2.12.0
* JAX 0.10.0
* vLLM 0.23.0 [\*]
* SGLang 0.5.13
* TensorFlow 2.21

The updated framework support replaces the previous PyTorch 2.9.1, JAX 0.8.2, vLLM 0.19.1, and SGLang 0.5.9 support.

For details, see [AI ecosystem support]().

### Developer tools, profiling, and validation

This release improves ROCm developer workflows with new HIP APIs, expanded profiling and tracing capabilities, and broader telemetry coverage.

#### HIP feature highlights

The following are notable enhancements to HIP:

* **HIP execution context support**: HIP now supports Execution Context APIs, enabling GPU compute resource partitioning and lightweight execution-context management on a single device. These APIs allow you to query and partition device resources (primarily CU count for HIP runtime), create execution contexts on resource subsets, and create streams and events scoped to those contexts. For more information, see [Execution Context Management](https://rocm.docs.amd.com/projects/HIP/en/latest/reference/hip_runtime_api/modules/execution_context_management.html).
* **HIP API additions for CUDA parity**:
  * **Batch memory management**: New batch asynchronous memory management APIs let applications discard (`hipMemDiscardBatchAsync`), prefetch (`hipMemPrefetchBatchAsync`), or combine both operations (`hipMemDiscardAndPrefetchBatchAsync`) across multiple memory ranges in a single call, reducing API call overhead. Both HIP runtime and HIP driver variants are available.
  * **Library management**: New library management APIs return the device pointer and size of a device global (`hipLibraryGetGlobal`) and the host pointer and size of a managed variable (`hipLibraryGetManaged`) defined in a `hipLibrary_t`, improving parity with CUDA library APIs.
* **Faster HIP graph replay for asynchronous memory allocations**: HIP graph replay now reduces overhead for graphs that interleave asynchronous memory allocations with compute. Allocation nodes no longer block during replay. Physical memory is reused across nodes instead of being mapped and unmapped on each launch, eliminating the gaps between kernels this pattern previously caused. For background on HIP graphs, see [Graph Management](https://rocm.docs.amd.com/projects/HIP/en/latest/reference/hip_runtime_api/modules/graph_management.html).
* **New HIP documentation topic on device properties**: The [HIP Device Properties and Topology on CDNA Architectures](https://rocm.docs.amd.com/projects/HIP/en/latest/how-to/hipDeviceProperties.html) provides an overview of the CDNA architecture topology and its key hardware characteristics, helping developers better understand the underlying architecture and optimize the performance of their HIP applications.

For more information, see the [HIP section]() in the ROCm component changelogs.

#### ROCprofiler-SDK feature highlights

The following are notable enhancements to ROCprofiler-SDK:

##### ROCprofiler-SDK integration with PyTorch Profiler

##### ROCprofiler-SDK beta support for Streaming Performance Monitors

Supported hardware: AMD Instinct MI300X, MI325X, MI350X, and MI355X GPUs.

#### WARNING
SPM is a beta capability under active development and may affect system stability, including unexpected reboots. Do not use in production environments. See [ROCm known issues]() for current limitations.

##### Selective ROCTx region profiling with counter collection

`rocprofiler-sdk` and `rocprofv3` include support for profiling selected ROCTx regions, allowing users to focus profiling on specific application phases instead of collecting data for the entire workload. By inserting `roctxProfilerPause` and `roctxProfilerResume` markers in application code and using the `--selected-regions` option, only the GPU activity within the marked regions is captured. This helps reduce profiling noise and output size while making it easier to isolate performance behavior in targeted code paths. This is particularly useful for long-running workloads where full-execution traces are impractical.

Counter collection for selected regions is available in ROCm 7.14.0. For details on `--selected-regions`, including usage with RCCL collectives and ROCTx markers, see [Using ROCprofiler-SDK ROCTx](https://rocm.docs.amd.com/projects/rocprofiler-sdk/en/latest/how-to/using-rocprofiler-sdk-roctx.html).

##### Improved attach and re-attach profiling workflows

`rocprofiler-sdk` and `rocprofv3` improve attach-based profiling workflows, allowing the profiler to connect to already-running GPU applications without requiring a restart. This supports production-style and long-running workloads where starting the application under the profiler is not always practical. Overall reliability and stability are also improved. Repeated attach and re-attach sessions now generate separate output files, making it easier to manage results from iterative profiling sessions. For details, see [Dynamic process attachment using rocprofv3](https://rocm.docs.amd.com/projects/rocprofiler-sdk/en/latest/how-to/using-rocprofv3-process-attachment.html).

##### Reduced profiling overhead for counter collection

`rocprofiler-sdk` reduces profiling overhead in ROCm 7.14.0, including customer-driven improvements for `rocprofv3` and SDK-based profiling paths. These changes make profiling more practical for performance-sensitive workflows and produce more representative trace data.

##### ROCprof Trace Decoder decoupled from the ROCprofiler-SDK API

`rocprof-trace-decoder` now has an independent public API, separate from the core `rocprofiler-sdk` API. The SQTT decoding API has moved into `rocprof-trace-decoder`; use the decoder’s public API directly rather than `rocprofiler-sdk` for SQTT decoding. Previously, `rocprof-trace-decoder` was an internal plugin within `rocprofiler-sdk`, tightly coupling the two. The independent API can be versioned separately, allowing tools such as `rocpd` to consume the decoder without depending on the full SDK runtime. The `rocprofv3` end-user experience is unchanged.

##### Quality and stability improvements

This release includes a range of quality and stability improvements across `rocprofiler-sdk` and `rocprofv3`. These include corrections to hardware counter reporting on specific GPU architectures, improvements to trace output accuracy, build fixes on newer GPU targets, expanded platform coverage in `aqlprofile`, and strengthened build and test support in TheRock CI. These changes improve the robustness and correctness of the profiling stack across supported hardware configurations.

##### Queue interposition as a lighter alternative to queue interception

`rocprofiler-sdk` introduces queue interposition, a mechanism that virtualizes HSA queue write-pointer operations without modifying the ROCR-Runtime or requiring full queue interception. For workloads that do not use dispatch counter collection, dispatch thread trace, or PC sampling, `rocprofiler-sdk` now defaults to this lighter interposition path, which reduces profiling overhead and improves overall stability. The legacy queue interception path remains in use when any of those features are active. The default behavior can be controlled via the `ROCPROFILER_QUEUE_INTERPOSITION` environment variable.

##### ROCprof Compute Viewer 0.2.0 release

ROCprof Compute Viewer (RCV) 0.2.0 adds the ability to open raw `.att` and `.out` thread trace directories directly without a JSON conversion step, and introduces a Flamegraph view with per-CU and SIMD source and ISA stack rollups replacing the previous Explorer view. This release also adds hidden latency analysis for gfx10+ and Navi thread traces, SQTT instrumentation marker visualization, and a heuristic GPU Utilization derived counter, alongside fixes for scaling issues, Global View misalignments, and real-time alignment for both JSON and raw `.att` inputs. Installers are available for Windows (`.exe`) and macOS ARM64 (`.sh`), with GitHub Actions CI and release workflows now in place.

#### ROCm Compute Profiler feature highlights

The following are notable enhancements to the ROCm Compute Profiler (rocprofiler-compute):

* **PyTorch operator statistics (experimental)**: The PyTorch tracing (`--torch-trace`) now includes a per-operator statistics summary table, making it easier to spot hot operators and per-dispatch variance. The trace now also captures backward-pass and nested operators that were previously missed or misattributed.
* **Faster analysis**: Analyze mode now processes profiling data more efficiently, reducing analysis time and memory usage on large workloads.
* **Improved metric averaging accuracy**: Metric values across multiple kernel dispatches are now correctly weighted-averaged, eliminating errors that occurred when aggregating metrics with varying values across individual dispatches.
* **pip installation support**: ROCm Compute Profiler is now available as a pip-installable Python package. A new `rocm-profiler` wheel on the ROCm Python package index lets you install ROCm Compute Profiler into a custom Python environment without building ROCm from source. The wheel package installs both ROCm Compute Profiler and ROCm Systems Profiler binaries. For installation instructions, see [Install ROCm Compute Profiler](https://rocm.docs.amd.com/projects/rocprofiler-compute/en/latest/install/core-install.html).

For more information, see the [ROCm Compute Profiler section]() in the ROCm component changelogs.

#### ROCm Systems Profiler feature highlights

The following are notable enhancements to ROCm Systems Profiler:

* **GPU hardware counter sampling**: ROCm Systems Profiler now supports periodic sampling of Performance Metric Counters (PMCs). This lets you collect performance metrics at regular time intervals without serializing kernel dispatches, enabling profiling with significantly reduced overhead. For details, see the GPU metrics section in [Configuring runtime options](https://rocm.docs.amd.com/projects/rocprofiler-systems/en/latest/how-to/configuring-runtime-options.html).
* **Unified memory profiling**: ROCm Systems Profiler now adds unified memory profiling with statistics on page migrations and page faults through a dedicated report section. The report displays aggregated transfer counts, sizes, and timing to help you identify performance bottlenecks caused by host-device memory migrations and page faults. Page table events are collected through ROCprofiler-SDK using the Kernel Fusion Driver (KFD) events API, replacing the previous page-migration API, with page migration and page fault events exposed directly in the profiler output. For details, see [Unified memory profiling](https://rocm.docs.amd.com/projects/rocprofiler-systems/en/latest/how-to/unified-memory-profiling.html).
* **SDMA engine activity profiling**: ROCm Systems Profiler can now collect System Direct Memory Access (SDMA) engine activity through AMD SMI, reporting per-process DMA usage to expose data-movement bottlenecks across GPU interconnects in multi-GPU workloads. This metric is opt-in and requires AMDGPU driver version 6.19.14 or later.
* **Selective MPI rank profiling**: In MPI jobs, you can now restrict profile and trace output to a chosen subset of ranks, while unselected ranks run undisturbed. This cuts data volume and speeds up post-run analysis, and works across MPI implementations such as MPICH and Open MPI, including heterogeneous and multi-node environments. For details, see the rank filtering section in [Communication runtime profiling](https://rocm.docs.amd.com/projects/rocprofiler-systems/en/latest/how-to/communication-runtime-profiling.html).
* **MPI rank console log control**: You can now limit console output to a specified subset of ranks while profiling and tracing continue on every rank. This reduces console log noise in large multi-rank runs without sacrificing collection coverage. Existing behavior is preserved when no rank-selection option is set.
* **Selective ROCTx region profiling**: When your application is instrumented with ROCTx region push and pop APIs, you can now include or exclude specific named regions to scope collection to the code paths you’re investigating.
* **pip installation support**: ROCm Systems Profiler is now available as a pip-installable Python package, letting you install and use the profiler in custom Python environments without rebuilding from source. The package supports multiple Python versions and provides the same functionality as traditional distributions, reducing setup time and complexity. For installation instructions, see [Install ROCm Systems Profiler](https://rocm.docs.amd.com/projects/rocprofiler-systems/en/latest/install/install.html).

For more information, see the [ROCm Systems Profiler section]() in the ROCm component changelogs.

#### AMD SMI feature highlights

The following are notable enhancements to AMD SMI:

* **Per-partition GPU metrics**: AMD SMI now reports temperature, clock, and usage at the partition level through the new `amd-smi metric --partition` flag, giving partition-level observability where previously only socket-level metrics were available. For CLI usage, see [AMD SMI CLI tool](https://rocm.docs.amd.com/projects/amdsmi/en/latest/how-to/amdsmi-cli-tool.html); for partitioning concepts, see [GPU partitioning](https://rocm.docs.amd.com/projects/amdsmi/en/latest/conceptual/partition.html).
* **Compute partition memory allocation mode**: AMD SMI now controls memory allocation behavior at the compute partition level through the new `amd-smi set --compute-partition-mem-alloc-mode` command. The current mode is visible in `amd-smi static --partition` output, and new C and Python APIs expose the same controls programmatically.
* **APU CLI metrics**: AMD SMI now surfaces APU-specific data through the existing `amd-smi metric` flags when APU metrics are available. `amd-smi monitor` adds temperature and clock fallbacks when standard discrete GPU sensors report N/A.
* **APU VRAM carve-out and GTT tuning**: AMD SMI now tunes APU memory from the command line, consolidating the get and set controls previously handled by the standalone `amd-ttm` tool and adding VRAM carve-out configuration. Carve-out and GTT changes take effect after the next reboot, and AMD SMI rebuilds the initramfs automatically so the new configuration is applied at boot. For details, see the memory tuning section in [AMD SMI CLI tool](https://rocm.docs.amd.com/projects/amdsmi/en/latest/how-to/amdsmi-cli-tool.html).
* **PID-grouped process listing**: AMD SMI now groups multi-GPU process output by PID with `amd-smi process --sort-by-pid` and `amd-smi monitor --sort-by-pid`, merging each process’s per-GPU usage into a single row. A new C and Python API, `amdsmi_get_gpu_process_list_by_pid()`, exposes the same data programmatically.
* **Fabric clock (FCLK) capping on MI300A**: You can now cap the maximum fabric clock (FCLK) on AMD Instinct MI300A APUs to steer power, using the new `fclk` clock type for `amd-smi set --clk-limit`. Only a maximum limit is supported.
* **Go bindings for CPU telemetry**: AMD SMI now exposes EPYC System Management Interface (ESMI) CPU functionality through its Go bindings, so Go applications can query CPU telemetry in-process without invoking external binaries or embedding C or Python runtimes. This simplifies integrating AMD CPU observability into Go-based control planes. For details, see [AMD SMI Go interface](https://rocm.docs.amd.com/projects/amdsmi/en/latest/how-to/amdsmi-go-lib.html).

For more information, see the [AMD SMI section]() in the ROCm component changelogs.

#### RDC expands telemetry coverage for DME parity

ROCm Data Center (RDC) adds 59 telemetry fields, bringing its metric coverage near parity with the Device Metrics Exporter (DME). New fields cover energy, temperature, clocks, memory, PCIe, engine activity, error correction code (ECC), and health and throttle metrics. Some metrics require recent driver and hardware support. For the available field groups and how to monitor them, see [Using RDC features](https://rocm.docs.amd.com/projects/rdc/en/latest/how-to/using_RDC_features.html).

For more information, see the [RDC section]() in the ROCm component changelogs.

#### ROCm Bandwidth Test (RBT) reaches end-of-life

ROCm Bandwidth Test (RBT) is deprecated and reaches end-of-life with the TheRock-based ROCm 7.14.0 release. Active development has ceased, and no further feature enhancements or fixes are planned. For equivalent and expanded functionality, transition to [TransferBench](https://rocm.docs.amd.com/projects/TransferBench/en/latest/) and the [ROCm Validation Suite (RVS)](https://rocm.docs.amd.com/projects/ROCmValidationSuite/en/latest/).

For more details, refer to the [ROCm Bandwidth Test](https://rocm.docs.amd.com/projects/rocm_bandwidth_test/en/latest/) documentation.

### Libraries

This release updates ROCm math, sparse compute, and communication libraries with additional routines, expanded datatype support, and performance improvements. It also adds the hipFile storage library.

#### hipFile direct storage I/O support

hipFile is now included in the ROCm Core SDK, enabling direct data transfers between storage and GPU memory as part of AMD Infinity Storage. hipFile enables storage-intensive workloads to bypass host-side copies, reducing latency and command overhead for high-throughput GPU I/O.

hipFile is supported on Linux with AMD Instinct GPUs. See the [ROCm hipFile examples](https://github.com/ROCm/rocm-examples/tree/release/therock-7.14/Systems/hipFile) and the [hipFile documentation](https://rocm.docs.amd.com/projects/hipFile/en/latest/) to get started.

#### Per-matrix bias support in hipBLASLt batched GEMM

hipBLASLt now supports applying a unique bias vector to each matrix in a strided batched GEMM. Set the new `HIPBLASLT_MATMUL_DESC_BIAS_BATCH_STRIDE` matmul descriptor attribute to specify the stride between consecutive bias vectors in device memory when `HIPBLASLT_EPILOGUE_BIAS` is set in the epilogue. A stride of 0 (the default) preserves the previous behavior of broadcasting a single bias vector to all matrices in the batch.

For more information, see the [hipBLASLt documentation](https://rocm.docs.amd.com/projects/hipBLASLt/en/latest/index.html).

#### Per-batch scalar coefficients for Level 2 batched BLAS

rocBLAS and hipBLAS now support per-batch scalar coefficients for Level 2 batched and strided-batched routines in device pointer mode. Each batch index uses its own device-resident scalar rather than a single value shared across the entire batch:

* **GEMV**: per-batch alpha and beta.
* **GER, GERU, and GERC**: per-batch alpha.

S, D, C, and Z precision variants are available for all routines.

For more information, see the [rocBLAS section]() and [hipBLAS section]() in the ROCm component changelogs.

#### Per-batch alpha for axpy_batched and axpy_strided_batched

rocBLAS adds per-batch alpha support for `axpy_batched`, `axpy_strided_batched`, and their `_ex` variants via `rocblas_set_batch_alpha_stride` in device pointer mode.

For more information, see the [rocBLAS section]() in the ROCm component changelogs.

#### hipSPARSE feature highlights

The following are notable enhancements to hipSPARSE:

* **BSR format support in hipSPARSE generic routines**: hipSPARSE adds Block Sparse Row (BSR) format support to its generic sparse compute routines: `hipsparseSpMM` (sparse matrix-matrix multiplication) and `hipsparseSpMV` (sparse matrix-vector multiplication). Two new descriptor functions, `hipsparseCreateBsr` and `hipsparseCreateConstBsr`, let you construct BSR-format sparse matrices for use with the generic API. This brings hipSPARSE to parity with the equivalent NVIDIA cuSPARSE routines, where BSR was previously available only through the rocSPARSE API.
* **Legacy SpGEAM routines deprecated**: The legacy hipSPARSE `csrgeam` routines (`hipsparseXcsrgeamNnz`, `hipsparseScsrgeam`, `hipsparseDcsrgeam`, `hipsparseCcsrgeam`, and `hipsparseZcsrgeam`) are deprecated and will be removed in a future release. Use the `csrgeam2` routines instead: `hipsparseScsrgeam2_bufferSizeExt`, `hipsparseDcsrgeam2_bufferSizeExt`, `hipsparseCcsrgeam2_bufferSizeExt`, `hipsparseZcsrgeam2_bufferSizeExt`, `hipsparseXcsrgeam2Nnz`, `hipsparseScsrgeam2`, `hipsparseDcsrgeam2`, `hipsparseCcsrgeam2`, and `hipsparseZcsrgeam2`.

#### rocSPARSE feature highlights

The following are notable enhancements to rocSPARSE:

* **Incomplete LDLᵀ factorization**: rocSPARSE adds the `rocsparse_spildlt0` routine, which computes an incomplete LDLᵀ factorization with zero fill-in (ILDLT(0)) for symmetric real or Hermitian complex sparse matrices in CSR format. The routine supports strided batched computation for factoring multiple matrices in a single call, a common building block for preconditioning iterative sparse solvers.
* **`rocsparse_indextype_u16` index type deprecated**: The `rocsparse_indextype_u16` field of the `rocsparse_indextype` enum is deprecated in this release. Code using `rocsparse_indextype_u16` now produces deprecation warnings at compile time. Migrate to `rocsparse_indextype_i32` or `rocsparse_indextype_i64`; `rocsparse_indextype_u16` will be removed in a future release.

#### RCCL feature highlights

The following are notable enhancements to RCCL:

* **Hierarchical AllGather**: RCCL adds a hierarchical AllGather algorithm for large multi-node jobs by separating inter-node from intra-node communication, relieving the concurrency pressure that constrains the existing ring and direct algorithms across many GPUs. On AMD Instinct MI350X GPUs, hierarchical AllGather is enabled by default for multi-node configurations. To disable it, set `RCCL_HIERARCHICAL_ALLGATHER=0`.
* **Direct reduce-scatter**: RCCL adds a direct reduce-scatter algorithm for small-to-medium message sizes on AMD Instinct MI350X GPUs, as an alternative to the existing ring-based implementation. RCCL selects it automatically for multi-node reduce-scatter operations within a configurable message-size threshold.
* **Copy Engine collectives (Preview)**: RCCL now offloads collective data movement to the GPU copy engine on AMD Instinct MI355X GPUs through new Copy Engine collectives. This frees compute units during communication-bound collectives, so compute and communication can overlap. RCCL uses a batched copy path when available, falls back to multi-stream or single-stream transfers otherwise, and preserves correct behavior during HIP graph capture.

For more information, see the [RCCL section]() in the ROCm component changelogs.

## AMD hardware support

The following table lists supported AMD Instinct GPUs, Radeon GPUs, and Ryzen APUs. Each supported device is listed with its corresponding GPU microarchitecture and LLVM target.

#### NOTE
If your GPU is not listed, it might be community-enabled through TheRock nightly builds. For more information, see [TheRock supported GPUs](https://github.com/ROCm/TheRock/blob/main/SUPPORTED_GPUS.md). For installation guidance, see [TheRock releases](https://github.com/ROCm/TheRock/blob/main/RELEASES.md).

**Instinct**

| Device series | Device | LLVM target | Architecture |
|-------------------------------|---------------------------------------------------------------------|---------------|----------------|
| AMD Instinct MI350 Series | Instinct MI355X Instinct MI350X Instinct MI350P | gfx950 | CDNA 4 |
| AMD Instinct MI300 Series | Instinct MI325X Instinct MI300X Instinct MI300A | gfx942 | CDNA 3 |
| AMD Instinct MI200 Series | Instinct MI250X Instinct MI250 Instinct MI210 | gfx90a | CDNA 2 |
| AMD Instinct MI100 Series | Instinct MI100 | gfx908 | CDNA |

**Radeon**

| Device series | Device | LLVM target | Architecture |
|---------------------------------|--------------------------------------------------------------------------------------------------------|---------------|----------------|
| AMD Radeon AI PRO R9000 Series | Radeon AI PRO R9700S Radeon AI PRO R9700 Radeon AI PRO R9600D | gfx1201 | RDNA 4 |
| AMD Radeon RX 9000 Series | Radeon RX 9070 XT Radeon RX 9070 GRE Radeon RX 9070 | gfx1201 | RDNA 4 |
| AMD Radeon RX 9000 Series | Radeon RX 9060 XT LP Radeon RX 9060 XT Radeon RX 9060 | gfx1200 | RDNA 4 |
| AMD Radeon PRO W7000 Series | Radeon PRO W7900 Dual Slot Radeon PRO W7900 Radeon PRO W7800 48GB Radeon PRO W7800 | gfx1100 | RDNA 3 |
| AMD Radeon PRO W7000 Series | Radeon PRO W7700 | gfx1101 | RDNA 3 |
| AMD Radeon RX 7000 Series | Radeon RX 7900 XTX Radeon RX 7900 XT Radeon RX 7900 GRE | gfx1100 | RDNA 3 |
| AMD Radeon RX 7000 Series | Radeon RX 7800 XT Radeon RX 7700 XT Radeon RX 7700 | gfx1101 | RDNA 3 |
| AMD Radeon RX 7000 Series | Radeon RX 7600 | gfx1102 | RDNA 3 |
| AMD Radeon PRO V Series | Radeon PRO V710 | gfx1101 | RDNA 3 |
| AMD Radeon PRO V Series | Radeon PRO V620 | gfx1030 | RDNA 2 |
| AMD Radeon PRO W6000 Series | Radeon PRO W6800 | gfx1030 | RDNA 2 |

**Ryzen**

| Device series | Device | LLVM target | Architecture |
|-------------------------------------|---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|---------------|----------------|
| AMD Ryzen AI Max PRO 400 Series | Ryzen AI Max+ PRO 495 (Radeon 8065S) Ryzen AI Max PRO 490 (Radeon 8050S) Ryzen AI Max PRO 485 (Radeon 8050S) | gfx1151 | RDNA 3.5 |
| AMD Ryzen AI Max PRO 300 Series | Ryzen AI Max+ PRO 395 (Radeon 8060S) Ryzen AI Max PRO 390 (Radeon 8050S) Ryzen AI Max PRO 385 (Radeon 8050S) Ryzen AI Max PRO 380 (Radeon 8040S) | gfx1151 | RDNA 3.5 |
| AMD Ryzen AI Max 300 Series | Ryzen AI Max+ 395 (Radeon 8060S) Ryzen AI Max+ 392 (Radeon 8060S) Ryzen AI Max+ 388 (Radeon 8060S) Ryzen AI Max 390 (Radeon 8050S) Ryzen AI Max 385 (Radeon 8050S) | gfx1151 | RDNA 3.5 |
| AMD Ryzen AI PRO 400 Series | Ryzen AI 5 PRO 435 (Radeon 840M) | gfx1153 | RDNA 3.5 |
| AMD Ryzen AI PRO 400 Series | Ryzen AI 9 HX PRO 475 (Radeon 890M) Ryzen AI 9 HX PRO 470 (Radeon 890M) Ryzen AI 9 PRO 465 (Radeon 880M) | gfx1150 | RDNA 3.5 |
| AMD Ryzen AI PRO 400 Series | Ryzen AI 7 PRO 450 (Radeon 860M) Ryzen AI 5 PRO 440 (Radeon 840M) | gfx1152 | RDNA 3.5 |
| AMD Ryzen AI 400 Series | Ryzen AI 5 435 (Radeon 840M) Ryzen AI 5 430 (Radeon 840M) Ryzen AI 7 445 (Radeon 840M) | gfx1153 | RDNA 3.5 |
| AMD Ryzen AI 400 Series | Ryzen AI 9 HX 475 (Radeon 890M) Ryzen AI 9 HX 470 (Radeon 890M) Ryzen AI 9 465 (Radeon 880M) | gfx1150 | RDNA 3.5 |
| AMD Ryzen AI 400 Series | Ryzen AI 7 450 (Radeon 860M) | gfx1152 | RDNA 3.5 |
| AMD Ryzen AI PRO 300 Series | Ryzen AI 9 HX PRO 375 (Radeon 890M) Ryzen AI 9 HX PRO 370 (Radeon 890M) | gfx1150 | RDNA 3.5 |
| AMD Ryzen AI PRO 300 Series | Ryzen AI 7 PRO 350 (Radeon 860M) Ryzen AI 5 PRO 340 (Radeon 840M) | gfx1152 | RDNA 3.5 |
| AMD Ryzen AI 300 Series | Ryzen AI 9 HX 375 (Radeon 890M) Ryzen AI 9 HX 370 (Radeon 890M) Ryzen AI 9 365 (Radeon 880M) | gfx1150 | RDNA 3.5 |
| AMD Ryzen AI 300 Series | Ryzen AI 7 350 (Radeon 860M) Ryzen AI 7 345 (Radeon 840M) Ryzen AI 5 340 (Radeon 840M) Ryzen AI 5 330 (Radeon 820M) | gfx1152 | RDNA 3.5 |
| AMD Ryzen PRO 200 Series | Ryzen 7 PRO 250 (Radeon 780M) Ryzen 5 PRO 230 (Radeon 760M) Ryzen 5 PRO 220 (Radeon 740M) Ryzen 5 PRO 215 (Radeon 740M) Ryzen 3 PRO 210 (Radeon 740M) | gfx1103 | RDNA 3 |
| AMD Ryzen 200 Series | Ryzen 9 270 (Radeon 780M) Ryzen 7 260 (Radeon 780M) Ryzen 7 250 (Radeon 780M) Ryzen 5 240 (Radeon 760M) Ryzen 5 230 (Radeon 760M) Ryzen 5 220 (Radeon 740M) Ryzen 3 210 (Radeon 740M) | gfx1103 | RDNA 3 |

## Operating system support

ROCm supports the following Linux distributions and Microsoft Windows versions. If you’re running ROCm on Linux, ensure your system is using a supported kernel version.

#### IMPORTANT
The following table is a general overview of supported operating systems. Actual support might vary by AMD GPU or APU. Use the [Compatibility matrix](https://rocm.docs.amd.com/en/latest/compatibility/compatibility-matrix.html) to verify support for your specific setup before installation.

**Instinct**

| Linux distribution                  | Supported versions   | Linux kernel version   |
|-------------------------------------|----------------------|------------------------|
| Ubuntu                              | 26.04                | GA 7.0                 |
| Ubuntu                              | 24.04.4              | GA 6.8                 |
| Ubuntu                              | 22.04.5              | GA 5.15                |
| Debian                              | 13                   | 6.12                   |
| Debian                              | 12                   | 6.1.0                  |
| Red Hat Enterprise Linux (RHEL)     | 10.2                 | 6.12.0-211             |
| Red Hat Enterprise Linux (RHEL)     | 10.0                 | 6.12.0-55              |
| Red Hat Enterprise Linux (RHEL)     | 9.8                  | 5.14.0-687             |
| Red Hat Enterprise Linux (RHEL)     | 9.6                  | 5.14.0-570             |
| Red Hat Enterprise Linux (RHEL)     | 9.4                  | 5.14.0-427             |
| Red Hat Enterprise Linux (RHEL)     | 8.10                 | 4.18.0-553             |
| Oracle Linux                        | 10                   | UEK 8.1                |
| Oracle Linux                        | 9                    | UEK 8                  |
| Oracle Linux                        | 8                    | UEK 7                  |
| Rocky Linux                         | 9                    | 5.14.0-570             |
| SUSE Linux Enterprise Server (SLES) | 16.0                 | 6.12                   |
| SUSE Linux Enterprise Server (SLES) | 15.7                 | 6.4.0-150700.51        |

**Radeon**

| Operating system                | Supported versions   | Linux kernel version   |
|---------------------------------|----------------------|------------------------|
| Ubuntu                          | 26.04                | GA 7.0                 |
| Ubuntu                          | 24.04.4              | GA 6.8                 |
| Ubuntu                          | 22.04.5              | GA 5.15                |
| Red Hat Enterprise Linux (RHEL) | 10.2                 | 6.12.0-211             |
| Red Hat Enterprise Linux (RHEL) | 9.8                  | 5.14.0-687             |
| Windows                         | 11 25H2              | —                      |

**Ryzen**

| Operating system   | Supported versions   | Linux kernel version   |
|--------------------|----------------------|------------------------|
| Ubuntu             | 26.04                | GA 7.0                 |
| Ubuntu             | 24.04.4              | HWE 6.17               |
| Windows            | 11 25H2              | —                      |

## Installation updates

ROCm 7.14.0 introduces several improvements to the Runfile Installer:

### Multi-architecture GPU support

The installer provides multi-architecture support, allowing you to install ROCm components for one or more GPU architectures. This is particularly useful for heterogeneous GPU environments or when deploying across multiple systems with different GPU types.

* Install single or multiple GPU architectures in one installation.
* Autodetect GPU and install matching architecture.
* Query available and installed architectures.
* Selectively uninstall specific architectures while keeping others.

### Flexible component selection

Choose exactly which ROCm components to install, reducing installation time and disk space requirements:

* **core**: Essential runtime libraries and tools (default)
* **core-dev**: Development headers and files
* **dev-tools**: Debugging and profiling utilities
* **core-sdk**: Comprehensive SDK with libraries and development tools
* **opencl**: OpenCL runtime support

### Graphics support

Optional graphics support for Mesa and OpenGL workloads is now available. When enabled, the installer includes the `amdgpu-lib` package for graphics capabilities.

### Build and manifest information

* Display TheRock build information including commit hash, GitHub run ID, and build date
* View complete manifest of all components and their versions included in the installer
* Query components by specific GPU architecture

### Universal installer

A single installer file now supports all Linux distributions, eliminating the need to download distribution-specific builds.

## Kernel driver and firmware bundle support

ROCm requires a coordinated stack of compatible firmware, driver, and user-space components. Maintaining version alignment between these layers ensures correct GPU operation and performance, especially for AMD data center products. While AMD publishes the AMD GPU driver and ROCm user space components, your server OEM (original equipment manufacturer) or infrastructure provider distributes the firmware packages. AMD supplies those firmware images (platform level data model (PLDM) bundles), which the OEM integrates and distributes.

**Instinct**

| AMD device | PLDM Bundle (Firmware) | Linux driver |
|-----------------|-------------------------------------------------------------|-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| Instinct MI355X | 01.26.01.03 (or later) [\*] 01.26.00.02 | AMD GPU Driver (amdgpu) 31.40.1 31.40.0 31.30.0 31.20.0 31.10.0 30.30.3 30.30.2 30.30.1 30.30.0 30.20.1 30.20.0 30.10.2 30.10.1 30.10.0 |
| Instinct MI350X | 01.26.01.03 (or later) [\*] 01.26.00.02 | AMD GPU Driver (amdgpu) 31.40.1 31.40.0 31.30.0 31.20.0 31.10.0 30.30.3 30.30.2 30.30.1 30.30.0 30.20.1 30.20.0 30.10.2 30.10.1 30.10.0 |
| Instinct MI350P | BKC12.0 (IFWI PRD1000A) or later [\*] IFWI 00189938 | AMD GPU Driver (amdgpu) 31.40.1 31.40.0 31.30.0 31.20.0 31.10.0 30.30.3 30.30.2 30.30.1 30.30.0 30.20.1 30.20.0 30.10.2 30.10.1 30.10.0 |
| Instinct MI325X | 01.26.01.03 (or later) [\*] 01.25.06.08 | AMD GPU Driver (amdgpu) 31.40.1 31.40.0 31.30.0 31.20.0 31.10.0 30.30.3 30.30.2 30.30.1 30.30.0 30.20.1 30.20.0 30.10.2 30.10.1 30.10.0 |
| Instinct MI300X | 01.26.00.04 (or later) [\*] 01.25.06.05 | AMD GPU Driver (amdgpu) 31.40.1 31.40.0 31.30.0 31.20.0 31.10.0 30.30.3 30.30.2 30.30.1 30.30.0 30.20.1 30.20.0 30.10.2 30.10.1 30.10.0 |
| Instinct MI300A | BKC 27 [\*] BKC 26.1 | AMD GPU Driver (amdgpu) 31.40.1 31.40.0 31.30.0 31.20.0 31.10.0 30.30.3 30.30.2 30.30.1 30.30.0 30.20.1 30.20.0 30.10.2 30.10.1 30.10.0 |
| Instinct MI250X | Maintenance update (MU) 5 with IFWI 75 (or later) | AMD GPU Driver (amdgpu) 31.40.1 31.40.0 31.30.0 31.20.0 31.10.0 30.30.3 30.30.2 30.30.1 30.30.0 30.20.1 30.20.0 30.10.2 30.10.1 30.10.0 |
| Instinct MI250 | Maintenance update (MU) 5 with IFWI 75 (or later) | AMD GPU Driver (amdgpu) 31.40.1 31.40.0 31.30.0 31.20.0 31.10.0 30.30.3 30.30.2 30.30.1 30.30.0 30.20.1 30.20.0 30.10.2 30.10.1 30.10.0 |
| Instinct MI210 | Maintenance update (MU) 5 with IFWI 75 (or later) | AMD GPU Driver (amdgpu) 31.40.1 31.40.0 31.30.0 31.20.0 31.10.0 30.30.3 30.30.2 30.30.1 30.30.0 30.20.1 30.20.0 30.10.2 30.10.1 30.10.0 |
| Instinct MI100 | VBIOS D3430401-037 | AMD GPU Driver (amdgpu) 31.40.1 31.40.0 31.30.0 31.20.0 31.10.0 30.30.3 30.30.2 30.30.1 30.30.0 30.20.1 30.20.0 30.10.2 30.10.1 30.10.0 |

**Radeon**

| Linux driver | Windows driver |
|-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------|
| AMD GPU Driver (amdgpu) 31.40.1 31.40.0 31.30.0 31.20.0 31.10.0 30.30.3 30.30.2 30.30.1 30.30.0 30.20.1 30.20.0 30.10.2 30.10.1 30.10.0 | AMD Software: Adrenalin Edition 26.6.4 Windows OEM Driver 26.10.28 |

**Ryzen**

| Linux driver | Windows driver |
|---------------------------------------------------------------------------------------|--------------------------------------------------------------------------------|
| Inbox kernel driver (Ubuntu 26.04) 6.14.0-1018 OEM kernel or newer (Ubuntu 24.04) | AMD Software: Adrenalin Edition 26.6.4 Windows OEM Driver 26.10.28 |

## GPU virtualization support

AMD Instinct data center GPUs support virtualization in the following configurations. Supported SR-IOV configurations require the AMD GPU Virtualization Driver (GIM) 9.1.0.K—see the [AMD Instinct Virtualization Driver documentation](https://instinct.docs.amd.com/projects/virt-drv/en/mainline-9.1.0.k/) for more information.

**Instinct**

| AMD GPU         | Hypervisor   | Virtualization technology   | Virtualization driver   | Host OS         | Guest OS     |
|-----------------|--------------|-----------------------------|-------------------------|-----------------|--------------|
| Instinct MI355X | KVM          | Passthrough                 | —                       | Ubuntu 24.04    | Ubuntu 24.04 |
| Instinct MI355X | KVM          | SR-IOV                      | GIM 9.1.0.K             | Ubuntu 24.04    | Ubuntu 24.04 |
| Instinct MI355X | KVM          | SR-IOV                      | GIM 9.1.0.K             | Ubuntu 24.04    | RHEL 10.0    |
| Instinct MI355X | KVM          | SR-IOV                      | GIM 9.1.0.K             | Ubuntu 24.04    | RHEL 9.6     |
| Instinct MI355X | ESXi         | SR-IOV                      | —                       | VMware ESXi 9.1 | Ubuntu 24.04 |
| Instinct MI350X | KVM          | Passthrough                 | —                       | Ubuntu 24.04    | Ubuntu 24.04 |
| Instinct MI350X | KVM          | SR-IOV                      | GIM 9.1.0.K             | Ubuntu 24.04    | Ubuntu 24.04 |
| Instinct MI350X | KVM          | SR-IOV                      | GIM 9.1.0.K             | Ubuntu 24.04    | RHEL 9.6     |
| Instinct MI350X | ESXi         | SR-IOV                      | —                       | VMware ESXi 9.1 | Ubuntu 24.04 |
| Instinct MI325X | KVM          | SR-IOV                      | GIM 9.1.0.K             | Ubuntu 22.04    | Ubuntu 22.04 |
| Instinct MI300X | KVM          | Passthrough                 | —                       | Ubuntu 22.04    | Ubuntu 22.04 |
| Instinct MI300X | KVM          | SR-IOV                      | GIM 9.1.0.K             | Ubuntu 24.04    | Ubuntu 24.04 |
| Instinct MI300X | KVM          | SR-IOV                      | GIM 9.1.0.K             | Ubuntu 22.04    | Ubuntu 22.04 |
| Instinct MI210  | KVM          | Passthrough                 | —                       | RHEL 9.4        | Ubuntu 22.04 |
| Instinct MI210  | KVM          | SR-IOV                      | GIM 9.1.0.K             | RHEL 9.4        | Ubuntu 22.04 |
| Instinct MI210  | KVM          | SR-IOV                      | GIM 9.1.0.K             | RHEL 9.4        | RHEL 9.4     |

**Radeon**

| AMD GPU              | Hypervisor   | Virtualization technology   | Virtualization driver   | Host OS      | Guest OS     |
|----------------------|--------------|-----------------------------|-------------------------|--------------|--------------|
| Radeon AI PRO R9700S | KVM          | Passthrough                 | —                       | Ubuntu 24.04 | Ubuntu 24.04 |
| Radeon PRO V710      | KVM          | SR-IOV                      | GIM 9.1.0.K             | Ubuntu 24.04 | Ubuntu 24.04 |

## GPU partitioning support

The following compute partition and NUMA-per-socket (NPS) configurations are
available on AMD Instinct GPUs in bare metal deployments.

| Deployment   | Device                  | Compute partition mode   | Memory partition mode   |
|--------------|-------------------------|--------------------------|-------------------------|
| Bare metal   | Instinct MI355X, MI350X | CPX                      | NPS2                    |
| Bare metal   | Instinct MI355X, MI350X | DPX                      | NPS2                    |
| Bare metal   | Instinct MI355X, MI350X | QPX                      | NPS2                    |
| Bare metal   | Instinct MI350P         | CPX                      | NPS1                    |
| Bare metal   | Instinct MI350P         | SPX                      | NPS1                    |
| Bare metal   | Instinct MI300X         | CPX                      | NPS4                    |
| Bare metal   | Instinct MI300X         | DPX                      | NPS2                    |

The following configurations are available on AMD Instinct GPUs in SR-IOV
deployments. See [GPU virtualization support](#release-virtualization-support) for driver support
information.

| Deployment | Device | VFs per GPU | Compute partition mode | Memory partition mode |
|--------------|-------------------------|---------------|--------------------------|-------------------------|
| KVM SR-IOV | Instinct MI355X, MI350X | 1 | SPX (Default) | NPS1 |
| KVM SR-IOV | Instinct MI355X, MI350X | 1 or 8[\*] | CPX | NPS2 |
| KVM SR-IOV | Instinct MI355X, MI350X | 1 or 2[\*] | DPX | NPS2 |
| KVM SR-IOV | Instinct MI325X | 1 | SPX | NPS1 |
| KVM SR-IOV | Instinct MI300X | 1 | SPX | NPS1 |
| KVM SR-IOV | Instinct MI300X | 8 [\*] | CPX | NPS4 |

See the [AMD GPU partitioning](https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/gpu-partitioning/index.html) topic in the AMD GPU Driver documentation to learn more.

## AI ecosystem support

ROCm 7.14.0 provides optimized support for popular deep learning frameworks and AI inference engines. The following table lists supported frameworks and libraries, their compatible operating systems, and validated versions.

#### IMPORTANT
The following table is a general overview of supported frameworks and AI inference engines. Actual support might vary by AMD GPU or APU. Use the [Compatibility matrix](https://rocm.docs.amd.com/en/latest/compatibility/compatibility-matrix.html) to verify support for your specific setup.

## ROCm Core SDK components

The following table lists core tools and libraries included in the ROCm 7.14.0 release.

#### IMPORTANT
The following table is a general overview of ROCm Core SDK components. Actual support for these libraries and tools can vary by GPU and OS. Use the [Compatibility matrix](https://rocm.docs.amd.com/en/latest/compatibility/compatibility-matrix.html) to verify support for your specific setup.

| Component group               | Component name                              | Version         | Supported platforms                                                                                                 |    |
|-------------------------------|---------------------------------------------|-----------------|---------------------------------------------------------------------------------------------------------------------|----|
| Math and compute libraries    | hipBLAS                                     | 3.4.0 ⇒ 3.5.0   | Linux/Windows · Instinct/Radeon/Ryzen                                                                               |    |
| Math and compute libraries    | hipBLASLt                                   | 1.3.0 ⇒ 1.4.1   | Linux/Windows · Instinct/Radeon/Ryzen                                                                               |    |
| Math and compute libraries    | hipCUB                                      | 4.4.0 ⇒ 4.5.0   | Linux/Windows · Instinct/Radeon/Ryzen                                                                               |    |
| Math and compute libraries    | hipFFT                                      | 1.0.23 ⇒ 1.0.24 | Linux/Windows · Instinct/Radeon/Ryzen                                                                               |    |
| Math and compute libraries    | hipRAND                                     | 3.3.0 ⇒ 3.4.0   | Linux/Windows · Instinct/Radeon/Ryzen                                                                               |    |
| Math and compute libraries    | hipSOLVER                                   | 3.4.0 ⇒ 3.5.0   | Linux/Windows · Instinct/Radeon/Ryzen                                                                               |    |
| Math and compute libraries    | hipSPARSE                                   | 4.5.0 ⇒ 4.6.0   | Linux/Windows · Instinct/Radeon/Ryzen                                                                               |    |
| Math and compute libraries    | MIOpen                                      | 3.5.1 ⇒ 3.5.2   | Linux/Windows · Instinct/Radeon/Ryzen                                                                               |    |
| Math and compute libraries    | rocBLAS                                     | 5.4.0 ⇒ 5.5.0   | Linux/Windows · Instinct/Radeon/Ryzen                                                                               |    |
| Math and compute libraries    | rocFFT                                      | 1.0.37 ⇒ 1.0.38 | Linux/Windows · Instinct/Radeon/Ryzen                                                                               |    |
| Math and compute libraries    | rocRAND                                     | 4.4.0 ⇒ 4.5.0   | Linux/Windows · Instinct/Radeon/Ryzen                                                                               |    |
| Math and compute libraries    | rocSOLVER                                   | 3.34.0 ⇒ 3.35.0 | Linux/Windows · Instinct/Radeon/Ryzen                                                                               |    |
| Math and compute libraries    | rocSPARSE                                   | 4.6.0 ⇒ 4.7.0   | Linux/Windows · Instinct/Radeon/Ryzen                                                                               |    |
| Math and compute libraries    | rocPRIM                                     | 4.4.0 ⇒ 4.5.0   | Linux/Windows · Instinct/Radeon/Ryzen                                                                               |    |
| Math and compute libraries    | rocThrust                                   | 4.4.0 ⇒ 4.5.0   | Linux/Windows · Instinct/Radeon/Ryzen                                                                               |    |
| Math and compute libraries    | rocWMMA                                     | 2.2.0 ⇒ 2.2.1   | Linux/Windows · Instinct/Radeon/Ryzen                                                                               |    |
| Math and compute libraries    | Composable Kernel                           | 1.2.0           | Linux/Windows · Instinct/Radeon                                                                                     |    |
| Math and compute libraries    | hipSPARSELt                                 | 0.2.8 ⇒ 0.2.9   | Linux · Instinct (gfx950/gfx942)                                                                                    |    |
| Communication libraries       | RCCL                                        | 2.28.3 ⇒ 2.30.4 | Linux · Instinct/Radeon/Ryzen                                                                                       |    |
| Communication libraries       | rocSHMEM                                    | 3.4.0 ⇒ 3.5.0   | Linux · Instinct (gfx950/gfx942/gfx90a) / Radeon (gfx1201/gfx1200/gfx1100/gfx1101 (except Radeon PRO V710)/gfx1102) |    |
| Media libraries               | rocDecode                                   | 1.8.0           | Linux · Instinct/Radeon/Ryzen (gfx1150/gfx1151/gfx1152/gfx1153)                                                     |    |
| Media libraries               | rocJPEG                                     | 1.5.0 ⇒ 1.6.0   | Linux · Instinct/Radeon/Ryzen (gfx1150/gfx1151/gfx1152/gfx1153)                                                     |    |
| Storage libraries             | hipFile                                     | 0.3.0           | Linux · Instinct                                                                                                    |    |
| Runtimes and compilers        | HIP                                         | 7.13 ⇒ 7.14     | Linux/Windows · Instinct/Radeon/Ryzen                                                                               |    |
| Runtimes and compilers        | HIPIFY                                      | 7.14            | Linux/Windows · Instinct/Radeon/Ryzen                                                                               |    |
| Runtimes and compilers        | LLVM                                        | 23.0.0          | Linux/Windows · Instinct/Radeon/Ryzen                                                                               |    |
| Runtimes and compilers        | SPIRV-LLVM-Translator                       | 23.0.0          | Linux/Windows · Instinct/Radeon/Ryzen                                                                               |    |
| Runtimes and compilers        | ROCr Runtime                                | 1.21.0          | Linux · Instinct/Radeon/Ryzen                                                                                       |    |
| Profiling and debugging tools | ROCm Compute Profiler (rocprofiler-compute) | 3.6.0 ⇒ 3.7.0   | Linux · Instinct/Ryzen (gfx1150/gfx1151/gfx1152/gfx1153)                                                            |    |
| Profiling and debugging tools | ROCm Systems Profiler (rocprofiler-systems) | 1.6.0 ⇒ 1.7.0   | Linux · Instinct/Ryzen (gfx1150/gfx1151/gfx1152/gfx1153)                                                            |    |
| Profiling and debugging tools | ROCprofiler-SDK                             | 1.3.0 ⇒ 1.3.2   | Linux · Instinct/Radeon/Ryzen (gfx1150/gfx1151/gfx1152/gfx1153)                                                     |    |
| Profiling and debugging tools | ROCdbgapi                                   | 0.80.0          | Linux · Instinct/Radeon                                                                                             |    |
| Profiling and debugging tools | ROCm Debugger (ROCgdb)                      | 16.3            | Linux · Instinct/Radeon                                                                                             |    |
| Profiling and debugging tools | ROCR Debug Agent                            | 2.1.0           | Linux · Instinct/Radeon                                                                                             |    |
| Control and monitoring tools  | AMD SMI (BM)                                | 26.4.0 ⇒ 26.5.0 | Linux · Instinct/Radeon                                                                                             |    |
| Control and monitoring tools  | rocminfo                                    | 1.0.0           | Linux · Instinct/Radeon/Ryzen                                                                                       |    |
| Control and monitoring tools  | ROCm Data Center Tool (RDC)                 | 1.3.0 ⇒ 1.3.1   | Linux · Instinct                                                                                                    |    |

### ROCm component changelogs

The following sections describe key changes to ROCm Core SDK components.

#### **AMD SMI** (26.5.0)

##### Added

- NIC processor discovery and information API surface.
 - New C APIs: `amdsmi_get_nic_processor_handles()`, `amdsmi_get_nic_device_bdf()`, `amdsmi_get_nic_fw_info()`, `amdsmi_get_nic_port_statistics()`, and `amdsmi_get_nic_vendor_statistics()`.
 - `amdsmi_get_nic_processor_handles()` enumerates NIC processors by socket; the BDF, firmware, and port/vendor statistics getters are reserved and currently return `AMDSMI_STATUS_NOT_YET_IMPLEMENTED`.
- Exposed APU metrics through the CLI and Python interface.
 - `amd-smi metric` now surfaces APU-specific data under `--usage`, `--power`, `--clock`, `--temperature`, `--fan`, `--voltage`, and `--throttle` when APU metrics are available.
 - `amd-smi monitor` provides APU temperature and clock fallbacks when standard dGPU sensors report N/A.
 - On APU systems, the `--pcie`, `--ecc-blocks`, `--voltage-curve`, `--overdrive`, `--xgmi-err`, and `--energy` sections are not applicable and are omitted.
- The `--partition` flag to `amd-smi metric` for partition-scoped metrics.
 - The `-X`/`--partition` flag switches the temperature, clock, and usage categories to partition-level data sources; throttle metrics are already partition-aware.
 - Reuses the existing temperature/clock/usage section schema and adds partition-only AID/XCP/MID entries within it; socket-only fields with no partition equivalent report `N/A`.
 - When `--partition` is set with `--temperature`: adds MID and per-XCP/XCD temperatures.
 - When `--partition` is set with `--clock`: sources GFX/VCLK/DCLK/SOCCLK from partition metrics and adds per-AID and per-XCP clock entries with their limits.
 - When `--partition` is set with `--usage`: reports per-XCP GFX/JPEG/VCN activity.
- `--folder` support to `amd-smi ras --afid`.
 - `amd-smi ras --afid --folder <DIR>` decodes every `*.cper` in a directory and prints a `file_name | list of afids` table (or a JSON array under `--json`).
 - Records with no AFIDs show `-`; files that cannot be parsed show `decode failed`.
- Wrapped ESMI functions in `amdsmi_go_shim`.
 - Go callers can now access ESMI CPU functionality through the existing `amdsmi_go_shim` interface.
- GPU partitioning conceptual guide and usage examples.
 - New guide at `docs/conceptual/partition.md` covering accelerator partition modes (SPX/DPX/TPX/QPX/CPX), memory partition modes (NPS1/NPS2/NPS4/NPS8), API generations, device enumeration after partition, and BDF encoding.
 - New C++ example: `example/amd_smi_partition_example.cc`.
 - New Python example: `example/amd_smi_partition_example.py`.
- An alias for `amd-smi set -C/--compute-partition` as `amd-smi set --accelerator-partition`.
 - Compute and accelerator partitions are fundamentally the same, so users can now use `--accelerator-partition` to set the compute/accelerator partition.
- Input validation for CPU `set` commands.
 - Out-of-range values are now rejected with a clear error showing the valid range:
 - `--cpu-xgmi-link-width` (0-1)
 - `--cpu-gmi3-link-width` (0-2)
 - `--cpu-lclk-dpm-level` (0-3)
 - `--cpu-disable-apb` (0-3)
 - `--cpu-pwr-limit` values above the socket maximum are now reduced to the maximum and applied, with a warning.
- A compute partition memory allocation mode API.
 - New `amd-smi static --partition` output includes `COMPUTE_PARTITION_MEM_ALLOC_MODE` field.
 - New `amd-smi set --compute-partition-mem-alloc-mode [CAPPING|ALL]` to control memory allocation mode (requires sudo).
 - New APIs: `amdsmi_get_gpu_compute_partition_mem_alloc_mode()`, `amdsmi_set_gpu_compute_partition_mem_alloc_mode()`.
 - New enum: `amdsmi_compute_partition_mem_alloc_mode_t` (`AMDSMI_COMPUTE_PARTITION_MEM_ALLOC_CAPPING`, `AMDSMI_COMPUTE_PARTITION_MEM_ALLOC_ALL`).
 - Reads/writes sysfs: `/sys/class/drm/cardN/device/compute_partition_mem_alloc_mode`.
- `AMDSMI_LINK_TYPE_NUMA` and `AMDSMI_LINK_TYPE_XNUMA` to `amdsmi_link_type_t`.
 - Represent NIC-to-GPU links that cross different PCIe switches on the same CPU (NUMA) or across CPUs (XNUMA).
- PID-grouped process listing across GPUs.
 - `amd-smi process --sort-by-pid` and `amd-smi monitor --sort-by-pid` group output by PID, merging each PID’s per-GPU usage into one row.
 - New C and Python API `amdsmi_get_gpu_process_list_by_pid()`.

##### Changed

- Normalized JSON/CSV key casing in `amd-smi metric` clock and temperature sections.
 - The `uclk_aid`, `socclks_mid`, and temperature `xcd` keys are now lowercase (`aid_<N>`, `mid_<N>`, `xcp_<N>`) in JSON and CSV output, matching the existing `xcp_<N>` usage keys; they were previously uppercase (`AID_<N>`, `MID_<N>`, `XCP_<N>`).
 - Human-readable output is unchanged, since it uppercases all keys.
- Normalized JSON/CSV key casing in the `amd-smi topology` NIC-GPU access table.
 - The per-GPU columns are now lowercase (`gpu_<N>` for the BDF header row, `gpu_<N>_topo` for each NIC’s status row) in JSON and CSV output, matching the existing `gpu_<N>` keys in the GPU-to-GPU access matrix; they were previously uppercase (`GPU<N>`, `GPU<N>_Topo`).
 - Human-readable output is unchanged, since it uppercases all keys.
- Renamed “AINIC version” to “ionic version” in `amd-smi version` output.
 - The label now correctly reflects that it shows the ionic kernel driver version.

##### Removed

- Removed the non-functional `--decode` flag from `amd-smi ras`. Out-of-band CPER decoding is available via `amd-smi ras --afid --cper-file <path>` or `--afid --folder <DIR>`.
- Removed the unused `amdsmi_nic_link_type_t` enum from the public header. No API or struct referenced it; NIC link types are reported through `amdsmi_link_type_t`, which gains `AMDSMI_LINK_TYPE_NUMA` and `AMDSMI_LINK_TYPE_XNUMA` in this release.

##### Optimized

- Improved Python test runner behavior:
  - Added `-l`/`--list` flag to list all available tests and exit without running them.
  - Added shadow detection: if `amdsmi` loads from a path other than the resolved expected path (`AMDSMI_PATH`, `ROCM_HOME`, `ROCM_PATH`, or `/opt/rocm` default), tests exit early with a clear error message and remediation steps.
  - Non-root invocations now exit with code 1 immediately with a clear message instead of failing mid-test.

##### Resolved issues

- `amd-smi set --power-cap` rejecting the minimum allowed value.
 - The lower bound is now inclusive, so setting the power cap to the exact minimum of the reported range (for example, `210` when the range is 210-300W) succeeds instead of failing validation, matching the inclusive range shown in the error message.
- Corrected invalid AMD SMI status-code names in exception messages and documentation.
 - Some `AmdSmiLibraryException` messages and API documentation entries were misspelled; they now use the correct `AMDSMI_STATUS_*` names.
- A crash in `amdsmi_get_gpu_vram_vendor()` and made `amdsmi_get_gpu_vram_info()` resilient to DRM failures.
 - `amdsmi_get_gpu_vram_vendor()` now validates the output buffer and only writes it on success, fixing a null-pointer dereference on the not-supported path.
 - `amdsmi_get_gpu_vram_info()` now reads the VRAM vendor from sysfs first and treats the DRM ioctl (VRAM type/bit width/bandwidth) as best effort, so the vendor is still returned when the DRM path is unavailable.
- AMD GPU manufacturer name display in `amd-smi static --board`.
 - The CLI now displays the canonical vendor name `Advanced Micro Devices, Inc. [AMD/ATI]` when the board manufacturer name is reported as the raw AMD PCI vendor ID (`0x1002`) because the host `pci.ids` lookup is unavailable. The C and Python APIs continue to return the raw value unchanged.
 - Standardized the hardcoded AMD vendor string on the canonical `pci.ids` spelling (with the comma) so `VENDOR_NAME` and `MANUFACTURER_NAME` are consistent with `lspci`.
- `amd-smi ras --cper` / `amdsmi_get_gpu_cper_entries()` crash (`free(): invalid pointer` / `SIGABRT`) when `libamd_smi.so` is `LD_PRELOAD`-ed under a host with a different libstdc++ (for example, device-metrics-exporter / `gpuagent`).
- `amd-smi ras --cper` failing with `AMDSMI_STATUS_FILE_ERROR` on an empty CPER ring. An empty ring (no RAS records) now reports no CPER records; `amdsmi_get_gpu_cper_entries()` returns `AMDSMI_STATUS_SUCCESS` with `entry_count == 0`.
- `amdsmi_init()` aborting entirely when CPU/ESMI initialization fails.
 - `populate_amd_cpus()` treated an `esmi_init()` failure (non-AMD CPU, missing/unsupported energy or HSMP driver, or a CPU/SMU in a bad state) as fatal, causing all of `amdsmi_init()` to fail so GPU and NIC functionality became unusable. ESMI/CPU discovery is now non-fatal and is skipped on failure, mirroring the NIC discovery paths.
 - Removed an incorrect `static_cast<amdsmi_status_t>(esmi_init())` that conflated the unrelated `esmi_status_t` and `amdsmi_status_t` enums.
 - Added checks for the previously ignored return values of `get_nr_cpu_sockets()`, `get_nr_cpu_cores()`, and `get_nr_threads_per_core()`, plus a guard against a divide-by-zero when a misbehaving driver reports zero sockets or threads.
- `amd-smi static` hanging indefinitely on gfx1153 and gfx950.
 - Added a 60-second timeout to `amdsmi_init()` in the CLI so the process exits with a clear error message instead of hanging when the GPU driver is unresponsive.
 - Added `O_NONBLOCK` to DRM device open during initialization so `open()` returns immediately if the device is wedged.
- `amd-smi ras --afid --cper-file <file>` not showing AFIDs for correctable errors.
 - `aca_decode_corrected_error` was receiving the count of `uint32_t` elements where `decode_afid` expected the count of `uint64_t` elements, causing `decode_error_info` to return `NULL` for all non-standard section types.
- `amd-smi ras --cper --json` producing invalid JSON.
 - Multi-GPU runs emitted a separate JSON array per GPU instead of a single unified array, and `--follow` mode printed an empty `[]` every iteration when no new entries existed. Both are now consolidated into a single JSON document.
- Exposed `amdsmi_get_afids_from_cper` in the Python package.
 - The CPER AFID API was implemented but missing from `py-interface/__init__.py`, making it unavailable to Python callers using `from amdsmi import ...`.
- Python unittest scripts now append a GTest-style summary after test output.
 - All `*_test.py` and `unit_tests.py` scripts print a colored `[PASSED]`/`[SKIPPED]`/`[FAILED]` block after the standard unittest output. Colors are automatically suppressed when output is not a TTY (for example, file redirection, CI log capture).
- Corrected the documented unit of `amdsmi_frequencies_t::frequency`.
 - The struct comment claimed frequencies were in MHz, but `amdsmi_get_clk_freq()` returns them in Hz. The comment now reads “List of frequencies in Hz”.
 - Also removed the incorrect “in MHz” note from the `current` field, which is a frequency index, not a frequency value.
 - Updated the Python API reference to state the unit is Hz.
- Fabric telemetry APIs returning the wrong status on non-IFoE systems.
 - `amdsmi_alloc_fabric_telemetry()`, `amdsmi_get_fabric_telemetry_data()`, and `amdsmi_free_fabric_telemetry()` now return `AMDSMI_STATUS_NOT_SUPPORTED` on systems without fabric hardware, consistent with `amdsmi_get_gpu_fabric_info()`.
- `amd-smi static --clock` CSV and human-readable formatting to output frequency levels as strings instead of dictionary objects.

##### Upcoming changes

- `amdsmi_get_gpu_vram_vendor()` is deprecated in favor of `amdsmi_get_gpu_vram_info()` and will be removed in a future ROCm release. It now emits a `DeprecationWarning` from the Python interface and functions as a wrapper of `amdsmi_get_gpu_vram_info()`.
- See [AMD SMI deprecations](#amd-smi-deprecations).

#### **HIP** (7.14)

##### Added

- New HIP APIs:
  - Execution Context Management: Support for the following APIs for parity with corresponding CUDA APIs:
    - `hipDeviceGetDevResource` returns the device resource of a given type for a device.
    - `hipDevSmResourceSplitByCount` splits SM resources into groups with at least a minimum SM count.
    - `hipDevSmResourceSplit` splits SM resources into groups with configurable per-group parameters.
    - `hipDevResourceGenerateDesc` generates a resource descriptor from one or more device resources.
    - `hipGreenCtxCreate` creates a green context from a resource descriptor.
    - `hipExecutionCtxDestroy` destroys an execution context.
    - `hipDeviceGetExecutionCtx` returns the default execution context for a device.
    - `hipExecutionCtxStreamCreate` creates a stream on an execution context with specified flags and priority.
    - `hipExecutionCtxGetDevResource` returns the device resource of a given type for an execution context.
    - `hipExecutionCtxGetDevice` returns the device associated with an execution context.
    - `hipExecutionCtxGetId` returns a unique identifier for an execution context.
    - `hipStreamGetDevResource` returns the device resource of a given type for a stream.
    - `hipExecutionCtxRecordEvent` records an event on an execution context.
    - `hipExecutionCtxSynchronize` blocks until all work on an execution context has completed.
    - `hipExecutionCtxWaitEvent` makes an execution context wait on an event.
  - Module Management: Support for the following APIs for parity with corresponding CUDA APIs:
    - `hipLibraryGetGlobal` returns the device pointer and size of a `__device__` global defined in a `hipLibrary_t`. Mirrors `cudaLibraryGetGlobal` / `cuLibraryGetGlobal`.
    - `hipLibraryGetManaged` returns the host pointer and size of a `__managed__` variable defined in a `hipLibrary_t`. Mirrors `cudaLibraryGetManaged` / `cuLibraryGetManaged`.
  - Memory Management: Support for the following APIs for parity with corresponding CUDA APIs:
    - `hipMemDiscardBatchAsync` discards a batch of memory ranges asynchronously, allowing the runtime to reclaim resources. Mirrors `cudaMemDiscardBatchAsync`.
    - `hipDrvMemDiscardBatchAsync` driver API variant of `hipMemDiscardBatchAsync`, using `hipDeviceptr_t` pointers. Mirrors `cuMemDiscardBatchAsync`.
    - `hipMemDiscardAndPrefetchBatchAsync` combines discard and prefetch in a single call, enabling the runtime to optimize data movement. Mirrors `cudaMemDiscardAndPrefetchBatchAsync`.
    - `hipDrvMemDiscardAndPrefetchBatchAsync` driver API variant of `hipMemDiscardAndPrefetchBatchAsync`, using `hipDeviceptr_t` pointers. Mirrors `cuMemDiscardAndPrefetchBatchAsync`.
- Support for non-Host Transparent (nHT) fabric handles in HIP Virtual Memory Management (VMM) APIs, enabling efficient cross-device memory sharing over Infinity Fabric over Ethernet (IFoE). This allows peer devices to directly access shared memory without routing data through the host, reducing data movement overhead and improving performance for multi-GPU and distributed workloads.
- Introduced an exported no-op function `__hipOnError(void *err_info)`, invoked from `HIP_UPDATE_ERROR_STATE` when an API returns a non-success status, enabling debuggers to set breakpoints on a stable symbol. The symbol is exported on ELF (Executable and Linkable Format) platforms via a version script and on Windows via `amdhip.def`. The `err_info` parameter is a pointer to a struct containing the error code, name, and descriptive string.

##### Optimized

- Enhanced HIP graph replay performance for asynchronous memory allocations. HIP graph replay now reduces overhead for graphs that interleave asynchronous memory allocations with compute. Allocation nodes no longer block during replay — physical memory is reused across nodes instead of being mapped and unmapped on each launch, eliminating the gaps between kernels this pattern previously caused.
- Enhanced debug information for illegal memory access errors. In multi-node and multi-GPU environments, it can be difficult to identify the source of a fault. The HIP runtime now includes the hostname, GPU index, and kernel name in GPU fault error messages, improving issue identification and debugging.

##### Resolved issues

- Resolved an issue where graph allocations that escape their originating graph (that is, allocation nodes without a corresponding free node) failed to remain valid after the graph and its executable
  instance were destroyed. Allocations created via stream capture were not properly tracked and were incorrectly classified as reusable, leading to premature unmapping during `hipGraphExecDestroy` and resulting in memory faults on subsequent access.
- Resolved an issue where an error propagated from the `hipModuleGetFunction` API, causing behavior inconsistent with the corresponding CUDA API. The HIP runtime now suppresses this propagated error to align with expected behavior.
- Resolved an issue where a stream entering an invalid state during capture could not recover, even after calling `hipStreamEndCapture`. The stream failed to return to a clean (None) state,
  and subsequent calls to `hipStreamIsCapturing` continued to report an invalidated state, preventing reuse. This behavior is now aligned with CUDA semantics.
- Resolved a race condition in HIP graph nodes. The HIP runtime now correctly manages graph node IDs within each `GraphNode` constructor to ensure thread safety.
  This prevents duplicate ID assignment when multiple threads concurrently construct graph nodes (for example, during XLA command-buffer fusion).
  As a result, nodes are no longer silently dropped from dispatched packets, eliminating uninitialized output buffers and preventing out-of-bounds or corrupted values.
- Segmentation fault in the `hipMemRetainAllocationHandle` API when a pointer allocated with `hipMalloc` was passed. The HIP runtime now validates non-VMM allocations and returns an appropriate error instead.
- Resolved an issue where `__managed__` global variables were misclassified by the `hipPointerGetAttributes` API both before and after kernel access. This behavior has been corrected to align with CUDA semantics.
- Resolved an issue in the classic graph execution path (RunOneNode and RunNodes) where missing synchronization for child graph nodes caused data races and incorrect results when executing graphs with child nodes under multi-stream parallelism.
  The HIP runtime now properly synchronizes child graph nodes within the execution path.
- Issue in `hipGraphMemsetNode` that caused incorrect validation for flat allocations. For 2D `memsets`, the `userData` `width/height/depth` extents are only initialized by `hipMallocPitch` and `hipMalloc3D`;
  allocations from `hipMalloc` leave these fields unset, leading to spurious validation failures. The HIP runtime now skips `userData`-based checks when extents are zero and relies on `ihipMemset3D_validate`
  for accurate size validation. Additionally, the exec flag is propagated through `ihipGraphNodeSetParams` to ensure executable graph updates use the correct validation path.
- Deadlock caused by `hipMemMap` and `hipMemUnmap` operations on the null stream that could lead to hangs. The HIP runtime now implements proper synchronization to all devices with access to a mapped pointer before unmapping it.
- Resolved an issue where streams created within an execution context remained usable after the context was destroyed, which did not align with CUDA behavior. The HIP runtime now flags such streams as detached when their execution context is destroyed and returns `hipErrorStreamDetached` if they are subsequently used.

##### Known issues

- Kernels using `cooperative_groups::reduce()` with block dimensions whose .y or .z component is different from 1 may produce incorrect results or fail to launch.

#### **hipBLAS** (3.5.0)

##### Added

- The following APIs have been added:
  - `hipblasSetBatchAlphaStride()`
  - `hipblasGetBatchAlphaStride()`
  - `hipblasSetBatchBetaStride()`
  - `hipblasGetBatchBetaStride()`
  - `hipblasGetVersion()`
  - `hipblasGetProperty()`

##### Resolved issues

- Guarded x86-specific code and compiler options.

#### **hipBLASLt** (1.4.1)

##### Added

- Introduced a new API: `hipBLASLt_ext::isSolutionSupported()`. This API is used by the new hipBLASLt integration from rocBLAS to check if a given solution is supported for a specific GPU and problem type.

#### **hipCUB** (4.5.0)

##### Added

- Support for the gfx1250 architecture.

##### Upcoming changes

- CCCL 2.8.x compatibility is deprecated. hipCUB and rocThrust will be brought forward to CCCL 3.0.x compatibility in an upcoming release.

#### **hipFFT** (1.0.24)

##### Added

- Support for the gfx1250 architecture.

#### **hipRAND** (3.4.0)

##### Added

- Support for the gfx1250 architecture.

#### **hipSOLVER** (3.5.0)

##### Changed

- Minor internal changes.

#### **hipSPARSE** (4.6.0)

##### Added

- `hipsparseCreateBsr` and `hipsparseCreateConstBsr` to enable BSR format support in generic routines.
- BSR format support to `hipsparseSpMV` and `hipsparseSpMM`.

##### Resolved issues

- Issue where calling `hipsparseSpMV` multiple times with different `hipsparseOperation_t`, `hipsparseSpMVAlg_t`, or compute-datatypes using the same sparse matrix descriptor resulted in errors.

##### Upcoming changes

- The routines `hipsparseXcsrgeamNnz`, `hipsparseScsrgeam`, `hipsparseDcsrgeam`, `hipsparseCcsrgeam`, and `hipsparseZcsrgeam` have been deprecated and will be removed in a future release. Use `hipsparseScsrgeam2_bufferSizeExt`, `hipsparseDcsrgeam2_bufferSizeExt`, `hipsparseCcsrgeam2_bufferSizeExt`, `hipsparseZcsrgeam2_bufferSizeExt`, `hipsparseXcsrgeam2Nnz`, `hipsparseScsrgeam2`, `hipsparseDcsrgeam2`, `hipsparseCcsrgeam2`, and `hipsparseZcsrgeam2` instead.

#### **hipSPARSELt** (0.2.9)

##### Added

- Support for the following data type combinations for the LLVM target gfx942:
  - FP8_FNUZ(E4M3_FNUZ) inputs, F32 output, and F32 Matrix Core accumulation.
  - BF8_FNUZ(E5M2_FNUZ) inputs, F32 output, and F32 Matrix Core accumulation.

#### **MIOpen** (3.5.2)

##### Changed

- [Conv] Naive convolution solvers are now skipped by default during find when any non-naive solver succeeds across any algorithm. Set `MIOPEN_NAIVE_DISABLE_IF_ALT=0` to restore the previous behavior.

##### Resolved issues

- [RNN] RNN workspace tensor descriptor integer overflow.
- [Conv] Enabled grouped Composable Kernel (CK) xdlops fwd, bwd, and wrw convolution (2D and 3D) for tensors whose strides exceed the int32 range.
- [Conv] `miopenStatusInternalError` thrown by Find on depthwise NHWC grouped convolutions under `MIOPEN_FIND_MODE=NORMAL`.

#### **RCCL** (2.30.4)

##### Added

- Compatibility with NCCL 2.30.4, NCCL 2.29.7, and NCCL 2.28.9
- Proxytrace profiler plugin and core proxy-diagnostics hooks (`RCCL_PROXYTRACE`).
- `ncclBarrierSession` LSA validation for barrier sessions.
- Symmetric-memory ReduceScatter kernel (`RailA2A_LsaLD`) on gfx942/gfx950.
- Bias (accumulation) `AllReduce` on gfx1250.
- Optimized scale-up `ReduceScatter`, `AllGather`, and `AllToAll` kernels.
- ROCprofiler-SDK coverage for `ncclCommGrow` and `ncclCommGetUniqueId`.
- Auto-enabled P2P batching for gfx950 in combination with non-AINIC NICs.
- Display HIP/ROCm runtime versions in `NCCL_DEBUG` output.
- Detect ROCm version via core symlink for multi-architecture installs.
- Skip DDA IPC initialization for directMode and MNNVL topologies.
- Load versioned `libamd_smi` SONAME instead of an unversioned symlink.
- Pythonic API bindings under `bindings/nccl4py/` (RCCL fork of NVIDIA `nccl4py` v0.2.0). Provides Python access to RCCL collectives via Cython bindings, an on-disk `cuda.core` HIP shim for ROCm hosts without `cuda-bindings` / `cuda-core`, and RCCL-only collective wrappers (`ncclAllReduceWithBias`, `ncclAllToAllv`).
- RCCL examples to the repository.
- `RCCL host API` pull-in from NCCL 2.30.

##### Changed

- Enabled WarpSpeed auto mode for grow communicators.
- Refactored AllGather algorithm selection; hierarchical AllGather now enabled by default for multi-node.
- Swapped legacy `net_ib` with the `net_ib` implementation from NCCL 2.29.
- Skip per-warp channel LDS copy when `warpComm` is disabled.
- Hardened proxy RPC setup against malformed peer input.
- Changed the bootstrap AllGather to use the bidirectional ring (N/2 steps) by default on the socket OOB path. `NCCL_BOOTSTRAP_BIDIR_ALLGATHER` now defaults to `1`; set it to `0` to fall back to the unidirectional ring. The net OOB path (`NCCL_OOB_NET_ENABLE`) and its bidirectional variant (`NCCL_BOOTSTRAP_BIDIR_NET`) remain off by default.
- `NCCL_PXN_C2C` is kept default-off (`0`); upstream NCCL defaults it to `1` since 2.28. The C2C PXN routing path is currently not applicable on AMD hardware.

##### Removed

- Removed NPKit profiling support (build option `ENABLE_NPKIT`, headers, device and proxy instrumentation, install script flag `--npkit-enable`, and related documentation and tooling). Use the profiler plugin API for profiling instead.
- Removed Kernel COLLTRACE support, including the `COLLTRACE` build option, device-side collective trace buffers, debug kernel variants, and related install/CI wiring. The host latency profiler is unchanged.
- Removed legacy `ENABLE_PROFILING` device profiling support and the `PROFILE` build option. Use the profiler plugin API instead.

##### Optimized

- Tuned symmetric memory kernels.
- Parallelized communicator destruction across child processes to reduce teardown latency.

##### Resolved issues

- `ncclCommGrow` channel-count divergence causing incorrect collective routing.
- A `ncclCommGrow` hang when growing to an 8-rank single-node communicator.
- Symmetric LDS under-reservation in legacy (non-device-linker) builds.
- LL128 protocol correctness for gfx1250.
- XGMI topology mapping for multi-system (NPS) nodes.
- gfx950 collective hang caused by a tuner race condition.
- `net_ib_cast`: gate CTS offload path on per-connection state.
- `net_ib`: avoid flagging a non-fatal Isend CTS no-match as a fatal error.
- Acquire-tail polling for gfx950 P2P host staging.
- LDS overflow in device-linker builds.
- Symmetric memory correctness issues.
- `ncclCommFree` to free symmetric window objects automatically (NCCL 2.29.7 defect).
- DDA IPC initialization skip on architectures that do not run DDA.
- Static build (`BUILD_SHARED_LIBS=OFF`) failing with `install(EXPORT "rccl-targets" ...)` error when `fmt` is fetched via `FetchContent`. The `fmt-header-only` target is now scoped to the build interface and excluded from RCCL’s exported usage requirements.
- Proxy channel staging buffers ignoring the new GDR mode selection on HIP < 7.12 builds. The legacy `#else` branch in `sendProxyConnect` / `recvProxyConnect` now honors `resources->useDmaBuf`, so peermem-equipped hosts on older HIP no longer fall through to `hsa_amd_portable_export_dmabuf` when peermem was selected in `*ProxySetup`. Workaround for affected RCCL builds: `NCCL_DMABUF_ENABLE=0`.
- RCCL initialization failing (`Failed to find ROCm runtime library`) on runtime-only ROCm trees that ship no unversioned `libhsa-runtime64.so` developer symlink (e.g. TheRock multi-arch pip-wheel `/opt/rocm-less` deployments). RCCL no longer `dlopen`s the HSA runtime by name; instead it directly links `hsa-runtime64::hsa-runtime64` (already a hard transitive dependency via the HIP runtime) and binds `hsa_init`, `hsa_system_get_info`, `hsa_status_string`, and `hsa_amd_portable_export_dmabuf` to those symbols. The linker records `DT_NEEDED libhsa-runtime64.so.1` and resolves it through librccl’s existing RPATH, removing the SONAME version-string fragility and load-scope (`RTLD_LOCAL`) issues. The `RCCL_ROCR_PATH` override is no longer needed and has been removed.

##### Known issues

- Elastic-buffer support for GIN (multi-segment symmetric memory windows backed by a mix of device and CPU/`HOST_NUMA` memory, exposed through `NCCL_ELASTIC_BUFFER_REGISTER` and `NCCL_SYM_REUSE_SYSMEM_HANDLES`) was newly synced from upstream and compiles on ROCm, but is unverified on AMD hardware.

#### **RDC** (1.3.1)

##### Added

- 59 new telemetry fields to close the gap with Device Metrics Exporter (DME).
  - Energy: `RDC_FI_GPU_ENERGY` — total energy consumed via `amdsmi_get_energy_count()`.
  - Temperature: `RDC_FI_GPU_JUNCTION_TEMP` — dedicated junction/hotspot temperature field.
  - Clock ranges: `RDC_FI_GPU_CLOCK_MIN`, `RDC_FI_GPU_CLOCK_MAX` — min/max GPU clock frequencies. Additional clock types: `RDC_FI_SOC_CLOCK`, `RDC_FI_VCLK0`, `RDC_FI_DCLK0`.
  - Memory: `RDC_FI_GPU_MEMORY_FREE` (free VRAM), visible VRAM (`RDC_FI_GPU_VIS_VRAM_TOTAL/USED/FREE`), GTT memory (`RDC_FI_GPU_GTT_TOTAL/USED/FREE`).
  - PCIe: `RDC_FI_PCIE_SPEED`, `RDC_FI_PCIE_MAX_SPEED`, `RDC_FI_PCIE_REPLAY_ROLLOVER`, `RDC_FI_PCIE_BANDWIDTH_BIDIR` with sentinel value handling for APU platforms.
  - Instantaneous activity: `RDC_FI_GPU_GFX_BUSY_INST`, `RDC_FI_GPU_VCN_BUSY_INST`, `RDC_FI_GPU_JPEG_BUSY_INST` from `gpu_metrics.xcp_stats`.
  - ECC deferred errors: 19 per-block deferred error fields (`RDC_FI_ECC_*_DE`) plus `RDC_FI_ECC_DEFERRED_TOTAL`, reading `deferred_count` from `amdsmi_error_count_t`.
  - Violation/throttle metrics: 19 new `RDC_HEALTH_*` fields covering accumulated counts and percentages for processor hot, PPT power, socket/VR/HBM thermal, gfx clock host limits, and low utilization violations via `amdsmi_get_violation_status()`. Driver 1.8 XCP/XCC fields return NOT_SUPPORTED on older platforms.
- An automated DME-RDC metric sync check.
  - New script `tools/dme_rdc_metric_sync_check.py` parses DME’s protobuf metric definitions and compares against RDC field enums via a curated mapping file (`tools/dme_rdc_metric_mapping.json`).
  - New GitHub Action (`.github/workflows/rdc-dme-sync-check.yml`) runs weekly and on PRs touching metric definitions. Automatically creates GitHub issues when DME adds metrics not yet tracked in RDC.

##### Changed

- Bumped gRPC from 1.67.1 to 1.78.1. See [ROCm/TheRock#4172](https://github.com/ROCm/TheRock/pull/4172).

##### Removed

- Removed RVS integration. [RVS](https://github.com/ROCm/ROCmValidationSuite) is built independently of RDC and TheRock, so its integration has been disabled.
  - `BUILD_RVS` now defaults to `OFF` (#7116).

##### Resolved issues

- The `Failed to insert module: N3amd3rdc10RdcRVSLibE` error.

#### **rocBLAS** (5.5.0)

##### Added

- Per-batch `alpha`/`beta` support for Level 2 batched and strided-batched `gemv` via `rocblas_set_batch_alpha_stride` and `rocblas_set_batch_beta_stride` (device pointer mode).
- Per-batch `alpha` support for Level 2 batched and strided-batched `ger`, `geru`, and `gerc` via `rocblas_set_batch_alpha_stride` (device pointer mode).
- Per-batch `alpha` (scalar vector) API support for `axpy_batched`, `axpy_strided_batched`, and their `_ex` forms through `rocblas_set_batch_alpha_stride` when `rocblas_handle` is in `rocblas_pointer_mode_device`.
- Support custom build with CMake arguments `GPU_TARGET=amdgcnspirv` when using `BUILD_WITH_TENSILE=OFF`.

##### Resolved issues

- Incorrect results on gfx12 in `trsv`, `asum`, and `nrm2` with large `batch_count` exceeding 65536.
- `gemm` with very large `K` or inner product leading dimension for which element byte offset overflowed `int32`.
- `install.sh/rmake.py` builds when `CMAKE_GENERATOR=Ninja` is set.

#### **rocFFT** (1.0.38)

##### Added

- Generalized multi-device computations for transforms such that each length dimension is fully covered either in all the input field’s bricks or in all the output field’s bricks, regardless of the type and placement of the transform. Specifically for real transforms, the innermost length dimension must be fully covered in all the input (respectively, output) field’s bricks for real forward (respectively, inverse) transforms.
- Support for the gfx1250 architecture.

##### Optimized

- Improved performance of even-length real transforms with real lengths between 512 and 8192, extending to larger lengths (up to around 32768) on devices with more LDS.

##### Changed

- Modified the `rocfft_plan_get_work_buffer_size` and `rocfft_execution_info_set_work_buffer` functions to get and set work memory for the current HIP device.
  - Multi-device transforms can require work memory on any of the devices used for input or output bricks, and the current device set at plan creation. Users should loop over the set of devices used by the input/output of the transform and check the work memory requirements for each device.

##### Resolved issues

- Possible incorrect results for multi-dimensional real transforms with small lengths (for example, smaller than 128) along the two fastest-varying dimensions.

#### **ROCgdb** (16.3)

##### Added

- Improved core dumping speed for AMD GPU programs with the `gcore` command,
  particularly for kernels that use small amounts of VRAM.
- A new `catch hiperr` command that stops the inferior when a HIP API
  call returns an error. The convenience variable `$_hiperr` holds
  the error code at the catchpoint.

#### **rocJPEG** (1.6.0)

##### Added

- A logging mechanism for core APIs that can be controlled by setting the `ROCJPEG_LOG_LEVEL` environment variable.

#### **ROCm Compute Profiler** (3.7.0)

##### Added

- `--bench-only` profile mode option to run the roofline microbenchmark standalone (without profiling an application or collecting performance counters). No application run is required. Useful for regenerating `roofline.csv` in an existing workload directory or running the microbenchmark on systems where only HIP is available but ROCprofiler-SDK is not.
- LDS arithmetic intensity as a roofline plot point and analysis database field.
- Backward compatibility for live attach mode to work with older ROCm 7.x.x releases.
- Support for GPU metrics on gfx1150 and gfx1152 hardware.
- Roofline benchmarking support for gfx1150 and gfx1152 hardware.
- Operator statistics and per-operator summary table in the analysis output of torch operator profiling, including the following statistics for every torch operator and its children:
  - Number of invocations.
  - Number of kernel dispatches.
  - Min/Max/Mean and Total duration of kernel dispatches.

##### Changed

- Moved `--gui` and `--tui` analyze options to experimental status. These features now require the `--experimental` flag to be enabled (for example, `rocprof-compute analyze --experimental --gui`).
- `--output-format csv` in analyze mode now uses the database analysis workflow and produces one CSV per analysis view. Requires `--format-rocprof-output rocpd` and no longer prints the report to the terminal (matching `db` format).
- Changed ratio metric aggregation from `AVG(A/B)` (arithmetic mean of per-dispatch ratios) to `SUM(A)/SUM(B)` (ratio of totals) across all analysis YAML configurations and all GPU architectures. `SUM(A)/SUM(B)` is a weighted average where each dispatch contributes proportionally to its denominator magnitude (duration, access count, cycle count). Single-dispatch workloads are unaffected (mathematically identical). Multi-dispatch workloads with different kernels or varying durations will see corrected values.
- `--torch-trace` now captures backward-pass and nested operators that were previously missed or misattributed. The first run builds and caches a helper under `~/.cache/rocprofiler-compute/`, so it takes longer than later runs.
- Profile workload output folder name for Strix Halo series (gfx1151) changed from `strix_halo` to `rdna35_halo`.
- Unified accumulator handling across profile and analyze so each `_ACCUM`-suffixed counter is preserved instead of collapsing to `SQ_ACCUM_PREV_HIRES`.
- Reworded the N/A metric-evaluation warning to “divide-by-zero or empty counter data” (the prior “missing counter data” message could only fire for non-missing causes).
- PC sampling in profile mode now opts in via the `--experimental --pc-sampling` option. Explicit `-b 21` / `--block 21` is no longer accepted on its own.
- PC sampling profiling now emits only `ps_file_results.json`. The per-sample, kernel-trace, and agent-info CSV artifacts are no longer produced or consumed by analysis.
- PC sampling analysis without `-k` now shows the full per-instruction table across all kernels (with a `Kernel_Name` column), identical in schema to the single-kernel view, instead of a collapsed source-line summary.
- `--pc-sampling-interval` now defaults to a method-appropriate value (512 microseconds for `host_trap`, 1048576 cycles for `stochastic`). Stochastic intervals are validated to be a power of 2 and at least 65536; previously invalid values were passed through silently.

##### Removed

- `--path` and `--subpath` options have been removed from profile mode. Use `--output-directory` instead.
- Removed redundant `if (X != 0) else None` divide-by-zero guards from metric equations across all analysis YAML configurations. Division by zero is already handled by the metric evaluation engine, which returns `"N/A"` for `inf` and `NaN` results.

##### Optimized

- Flattened the analyze-mode PMC dataframe to a single-index frame.
- Eliminated “missing counter” warnings during analyze when profile-mode `-b` was used. Analyze now skips metrics outside the selected blocks.

##### Resolved issues

- Roofline panel L1/L2 bandwidth and arithmetic intensity on gfx942 and gfx950 now use the correct 128B cache line, matching the values reported in the Speed-of-Light and vL1D/L2 cache panels for the same run. Bandwidth values on these architectures are 2x and AI values are 0.5x compared to prior releases.
- “ROCPROF_OUTPUT_PATH environment variable must be set” crash that aborted profiling when `ROCPROF_OUTPUT_PATH` was unset or empty (observed when profiling shell-script targets such as `rocprof-compute profile -o /tmp/out -- bash run.sh`). The collector now silently falls back to a documented default instead of aborting.
- `inf` display for metrics with zero-denominator counters (e.g., L2-Fabric Write Latency when no write requests are issued). The metric evaluation path now catches `inf` scalar results and returns `"N/A"`, consistent with existing `NaN` handling.
- Kernels with missing counter data after iteration multiplexing imputation are now excluded from metrics calculations. A warning at analysis time lists the affected kernels. Their execution times remain visible in Top Stats.
- Empirical roofline benchmark to correctly produce double the Matrix BF16 Gflop/s on gfx90a (AMD Instinct MI200 Series) GPUs.
- PC sampling collection now runs when requested via the `pc_sampling` block alias (`--block pc_sampling`), instead of being silently skipped.

##### Upcoming changes

- Roofline support for gfx1153 devices.

##### Known issues

- On gfx1151, `TCP_REQ_sum` is zero in single-pass counter collection, so the related `GL0` metrics always report zero. This issue will be fixed in a future release.
- On gfx1151, `$max_mclk` is not automatically populated in sysinfo, so the related bandwidth metrics may be incorrect. Use `amd-smi` to obtain the maximum memory clock and provide it via `--specs-correction`.
- In analyze mode, `--nodes` is not suitable for multi-rank analysis. Use `--path` with the rank-specific path (for example, `--path workload/1`) instead of `--path workload --nodes 1`.

#### **ROCm Systems Profiler** (1.7.0)

##### Added

- `--output-format` flag for `rocprof-sys-run` and `rocprof-sys-sample` to select
 output format(s) in a single, intuitive option: `proto` (Perfetto), `rocpd`
 (RocPD database), and `json` / `text` (Timemory profile; `txt` aliases `text`).
 Tokens are space- or comma-separated and authoritative — only the listed
 formats are produced. The existing `--trace`, `--profile`, `--flat-profile`,
 and `--profile-format` flags and their environment variables remain available,
 but cannot be combined with `--output-format` on the same command line.
- Unified-memory profiling reports (`unified_memory.txt` and
 `unified_memory.json`) summarizing KFD page fault and page migration events,
 including per-GPU counts, trigger breakdown (`gpu_page_fault`,
 `cpu_page_fault`, `prefetch`), and Host-to-Device/Device-to-Host migration
 bandwidth. Enable with `ROCPROFSYS_USE_UNIFIED_MEMORY_PROFILING=ON`; requires
 `HSA_XNACK=1` on an XNACK-capable AMD GPU and ROCm 7.13 or later. For standalone ROCprofiler-SDK installations, ROCprofiler-SDK 1.2.2 or later. The required KFD tracing domains are enabled automatically.
- Dedicated `ROCPROFSYS_UNIFIED_MEMORY_OUTPUT_PATH` setting for routing
 unified-memory profiling reports to an explicit output directory.
- MPI-rank-based console output filtering features controlled with CLI arguments:
 `--rank-filter-logs` and `--rank-filter-id`.
- GPU Hardware Performance Counter (PMC) sampling via the ROCprofiler-SDK device
 counting service. Periodic per-GPU hardware counters are collected alongside
 existing PMC sources and exposed in both Perfetto and RocPD outputs. Specify
 counters with `ROCPROFSYS_GPU_PERF_COUNTERS` (comma-separated; suffix
 `:device=N` to target a specific GPU). Requires ROCprofiler-SDK 0.6.0 or
 later (ROCm 6.4.0+).
- GPU graphics and memory clock frequency metrics (`gfx_clock`, `mem_clock`) via
 AMD SMI, exposing `current_gfxclk` and `current_uclk` in MHz as PMC samples.
 Configure via `ROCPROFSYS_AMD_SMI_METRICS=gfx_clock,mem_clock`.
- Progress bars during trace cache post-processing: perfetto generation
 (`sequential dispatch`) shows one bar per buffered_storage file in turn; rocpd
 generation (`multithreaded dispatch`) shows a single aggregate bar accumulating
 updates from all worker threads.
- Per-stream Perfetto tracks (`HIP Activity Stream {N}`) for kernel dispatch,
 scratch memory, and memory copy events in the trace-cache path, matching the
 buffered tracing behavior. Controlled via `ROCPROFSYS_ROCM_GROUP_BY_QUEUE`
 (default: `false` — group by HIP stream).
- `--list-domains` and `--list-operations <domain>` to `rocprof-sys-avail`.
 These new options allow the user to query more information about available
 ROCm domains (used in `ROCPROFSYS_ROCM_DOMAINS`) and their operations.
- `rocprofsys_push_trace_with_args`, a public API for pushing a user trace region
 with a pre-serialized argument string attached. The arguments are recorded in cached
 tracing mode (the default); in legacy tracing (`ROCPROFSYS_TRACE_CACHED=OFF`) they are
 ignored.

##### Changed

- Split PMC AMD SMI, ROCprofiler-SDK, and procfs wrappers into standalone
  internal backend targets under `source/lib/backends`, replacing the old
  PMC `drivers` layout.
- Removed Boost as a Dyninst dependency by replacing Boost usage with in-tree
  `dyncompat` shims and C++17 standard library equivalents; Bundled Dyninst now
  requires GCC ≥ 10.
- The `trace-openmp` configuration preset no longer includes `HSA_API`,
  by default.
- `rocprof-sys-sample` — Aligned flags with `rocprof-sys-run`. Renamed `--freq`,
  `--cputime` and `--realtime` to `--sampling-freq`, `--sampling-cputime` and
  `--sampling-realtime`, respectively. Old flags are still handled as a part of
  backward compatibility.
- Allowed presets to use `--gpus`/`--cpu`/`--ai-nics` flags without
  `--device`/`--host` flags.
- Minimum required C++ standard raised from C++17 to C++20. timemory now builds
  against the `rocprofiler-systems-cppstd20` branch and spdlog was bumped to
  v1.17.0 (bundled fmt v12).
- Supported environment variables for rank detection: removed `MPI_RANK` and
  `MPI_LOCALRANKID`, added `PMI_RANK` and `SLURM_PROCID`.

##### Resolved issues

- An issue affecting the ElfUtils build on GCC 15.
- Output directory of `rocpd` files was not unique when re-attaching to the same process
  with `rocprof-sys-attach`. Now, each session will have a unique output folder.
- CPU-related counters (like CPU frequency) were missing from `rocpd` output.
- The “group-by-queue” option was not handled correctly in the Perfetto generator.
- The visualization of GPU counters made it look like there was activity
  between kernel dispatches.
- A hang due to mismatched versions of `binutils` between system and bundled
  versions. Ensure that the vendored version of `binutils`’s symbols are hidden.
- The ASAN build on TheRock.
- An issue that could cause certain events to appear in trace, when they should
  have been excluded due to roctx region filtering.
- A CMake issue that caused the wrong version of `elfutils` to be linked when
  building for TheRock. The system version of `elfutils` was used, rather than
  the vendored version causing package install failures.
- Documentation and internal config handling that referenced the non-existent
  `ROCPROFSYS_USE_TRACE`. The Perfetto tracing backend is controlled by
  `ROCPROFSYS_TRACE`; setting `ROCPROFSYS_USE_TRACE` had no effect.
- A pre-main `rocprof-sys-run` `SIGSEGV` in `rocprofiler_configure()` when
  profiling OpenMPI GPU-aware MPI workloads.

##### Known issues

- A push/pop trace count imbalance can occur for workloads that instrument runtime
  internals such as OMPT. When pushes exceed pops, rocprof-sys completes
  finalization, emits a warning, and omits any still-open trace regions from the
  generated trace output.

#### **rocPRIM** (4.5.0)

##### Added

- `generate_resource_spec.cpp` to the test directory, built as a new target by CMake. It generates the resource spec file required by CTest when running tests in parallel.
- Support for the gfx1250 architecture.
- A parallel `device_topk`, which finds the largest/smallest K elements from an input array of keys.

##### Changed

- Updated the documentation on how to run rocPrim tests on multiple GPUs in parallel.

##### Removed

- Removed the `GenerateResourceSpec.cmake` script - it is replaced by the added `generate_resource_spec.cpp` code above.

#### **ROCprofiler-SDK** (1.3.2)

##### Added

**API:**

- Streaming Performance Monitor (SPM) counter collection support (beta):
  - New experimental API in `rocprofiler-sdk/experimental/spm.h`.
  - GPU-timestamped counter values alongside kernel dispatch information.
- `spm_support` along with reserved padding to `rocprofiler_counter_info_v1_t`.

**rocprofv3 (CLI):**

- SPM counter collection support in `rocprofv3` (beta):
 - `--spm <counter>` flag to specify counters for SPM collection.
 - `--spm-sample-interval` and `--spm-sample-interval-unit` parameters to configure sampling rate.
 - `--spm-beta-enabled` flag or the `ROCPROFILER_SPM_BETA_ENABLED` environment variable to opt in to the beta SPM feature via `rocprofv3`. For API-based usage, set `ROCPROFILER_SPM_BETA_ENABLED`.
 - `--spm-config` option in `rocprofv3-avail` to list available SPM configurations.
- JSON and rocpd output format support for SPM.

**Documentation:**

- [SPM API reference guide](https://rocm.docs.amd.com/projects/rocprofiler-sdk/en/docs-7.14.0/api-reference/spm.html).
- [SPM usage guide for rocprofv3](https://rocm.docs.amd.com/projects/rocprofiler-sdk/en/docs-7.14.0/how-to/using-spm.html).
- `--spm-config` documentation to [`rocprofv3-avail` usage guide](https://rocm.docs.amd.com/projects/rocprofiler-sdk/en/develop/how-to/using-rocprofv3-avail.html).

##### Changed

- Bumped rocpd schema to version 3.0.1 which supports NIC agent types.

#### **rocRAND** (4.5.0)

##### Added

- Support for the gfx1250 architecture.

#### **rocSHMEM** (3.5.0)

##### Added

- New APIs:
  - `rocshmem_align`
  - `rocshmem_calloc`
  - `rocshmem_buffer_unregister_all`
  - `rocshmem_buffer_register/unregister` for GDA backend
  - `rocshmem_reduce_on_stream`
  - `rocshmem_team_split_2D`
- Tile-granular RMA operations for the IPC backend.
- Host-initiated RMA operations in the IPC backend for the non-MPI bootstrapping path.
- Team creation using non-contiguous parent teams in the IPC backend.
- Python bindings for memory-management APIs.
- Python bindings coverage for team APIs.
- Support for GPU-initiated operations using the SDMA engines.
- ASAN build support.

##### Changed

- Changed default `ROCSHMEM_DEBUG_LEVEL` from `WARN` to `ERROR`.
- Performance optimizations:
  - Separated put/get memcpy primitives to apply correct cache coherence semantics and fences.
  - Use constmem for backend variables and provider muxing.
  - Updated O(1) IPC availability check using pattern detection.

#### **rocSOLVER** (3.35.0)

##### Added

- Support for the gfx1250 architecture.

##### Optimized

- Refined `potf2_run_small` dispatch by `BS2` to avoid over-generating specialized kernels while preserving runtime bounds checks on `nb`.

##### Resolved issues

- An out-of-bounds read in `bdsqr_lower2upper`.
- An invalid kernel launch in the small-matrix LU factorization (GETF2/GETRF) for large batch counts.
- A synchronization issue in GETRI and TRTRI on wave 32 architectures.

#### **rocSPARSE** (4.7.0)

##### Added

- `rocsparse_spildlt0` routine for incomplete LDL’ factorization with zero fill-in (ILDLT(0)) for symmetric (real) or Hermitian (complex) sparse matrices in CSR format, with strided batched computations enabled.

##### Known issues

- The HIP graph capture/launch path for the factorization routines `bsric0`, `bsrilu0`, `csric0` and `csrilu0` can fail with `hipErrorOutOfMemory` at `hipGraphLaunch` on memory-constrained GPUs such as the gfx110X family. The corresponding `graph_test` cases are marked `known_bug` and excluded from the standard test suites until the fix lands.

##### Upcoming changes

- Deprecated the `rocsparse_indextype_u16` index type and will be removed in a future release. Use `rocsparse_indextype_i32` or `rocsparse_indextype_i64` instead.

#### **rocThrust** (4.5.0)

##### Added

- Support for the gfx1250 architecture.
- One-time runtime warning for hipstdpar algorithms running on GPUs that support xnack when `__HIPSTDPAR_INTERPOSE_ALLOC__` or `__HIPSTDPAR_INTERPOSE_ALLOC_V1__` are not enabled and xnack is off.

##### Upcoming changes

- CCCL 2.8.x compatibility is deprecated. hipCUB and rocThrust will be brought forward to CCCL 3.0.x compatibility in an upcoming version.

#### **rocWMMA** (2.2.1)

##### Added

- Support for the gfx1250 architecture.

## ROCm known issues

ROCm known issues are noted on  [GitHub](https://github.com/ROCm/ROCm/labels/Verified%20Issue). These issues will be fixed in a future ROCm release. For known issues related to individual components, review the [ROCm component changelogs]().

### PyTorch might display a warning when libnuma is not installed

PyTorch might display a warning when importing on Linux if the system libnuma package is not installed on some Radeon graphics products, such as Radeon AI PRO R9600D. As a workaround, install the system libnuma package or configure the library path to use the ROCm-bundled NUMA libraries. See [GitHub issue #6485](https://github.com/ROCm/ROCm/issues/6485).

### Significantly longer LLM warmup times on some Radeon GPUs

Significantly longer warmup times might be observed in some large language model inference workloads on AMD Radeon GPUs using vLLM versions v0.21.0 through v0.25.0. As a workaround, use a vLLM release earlier than v0.21.0 or upgrade to vLLM v0.26.0 or later, which includes a fix for this issue. See [GitHub issue #6486](https://github.com/ROCm/ROCm/issues/6486).

### SGLang default settings and some models might cause failures on Radeon GPUs

ROCm 7.14 introduces initial SGLang support for AMD Radeon GPUs. Radeon GPU users should disable AITER and unset `SGLANG_ROCM_FUSED_DECODE_MLA`, as both are enabled by default in the SGLang Docker image and might cause some workloads to fail:

```bash
export SGLANG_USE_AITER=false
export SGLANG_ROCM_FUSED_DECODE_MLA=false
```

Additionally, some models might not function correctly on Radeon GPUs, including certain Mixture-of-Experts (MoE) models (such as GPT-OSS-20B and MiniMax-M2.7) and Qwen3-ASR models. Users experiencing these issues are recommended to use the latest upstream SGLang versions, which will include the necessary fixes once they are merged. See the [SGLang environment variables reference](https://docs.sglang.io/docs/references/environment_variables#environment-variables) for more details. See [GitHub issue #6487](https://github.com/ROCm/ROCm/issues/6487).

### Lower-than-expected LLM inference performance on RDNA3 Radeon GPUs and Ryzen AI Max / Max+ Series Processors

Lower-than-expected performance might be observed in some large language model inference workloads, including vLLM FP16 decode workloads with batch sizes of 8 or greater, on AMD Radeon RX 7900 Series Graphics, AMD Radeon RX 7800 XT Graphics, and AMD Ryzen AI Max / Max+ Series Processors when using PyTorch versions earlier than 2.14. As a workaround, set the `TORCH_BLAS_PREFER_HIPBLASLT=1` environment variable to use the hipBLASLt backend. This setting becomes the default for these architectures in PyTorch 2.14. See [GitHub issue #6488](https://github.com/ROCm/ROCm/issues/6488).

### ROCprofiler-SDK SPM sessions can remain in a stale state after abrupt termination

If a Streaming Performance Monitors (SPM) session is terminated abruptly (for example, with `Ctrl+C`), KFD-side SPM resources might not be released cleanly. When this happens, the KFD-side SPM resources can remain in a stale state, which might cause subsequent SPM profiling sessions to hang or fail to start with the error `Unable to acquire KFD thread: 4096`. To recover, if the profiling process is still running, terminate it manually. If the error persists, a system reboot is currently required to restore the GPU to a usable state for SPM profiling. This issue is under active investigation for a fix. See [GitHub issue #6489](https://github.com/ROCm/ROCm/issues/6489).

### ROCm Compute Profiler might report inflated Avg values with per_kernel normalization

When using ROCm Compute Profiler with `per_kernel` normalization, the reported Avg value for certain normalized metrics might be incorrectly inflated and can exceed the corresponding Min and Max values. This issue affects analysis results only. As a workaround, use an alternative normalization unit (`-n`/`--normal-unit`) until a fix is available. See [GitHub issue #6490](https://github.com/ROCm/ROCm/issues/6490).

### rocALUTION and hipTensor have no dedicated HPC Expansion tarball

The `amdrocm-hpc` meta-package installs rocALUTION and hipTensor, but there is no dedicated HPC Expansion tarball for tarball-based installations. The standard ROCm tarballs include both libraries. See [GitHub issue #6491](https://github.com/ROCm/ROCm/issues/6491).

### HIP SPIR-V kernels might segfault on first launch

HIP kernels compiled with the SPIR-V target (`--offload-arch=amdgcnspirv`) might segfault on first kernel launch at `hipLaunchKernel`. The failure affects both library-level workloads such as rocBLAS and standalone HIP applications built against the SPIR-V offload bundle. Applications compiled for a native GPU architecture target are not affected. As a workaround, compile using a direct GPU architecture target instead of `--offload-arch=amdgcnspirv`. See [GitHub issue #6492](https://github.com/ROCm/ROCm/issues/6492).

### RCCL might show degraded performance on multi-node configurations

RCCL operations with message sizes in the 64 MB to 512 MB range might show suboptimal performance on multi-node, multi-threaded configurations. This issue affects the packaged binary distribution and might severely impact production workloads. Known affected workloads include Llama 3 405B and JAX stack; additional workloads might also be affected. Single-node (scale-up) configurations are not affected.
To work around this issue, recompile RCCL from the ROCm 7.14 source with fault injection disabled (see Building and installing RCCL). You can either set the option in the RCCL CMake file:

```cmake
option(FAULT_INJECTION         "Enable fault injection"           OFF)
```

Alternatively, add the following CMake flag during compilation:

```text
-DFAULT_INJECTION=OFF
```

See [GitHub issue #6493](https://github.com/ROCm/ROCm/issues/6493).

### AMD SMI NIC telemetry supports Pollara 400 adapters only

In ROCm 7.14.0, AMD SMI NIC telemetry only supports AMD AI NIC Pollara 400 adapters. Broadcom NIC support is planned for a future release. See [GitHub issue #6497](https://github.com/ROCm/ROCm/issues/6497).

## ROCm resolved issues

The following notable issues have been fixed in ROCm 7.14.0.

### ROCm Compute Profiler failed when profiling bash scripts or commands

Previously, running a bash script or command as a target for ROCm Compute Profiler failed because bash overwrote the required environment variables.

### LLVM-based compilers failed when compiling half-precision vector operations

Previously, LLVM-based compilers failed, returning the `Failed to find subregs!` error message in `SIInstrInfo::copyPhysReg`, when compiling half-precision vector operations with optimization enabled at levels `-O1` to `-O3`.

### hipBLAS test suites returned non-zero exit codes on Windows

Previously, when using hipBLAS on Windows, the test suites returned non-zero exit codes even when all mathematical correctness tests passed, blocking automated testing workflows.

### Illegal memory address error when using placement new with device function returns

Previously, HIP kernels that used placement new to construct objects in `hipMalloc`-allocated device memory crashed with a `hipErrorIllegalAddress` error when a `__device__` function return value was passed as the constructor argument for non-trivially copyable types.

### GPU kernels failed to launch in ASan builds with large thread counts

Previously, when building GPU libraries with ASan enabled, kernels configured with large thread counts failed to launch with an `HSA_STATUS_ERROR_INVALID_ISA` error.

### ASan prevented multi-architecture HIP binary builds from launching

Previously, HIP applications built with ASan enabled and targeting multiple GPU architectures failed to launch with `RuntimeError: .hipFatBinSegment size N is not a multiple of wrapper size (24)` and `RuntimeError: Unexpected magic 0x00000000 at wrapper i` error messages.

### ROCm Systems Profiler overwrote ROCPD output after process re-attachment

Previously, when using `rocprof-sys-attach` to re-attach to a previously profiled process, the ROCPD output database files (`.db`) were written to the initial session’s output directory instead of a new timestamped directory.

### hipCUB DeviceMerge large-size stress test failed with out-of-memory error on gfx1150

Previously, on gfx1150 APUs, the hipCUB DeviceMerge large-size stress test (`MergeLargeSizeIterators`) failed with an out-of-memory error when running ROCm 7.12.0. Standard DeviceMerge test cases were not affected.

### HIP kernel launch limit caused failures for some models

Previously, with PyTorch 2.10, some models hit the HIP kernel launch limit of 2³² kernel launches within a single process, causing HIP kernel launch errors. One known affected model was `black-forest-labs/flux`.

### Non-deterministic GPU memory faults when passing large data structures on MI300X

Previously, applications running on AMD Instinct MI300X GPUs that passed large, complex data structures between device functions using scratch memory encountered non-deterministic GPU memory access faults and became unresponsive when compiler optimizations minimized the number of copy operations.

## ROCm upcoming changes

Future releases will add support for:

* Additional ROCm Core SDK components.
* Domain-specific expansion toolkits (data science, life sciences, finance, simulation, and other HPC domains).
* More AMD hardware support.

### AMD SMI deprecations

The AMD SMI library will deprecate the following APIs. Certain APIs will be
deprecated with or without a replacement; see the following tables for details.
We suggest updating your code to use the replacement identifiers before the
targeted removal releases.

#### Planned removal in the next major release

The following APIs, defines, enums, and struct fields are deprecated and
scheduled for removal in the next major release.

##### APIs

| Deprecated                       | Replacement                                           |
|----------------------------------|-------------------------------------------------------|
| `amdsmi_get_cpusocket_handles()` | No replacement; functionality removed                 |
| `amdsmi_get_gpu_vram_vendor()`   | `amdsmi_get_gpu_vram_info()`                          |
| `amdsmi_gpu_driver_reload()`     | No replacement; functionality removed                 |
| `amdsmi_get_xgmi_plpd()`         | Python: use the `policy` attribute instead of `plpds` |
| `amdsmi_set_gpu_clk_range()`     | `amdsmi_set_gpu_clk_limit()`                          |

##### Types

| Deprecated                 | Replacement                         |
|----------------------------|-------------------------------------|
| `amdsmi_fabric_info_ver_t` | Moved inside `amdsmi_fabric_info_t` |
| `amdsmi_nic_fw_t`          | `amdsmi_nic_fw_entry_t`             |

##### Defines and enums

| Deprecated                        | Replacement                                |
|-----------------------------------|--------------------------------------------|
| `MAX_NUMBER_OF_AFIDS_PER_RECORD`  | `AMDSMI_MAX_NUMBER_OF_AFIDS_PER_RECORD`    |
| `MAX_SVI3_RAIL_INDEX`             | `AMDSMI_MAX_SVI3_RAIL_INDEX`               |
| `MAX_SVI3_RAIL_SELECTION`         | `AMDSMI_MAX_SVI3_RAIL_SELECTION`           |
| `POWER_EFFICIENCY_MODE_4`         | `AMDSMI_POWER_EFFICIENCY_MODE_4`           |
| `POWER_EFFICIENCY_MODE_5`         | `AMDSMI_POWER_EFFICIENCY_MODE_5`           |
| `CENTRIGRADE_TO_MILLI_CENTIGRADE` | No replacement; constant removed           |
| `_AMDSMI_MAX_STRING_LENGTH`       | No replacement; private symbol, do not use |
| `_AMDSMI_STRING_LENGTH`           | No replacement; private symbol, do not use |

##### `amdsmi_gpu_metrics_t` field type widening

The following fields in `amdsmi_gpu_metrics_t` will change from `uint32_t` to `uint64_t` to support next generation AMD Instinct counter ranges:

* `gfx_activity_acc`
* `mem_activity_acc`
* `pcie_nak_sent_count_acc`
* `pcie_nak_rcvd_count_acc`
* `pcie_lc_perf_other_end_recovery`

Recompile any code that reads these fields. Any assignments into fixed-width 32-bit variables must be updated to use 64-bit types.

#### Future deprecation notice: planned removal after the next major release

The following APIs, types, and enums are deprecated and will be removed sometime **after** the next major release.

##### APIs

| Deprecated                                          | Replacement                                             |
|-----------------------------------------------------|---------------------------------------------------------|
| `amdsmi_get_gpu_compute_partition_mem_alloc_mode()` | `amdsmi_get_gpu_accelerator_partition_mem_alloc_mode()` |
| `amdsmi_set_gpu_compute_partition_mem_alloc_mode()` | `amdsmi_set_gpu_accelerator_partition_mem_alloc_mode()` |
| `amdsmi_get_gpu_compute_partition()`                | `amdsmi_get_gpu_accelerator_partition_profile()`        |
| `amdsmi_set_gpu_compute_partition()`                | `amdsmi_set_gpu_accelerator_partition_profile()`        |
| `amdsmi_set_gpu_memory_partition()`                 | `amdsmi_set_gpu_memory_partition_mode()`                |

##### Types

* `amdsmi_compute_partition_type_t`
* `amdsmi_compute_partition_mem_alloc_mode_t`

##### Enums

| Deprecated      | Replacement            |
|-----------------|------------------------|
| `CLK_LIMIT_MIN` | `AMDSMI_CLK_LIMIT_MIN` |
| `CLK_LIMIT_MAX` | `AMDSMI_CLK_LIMIT_MAX` |
| `AGG_BW0`       | `AMDSMI_AGG_BW0`       |
| `RD_BW0`        | `AMDSMI_RD_BW0`        |
| `WR_BW0`        | `AMDSMI_WR_BW0`        |

---

Source: https://rocm.docs.amd.com/en/latest/about/transition-guide-TheRock.html

# Transition guide from legacy ROCm release stream

[ROCm Core SDK 7.14.0](https://rocm.docs.amd.com/en/latest/index.html#rocm-core-sdk) marks a step change from the ROCm legacy release stream. It is built on our new build system, TheRock.

## Major changes

| Feature                | ROCm Core SDK        | ROCm Legacy                                             | Description                                                                                     |
|------------------------|----------------------|---------------------------------------------------------|-------------------------------------------------------------------------------------------------|
| Installation directory | /opt/rocm/core       | /opt/rocm/                                              | To support additional release streams downstream of the ROCm Core SDK                           |
| Package names          | amdrocm-[$component] | rocm-[$component] or roc[$component] or hip[$component] | Unique package prefix to avoid conflicts with upstream packages                                 |
| Extras directory       | /opt/rocm/extras-7/  | N/A                                                     | Shared install prefix scoped to each ROCm major version for projects built on the ROCm Core SDK |

## Paths and linking

ROCm Core SDK 7.14.0 maintains ABI and API compatibility with the ROCm 7.2
legacy releases, so recompilation is not required. For installations using your
Linux distribution’s package manager, the `amdrocm` meta package configures
`update-alternatives` and provides backward-compatible symlinks for
`/opt/rocm/bin`, `/opt/rocm/lib`, and other `/opt/rocm/` directories. For
tarball installs, update `PATH`, `LD_LIBRARY_PATH`, `ROCM_PATH`, or other
environment variables to reflect the new installation path (`/opt/rocm/core`).

## Software packages

ROCm Core SDK packages are more consolidated than the legacy ROCm release
stream. For example, hipBLAS and rocBLAS are now combined into one package,
`amdrocm-blas`. The table below lists new packages, their contents, and the
corresponding legacy packages.

> **Note:** ASAN packages are not available in 7.14.0 and are planned for a future release.

### Linux packages available in ROCm 7.14.0

| ROCm Core SDK Package   | Package Contents                                                     | ROCm Legacy Package                                                               |
|-------------------------|----------------------------------------------------------------------|-----------------------------------------------------------------------------------|
| amdrocm-amdsmi          | amd-smi                                                              | amd-smi-lib, rocm-smi-lib                                                         |
| amdrocm-llvm            | amdclang++, hipcc, flang                                             | rocm-llvm, rocm-llvm-dev, Fortran compiler (included in rocm-llvm OpenMP runtime) |
| amdrocm-runtime         | HIP, ROCR, runtime compilation                                       | hip-runtime-amd, rocm-hip-runtime, rocm-language-runtime, hsa-rocr, comgr         |
| amdrocm-fft             | rocFFT, hipFFT, hipFFTW                                              | rocfft, hipfft                                                                    |
| amdrocm-blas            | rocBLAS, hipBLAS, hipBLASLt, hipSPARSELt                             | rocblas, hipblas, hipblaslt, hipsparselt                                          |
| amdrocm-sparse          | rocSPARSE, hipSPARSE                                                 | rocsparse, hipsparse                                                              |
| amdrocm-solver          | rocSOLVER, hipSOLVER                                                 | rocsolver, hipsolver, rocalution                                                  |
| amdrocm-dnn             | hipDNN, MIOpen                                                       | miopen-hip                                                                        |
| amdrocm-rand            | rocRAND, hipRAND                                                     | rocrand, hiprand                                                                  |
| amdrocm-ccl             | rocPRIM, rocThrust, hipCUB                                           | rocprim, rocthrust, hipcub, rocwmma                                               |
| amdrocm-profiler        | rocprofiler-systems, rocprofiler-compute, rocprofiler-sdk, roctracer | rocprofiler, rocprofiler-compute, rocprofiler-systems, rocprofiler-sdk, roctracer |
| amdrocm-profiler-base   | rocprofiler-sdk, roctracer                                           | rocprofiler-register, roctracer, hsa-amd-aqlprofile                               |
| amdrocm-base            | rocminfo, rocm-core                                                  | rocm-core, rocminfo, rocm-cmake, half                                             |
| amdrocm-ck              | Composable Kernel                                                    | composablekernel                                                                  |
| amdrocm-debugger        | rocgdb, ROCdbgapi, ROCR Debug Agent                                  | rocm-gdb, rocm-dbgapi, rocm-debug-agent                                           |
| amdrocm-hipify          | HIPIFY                                                               | hipify-clang                                                                      |
| amdrocm-opencl          | OpenCL runtime and ICD loader                                        | rocm-opencl-runtime, rocm-opencl, hip-opencl                                      |
| amdrocm-decode          | rocDecode (newly included in the ROCm Core SDK)                      | rocdecode                                                                         |
| amdrocm-jpeg            | rocJPEG (newly included in the ROCm Core SDK)                        | rocjpeg                                                                           |
| amdrocm-rccl            | rccl                                                                 | rccl                                                                              |
| amdrocm-rocshmem        | rocSHMEM                                                             | rocshmem                                                                          |
| amdrocm-rdc             | ROCm Data Center Tool (newly included in the ROCm Core SDK)          | rdc                                                                               |
| amdrocm-sysdeps         | Bundled third-party dependencies (libdrm, libelf, numa, libVA)       | System dependencies                                                               |

Packages are offered in the following variants:

- **For all supported GPUs** – works across all GPUs supported by ROCm (for example, `apt install amdrocm-core-sdk7.14`).
- **For a specific GPU architecture** – smaller install size, but requires you to know the GPU installed in your system (for example, `apt install amdrocm-core-sdk7.14-gfx110x`).

Installing all GPU architectures is not required. You can install packages for a specific architecture, multiple architectures side by side, or all supported GPU architectures.

When redistributing software built on the ROCm Core SDK (for example, via containers), we recommend the all GPU package variant for broad hardware support. If disk footprint is a concern, you can use a single GPU architecture package variant instead.

### Architecture-specific packages available in ROCm 7.14.0

| Architecture Family | Package Suffix | Product Name (Not Exhaustive) |
|-----------------------|------------------------------------|----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| CDNA4 | -gfx950 | AMD Instinct MI355X / MI350X |
| CDNA3 | -gfx94x | AMD Instinct MI325X / MI300X / MI300A |
| CDNA2 | -gfx90a | AMD Instinct MI250X / MI250 / MI210 |
| CDNA | -gfx908 | AMD Instinct MI100 |
| RDNA4 | -gfx120x | AMD Radeon RX 9070 / AMD Radeon RX 9060 / AMD Radeon RX 9070 XT / AMD Radeon RX 9060 XT / AMD Radeon RX 9070 GRE / AMD Radeon AI PRO R9700S / AMD Radeon AI PRO R9700 / AMD Radeon AI PRO R9600D / AMD Radeon RX 9060 XT LP |
| RDNA3.5 | -gfx1150 -gfx1151 -gfx1152 | AMD Ryzen AI 9 465 / AMD Ryzen AI 9 365 / AMD Ryzen AI 9 HX 475 / AMD Ryzen AI 9 HX 470 / AMD Ryzen AI 9 HX 375 / AMD Ryzen AI 9 HX 370 / AMD Ryzen AI 9 PRO 465 / AMD Ryzen AI 9 PRO HX 475 / AMD Ryzen AI 9 PRO HX 470 / AMD Ryzen AI 9 HX PRO 375 / AMD Ryzen AI 9 HX PRO 370 / AMD Ryzen AI Max 390 / AMD Ryzen AI Max 385 / AMD Ryzen AI Max+ 395 / AMD Ryzen AI Max+ 392 / AMD Ryzen AI Max+ 388 / AMD Ryzen AI Max PRO 390 / AMD Ryzen AI Max PRO 385 / AMD Ryzen AI Max PRO 380 / AMD Ryzen AI Max+ PRO 395 / AMD Ryzen AI 7 450 / AMD Ryzen AI 7 350 / AMD Ryzen AI 7 345 / AMD Ryzen AI 5 340 / AMD Ryzen AI 5 330 / AMD Ryzen AI 7 PRO 450 / AMD Ryzen AI 5 PRO 440 / AMD Ryzen AI 7 PRO 350 / AMD Ryzen AI 5 PRO 340 |
| RDNA3 | -gfx110x | AMD Radeon RX 7700 / AMD Radeon RX 7600 / AMD Radeon PRO V710 / AMD Radeon PRO W7900 / AMD Radeon PRO W7800 / AMD Radeon PRO W7700 / AMD Radeon RX 7900 XT / AMD Radeon RX 7800 XT / AMD Radeon RX 7700 XT / AMD Radeon RX 7700 XE / AMD Radeon RX 7900 XTX / AMD Radeon RX 7900 GRE / AMD Radeon PRO W7800 48GB / AMD Radeon PRO W7900 Dual Slot |
| RDNA2 | -gfx1030 | AMD Radeon PRO V620 / AMD Radeon PRO W6800 |

## ROCm Core SDK component changes (moved or removed)

### Planned for future releases

- ROCm Core SDK: RPP
- ROCm-Extras: hipfort, rocPyDecode, rocAL, MIVisionX

### Moved to ROCm-Extras

- ROCm Validation Suite
- TransferBench
- MIGraphX

### Moved to Standalone/ONNX

- ONNX runtime

### Removed

- [ROCm SMI](https://rocm.docs.amd.com/en/latest/about/release-notes.html#rocm-smi-deprecation) (replaced by AMD SMI)
- ROCm Bandwidth Test (deprecated; reached end-of-life with the TheRock-based ROCm 7.14.0 release — use TransferBench or RVS instead)

## Notable package relocations

- rocMLIR (now included in MIGraphX)
- HIPCC (now included in `amdrocm-llvm`)
- FLANG (now included in `amdrocm-llvm`)
- ROCm CMake (now in `amdrocm-base`)
- ROCTracer (now in `amdrocm-profiler-base`)
- ROCProfiler (functionality in `amdrocm-profiler`)

## Components available in the ROCm Core SDK, ROCm-Extras, and Standalone/ONNX

|                 | Category                        | Present                                                                                                                                                             | Absent/Moved                                                             |
|-----------------|---------------------------------|---------------------------------------------------------------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------|
| ROCm Core SDK   | Math and compute libraries      | CK, hipBLAS, hipBLASLt, hipCUB, hipFFT, hipRAND, hipSOLVER, hipSPARSE/SPARSELt, MIOpen, rocBLAS, rocFFT, rocRAND, rocSOLVER, rocSPARSE, rocPRIM, rocThrust, rocWMMA | hipfort, rocALUTION                                                      |
| ROCm Core SDK   | Communication libraries         | RCCL, rocSHMEM                                                                                                                                                      | —                                                                        |
| ROCm Core SDK   | Media libraries                 | rocDecode, rocJPEG, ROCm Performance Primitives (RPP planned for a future release)                                                                                  | rocPyDecode, rocAL, MIVisionX, MIGraphX, CK (moved to math and compute)  |
| ROCm Core SDK   | Runtime, compilers, build tools | HIP, HIPIFY, LLVM                                                                                                                                                   | HIPCC, FLANG, ROCm CMake                                                 |
| ROCm Core SDK   | Profiling and debugging tools   | ROCm Compute Profiler, ROCm Systems Profiler, ROCprofiler-SDK, ROCdbgapi, ROCm Debugger, ROCR Debug Agent                                                           | ROCTracer, ROCProfiler                                                   |
| ROCm Core SDK   | Control and monitoring tools    | AMD SMI, ROCm Data Center Tool, rocminfo, hipinfo                                                                                                                   | ROCm SMI (removed), ROCm Validation Suite, ROCm Bandwidth Test (removed) |
| ROCm-Extras     | —                               | ROCm Validation Suite, TransferBench, MIGraphX                                                                                                                      | —                                                                        |
| Standalone/ONNX | —                               | rocMLIR, ONNX runtime                                                                                                                                               | —                                                                        |

---

Source: https://rocm.docs.amd.com/en/latest/about/what-is-rocm.html

# What is ROCm?

ROCm is the AMD open software stack for GPU‑accelerated computing. It provides
the tools needed to program AMD GPUs — including runtimes, compilers,
performance and system utilities, and optimized math and compute libraries. The
wider ROCm ecosystem includes ROCm‑enabled HPC applications and deep learning
frameworks such as PyTorch.

ROCm 7.14.0 is built through the [TheRock](https://github.com/ROCm/TheRock), AMD’s open build and release system.
TheRock replaces the previous monolithic release process with a modular
workflow that makes ROCm components easier to build, integrate, and distribute.
See the [release notes](https://rocm.docs.amd.com/en/latest/about/release-notes.html) for more information.

## ROCm Core SDK

The ROCm Core SDK provides the foundational components that power the ROCm
ecosystem — runtimes, compilers, math libraries, and system utilities for GPGPU
computing. See [ROCm Core SDK components](https://rocm.docs.amd.com/en/latest/components/core.html) for more information.

## ROCm Extras

ROCm Extra components are supplementary tools for benchmarking, validating, and managing ROCm
deployments. These tools are not required for GPU application development but are useful
for verifying hardware health, measuring system performance, and managing GPU fleets.
For more information, see [ROCm Extra components](https://rocm.docs.amd.com/en/latest/components/extras.html).

## Get started

* See the release notes – [ROCm Core SDK 7.14.0 release notes](https://rocm.docs.amd.com/en/latest/about/release-notes.html) – to learn about the
  latest changes and the current state of ROCm.
* See the [ROCm 7.14.0 compatibility matrix](https://rocm.docs.amd.com/en/latest/compatibility/compatibility-matrix.html#compat-matrix) – for system requirements and AMD hardware
  compatibility information.
* Follow [Install AMD ROCm 7.14.0](https://rocm.docs.amd.com/en/latest/install/rocm.html) to set up ROCm on your system.

---

Source: https://rocm.docs.amd.com/en/latest/compatibility/compatibility-matrix.html

# ROCm 7.14.0 compatibility matrix

To plan your ROCm 7.14.0 installation, use the following selector to
view ROCm compatibility and system requirements information for your AMD
hardware configuration and system environment. For installation instructions,
see [Install AMD ROCm 7.14.0](https://rocm.docs.amd.com/en/latest/install/rocm.html).

**w=compute**

**fam=instinct fam=radeon fam=ryzen**

**w=compute**

**fam=instinct**

**fam=radeon**

**gfx=gfx1201 gfx=gfx1200**

**gfx=gfx1101**

**fam=ryzen**

**fam=all**

**w=graphics**

---

## System requirements and information

ROCm depends on a coordinated stack of compatible firmware, driver, and user
space components. Maintaining version alignment between these layers ensures
expected GPU operation and performance, especially for AMD data center
products. This table lists GPU details followed by supported operating systems,
kernel driver, and firmware versions.

| AMD GPU series | [AMD Instinct MI350 Series](https://www.amd.com/en/products/accelerators/instinct/mi350.html) | [AMD Instinct MI300 Series](https://www.amd.com/en/products/accelerators/instinct/mi300.html) | [AMD Instinct MI200 Series](https://www.amd.com/en/products/accelerators/instinct/mi200.html) | [AMD Instinct MI100 Series](https://www.amd.com/en/products/accelerators/instinct/mi100.html) | | | |
|-------------------------------------------------|-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|-------------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|-------------------------------------------------------------------------------------------------|--------------------------|---------------------------------------------------|--------------------|
| Architecture | CDNA 4 | CDNA 3 | CDNA 2 | CDNA | | | |
| LLVM target | gfx950 | gfx942 | gfx90a | gfx908 | | | |
| Supported Ubuntu versions | Ubuntu 26.04 (GA kernel: 7.0) Ubuntu 24.04.4 (GA kernel: 6.8) Ubuntu 22.04.5 (GA kernel: 5.15) | Ubuntu 26.04 (GA kernel: 7.0) Ubuntu 24.04.4 (GA kernel: 6.8) | | | | | |
| Supported Red Hat Enterprise Linux versions | RHEL 10.2 (kernel: 6.12.0-211) RHEL 10.0 (kernel: 6.12.0-55) RHEL 9.8 (kernel: 5.14.0-687) RHEL 9.6 (kernel: 5.14.0-570) RHEL 9.4 (kernel: 5.14.0-427) RHEL 8.10 (kernel: 4.18.0-553) | RHEL 10.2 (kernel: 6.12.0-211) RHEL 9.8 (kernel: 5.14.0-687) RHEL 9.6 (kernel: 5.14.0-570) | RHEL 10.2 (kernel: 6.12.0-211) RHEL 10.0 (kernel: 6.12.0-55) RHEL 9.8 (kernel: 5.14.0-687) RHEL 9.6 (kernel: 5.14.0-570) RHEL 9.4 (kernel: 5.14.0-427) | | | | |
| Supported Debian version | Debian 13 (kernel: 6.12) Debian 12 (kernel: 6.1.0) | Debian 13 (kernel: 6.12) | Debian 12 (kernel: 6.1.0) | | | | |
| Supported Oracle Linux versions | Oracle Linux 10 (kernel: UEK 8.1) Oracle Linux 9 (kernel: UEK 8) | Oracle Linux 10 (kernel: UEK 8.1) Oracle Linux 9 (kernel: UEK 8) Oracle Linux 8 (kernel: UEK 7) | | | | | |
| Supported Rocky Linux versions | Rocky Linux 9 (kernel: 5.14.0-570) | | | | | | |
| Supported SUSE Linux Enterprise Server versions | SLES 16.0 (kernel: 6.12) SLES 15.7 (kernel: 6.4.0-150700.51) | SLES 15.7 (kernel: 6.4.0-150700.51) | | | | | |
| Supported AMD GPU Driver (amdgpu) versions | [31.40.1](https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-31.40.1/documentation/release-notes.html) [31.40.0](https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-31.40.0/documentation/release-notes.html) [31.30.0](https://instinct.docs.amd.com/projects/amdgpu-docs/en/31.30.0-preview/documentation/release-notes.html) [31.20.0](https://instinct.docs.amd.com/projects/amdgpu-docs/en/31.20.0-preview/documentation/release-notes.html) [31.10.0](https://instinct.docs.amd.com/projects/amdgpu-docs/en/31.10.0-preview/documentation/release-notes.html) [30.30.3](https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-30.30.3/documentation/release-notes.html) [30.30.2](https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-30.30.2/documentation/release-notes.html) [30.30.1](https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-30.30.1/documentation/release-notes.html) [30.30.0](https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-30.30.0/documentation/release-notes.html) [30.20.1](https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-30.20.1/documentation/release-notes.html) [30.20.0](https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-30.20.0/documentation/release-notes.html) [30.10.2](https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-30.10.2/documentation/release-notes.html) [30.10.1](https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-30.10.1/documentation/release-notes.html) [30.10.0](https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-30.10/documentation/release-notes.html) | | | | | | |
| Supported PLDM bundle (firmware) versions | 01.26.01.03 (or later) 01.26.00.02 | BKC12.0 (IFWI PRD1000A) or later IFWI 00189938 | 01.26.01.03 (or later) 01.25.06.08 | 01.26.00.04 (or later) 01.25.06.05 | BKC 27 BKC 26.1 | Maintenance update (MU) 5 with IFWI 75 (or later) | VBIOS D3430401-037 |

| AMD GPU series | [AMD Radeon AI PRO R9000 Series](https://www.amd.com/en/products/graphics/workstations/radeon-ai-pro.html#tabs-95fa144b96-item-b95ec9e1ca-tab) | [AMD Radeon PRO W7000 Series](https://www.amd.com/en/products/graphics/workstations/radeon-pro.html#tabs-990fdead92-item-20daa37284-tab) | [AMD Radeon PRO W6000 Series](https://www.amd.com/en/products/graphics/workstations/radeon-pro/w6800.html) | [AMD Radeon PRO V Series](https://www.amd.com/en/products/accelerators/radeon-pro.html) | [AMD Radeon RX 9000 Series](https://www.amd.com/en/products/graphics/desktops/radeon.html#tabs-ff9c5c3863-item-37fb38a236-tab) | [AMD Radeon RX 7000 Series](https://www.amd.com/en/products/graphics/desktops/radeon/7000-series.html) |
|--------------------------------------------|-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------|-------------------------------------------------------------------------------------------|----------------------------------------------------------------------------------------------------------------------------------|----------------------------------------------------------------------------------------------------------|
| Architecture | RDNA 4 | RDNA 3 | RDNA 2 | | | |
| LLVM target | gfx1201 | gfx1200 | gfx1100 | gfx1101 | gfx1102 | gfx1030 |
| Supported Ubuntu versions | Ubuntu 26.04 (GA kernel: 7.0) Ubuntu 24.04.4 (GA kernel: 6.8) Ubuntu 22.04.5 (GA kernel: 5.15) | | | | | |
| Supported RHEL versions | RHEL 10.2 (kernel: 6.12.0-211) RHEL 9.8 (kernel: 5.14.0-687) | | | | | |
| Supported Windows version | Windows 11 25H2 | | | | | |
| Supported AMD GPU Driver (amdgpu) versions | [31.40.1](https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-31.40.1/documentation/release-notes.html) [31.40.0](https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-31.40.0/documentation/release-notes.html) [31.30.0](https://instinct.docs.amd.com/projects/amdgpu-docs/en/31.30.0-preview/documentation/release-notes.html) [31.20.0](https://instinct.docs.amd.com/projects/amdgpu-docs/en/31.20.0-preview/documentation/release-notes.html) [31.10.0](https://instinct.docs.amd.com/projects/amdgpu-docs/en/31.10.0-preview/documentation/release-notes.html) [30.30.3](https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-30.30.3/documentation/release-notes.html) [30.30.2](https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-30.30.2/documentation/release-notes.html) [30.30.1](https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-30.30.1/documentation/release-notes.html) [30.30.0](https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-30.30.0/documentation/release-notes.html) [30.20.1](https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-30.20.1/documentation/release-notes.html) [30.20.0](https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-30.20.0/documentation/release-notes.html) [30.10.2](https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-30.10.2/documentation/release-notes.html) [30.10.1](https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-30.10.1/documentation/release-notes.html) [30.10.0](https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-30.10/documentation/release-notes.html) | | | | | |
| Supported Adrenalin Driver version | [26.6.4](https://www.amd.com/en/resources/support-articles/release-notes/RN-RAD-WIN-26-6-4.html) | | | | | |
| Supported Windows OEM Driver version | 26.10.28 | | | | | |

| AMD APU series | [AMD Ryzen AI Max PRO 400 Series](https://www.amd.com/en/products/processors/workstations/mobile.html#tabs-7f0c432fb2-item-296e08633e-tab) | [AMD Ryzen AI Max PRO 300 Series](https://www.amd.com/en/products/processors/workstations/mobile.html#tabs-7f0c432fb2-item-5116ab7a74-tab) | [AMD Ryzen AI Max 300 Series](https://www.amd.com/en/products/processors/laptop/ryzen.html#tabs-1181ea0b44-item-6ccfea5f65-tab) | [AMD Ryzen AI PRO 400 Series](https://www.amd.com/en/products/processors/workstations/mobile.html#tabs-7f0c432fb2-item-0c42136112-tab) | [AMD Ryzen AI PRO 400 Series](https://www.amd.com/en/products/processors/laptop/ryzen-for-business.html#tabs-0d174caf43-item-87690677fc-tab) | [AMD Ryzen AI 400 Series](https://www.amd.com/en/products/processors/consumer/ryzen-ai.html#tabs-f556098628-item-808b56dca3-tab) | [AMD Ryzen AI PRO 300 Series](https://www.amd.com/en/products/processors/workstations/mobile.html#tabs-7f0c432fb2-item-387526c6cc-tab) | [AMD Ryzen AI 300 Series](https://www.amd.com/en/products/processors/consumer/ryzen-ai.html#tabs-f556098628-item-54e149d850-tab) | [AMD Ryzen PRO 200 Series](https://www.amd.com/en/products/processors/laptop/ryzen-for-business.html#tabs-0d174caf43-item-a8ec88d07e-tab) | [AMD Ryzen 200 Series](https://www.amd.com/en/products/processors/laptop/ryzen.html#tabs-1181ea0b44-item-895d56feed-tab) | | |
|--------------------------------------|----------------------------------------------------------------------------------------------------------------------------------------------|----------------------------------------------------------------------------------------------------------------------------------------------|-----------------------------------------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------------------------------------------|---------------------------------------------------------------------------------------------------------------------------------------------|----------------------------------------------------------------------------------------------------------------------------|-------------|-------------|
| Graphics model (iGPU) | Radeon 8065S | Radeon 8060S | Radeon 8050S | Radeon 8040S | Radeon 890M | Radeon 880M | Radeon 860M | Radeon 840M | Radeon 820M | Radeon 780M | Radeon 760M | Radeon 740M |
| Architecture | RDNA 3.5 | RDNA 3 | | | | | | | | | | |
| LLVM target | gfx1151 | gfx1150 | gfx1152 | gfx1153 | gfx1103 | | | | | | | |
| Supported Ubuntu versions | 26.04 (GA kernel: 7.0) 24.04.4 (HWE kernel: 6.17) | | | | | | | | | | | |
| Supported Windows version | Windows 11 25H2 | | | | | | | | | | | |
| Supported kernel driver version | Inbox kernel driver in supported Ubuntu version | | | | | | | | | | | |
| Supported Adrenalin Driver version | [26.6.4](https://www.amd.com/en/resources/support-articles/release-notes/RN-RAD-WIN-26-6-4.html) | | | | | | | | | | | |
| Supported Windows OEM Driver version | 26.10.28 | | | | | | | | | | | |

For hardware specifications, see [AMD GPU specifications](https://rocm.docs.amd.com/en/latest/reference/gpu-specs.html#gpu-specs).

---

## ROCm Core SDK components

The following table lists core components included in the ROCm 7.14.0
release. Expect future releases in this stream to expand the list of
components.

| Component group                                                                                                                          | Component name                                                                                            |                                                                                              |                                                                                                                                          |                                                                                                          |                                                                                                                                          |
|------------------------------------------------------------------------------------------------------------------------------------------|-----------------------------------------------------------------------------------------------------------|----------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------------------------------------------------|----------------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------------------------------------------------|
| Math and compute libraries                                                                                                               | [Composable Kernel 1.2.0](https://rocm.docs.amd.com/projects/composable_kernel/en/docs-7.14.0/index.html) |                                                                                              |                                                                                                                                          |                                                                                                          |                                                                                                                                          |
| [hipBLAS 3.5.0](https://rocm.docs.amd.com/projects/hipBLAS/en/docs-7.14.0/index.html)                                                    |                                                                                                           |                                                                                              |                                                                                                                                          |                                                                                                          |                                                                                                                                          |
| [hipBLASLt 1.4.1](https://rocm.docs.amd.com/projects/hipBLASLt/en/docs-7.14.0/index.html)                                                |                                                                                                           |                                                                                              |                                                                                                                                          |                                                                                                          |                                                                                                                                          |
| [hipCUB 4.5.0](https://rocm.docs.amd.com/projects/hipCUB/en/docs-7.14.0/index.html)                                                      |                                                                                                           |                                                                                              |                                                                                                                                          |                                                                                                          |                                                                                                                                          |
| [hipFFT 1.0.24](https://rocm.docs.amd.com/projects/hipFFT/en/docs-7.14.0/index.html)                                                     |                                                                                                           |                                                                                              |                                                                                                                                          |                                                                                                          |                                                                                                                                          |
| [hipRAND 3.4.0](https://rocm.docs.amd.com/projects/hipRAND/en/docs-7.14.0/index.html)                                                    |                                                                                                           |                                                                                              |                                                                                                                                          |                                                                                                          |                                                                                                                                          |
| [hipSOLVER 3.5.0](https://rocm.docs.amd.com/projects/hipSOLVER/en/docs-7.14.0/index.html)                                                |                                                                                                           |                                                                                              |                                                                                                                                          |                                                                                                          |                                                                                                                                          |
| [hipSPARSE 4.6.0](https://rocm.docs.amd.com/projects/hipSPARSE/en/docs-7.14.0/index.html)                                                |                                                                                                           |                                                                                              |                                                                                                                                          |                                                                                                          |                                                                                                                                          |
| [hipSPARSELt 0.2.9](https://rocm.docs.amd.com/projects/hipSPARSELt/en/docs-7.14.0/index.html)                                            |                                                                                                           |                                                                                              |                                                                                                                                          |                                                                                                          |                                                                                                                                          |
| [MIOpen 3.5.2](https://rocm.docs.amd.com/projects/MIOpen/en/docs-7.14.0/index.html)                                                      |                                                                                                           |                                                                                              |                                                                                                                                          |                                                                                                          |                                                                                                                                          |
| [rocBLAS 5.5.0](https://rocm.docs.amd.com/projects/rocBLAS/en/docs-7.14.0/index.html)                                                    |                                                                                                           |                                                                                              |                                                                                                                                          |                                                                                                          |                                                                                                                                          |
| [rocFFT 1.0.38](https://rocm.docs.amd.com/projects/rocFFT/en/docs-7.14.0/index.html)                                                     |                                                                                                           |                                                                                              |                                                                                                                                          |                                                                                                          |                                                                                                                                          |
| [rocPRIM 4.5.0](https://rocm.docs.amd.com/projects/rocPRIM/en/docs-7.14.0/index.html)                                                    |                                                                                                           |                                                                                              |                                                                                                                                          |                                                                                                          |                                                                                                                                          |
| [rocRAND 4.5.0](https://rocm.docs.amd.com/projects/rocRAND/en/docs-7.14.0/index.html)                                                    |                                                                                                           |                                                                                              |                                                                                                                                          |                                                                                                          |                                                                                                                                          |
| [rocSOLVER 3.35.0](https://rocm.docs.amd.com/projects/rocSOLVER/en/docs-7.14.0/index.html)                                               |                                                                                                           |                                                                                              |                                                                                                                                          |                                                                                                          |                                                                                                                                          |
| [rocSPARSE 4.7.0](https://rocm.docs.amd.com/projects/rocSPARSE/en/docs-7.14.0/index.html)                                                |                                                                                                           |                                                                                              |                                                                                                                                          |                                                                                                          |                                                                                                                                          |
| [rocThrust 4.5.0](https://rocm.docs.amd.com/projects/rocThrust/en/docs-7.14.0/index.html)                                                |                                                                                                           |                                                                                              |                                                                                                                                          |                                                                                                          |                                                                                                                                          |
| [rocWMMA 2.2.1](https://rocm.docs.amd.com/projects/rocWMMA/en/docs-7.14.0/index.html)                                                    |                                                                                                           |                                                                                              |                                                                                                                                          |                                                                                                          |                                                                                                                                          |
| Communication libraries                                                                                                                  | Communication libraries                                                                                   | Communication libraries                                                                      | [RCCL 2.30.4](https://github.com/ROCm/rocm-systems/tree/therock-7.14/projects/rccl)                                                      |                                                                                                          |                                                                                                                                          |
| [rocSHMEM 3.5.0](https://github.com/ROCm/rocm-systems/tree/therock-7.14/projects/rocshmem)                                               |                                                                                                           |                                                                                              |                                                                                                                                          |                                                                                                          |                                                                                                                                          |
| Media libraries                                                                                                                          | Media libraries                                                                                           | [rocDecode 1.8.0](https://github.com/ROCm/rocm-systems/tree/therock-7.14/projects/rocdecode) | [rocDecode 1.8.0](https://github.com/ROCm/rocm-systems/tree/therock-7.14/projects/rocdecode)                                             |                                                                                                          |                                                                                                                                          |
| [rocJPEG 1.6.0](https://github.com/ROCm/rocm-systems/tree/therock-7.14/projects/rocjpeg)                                                 | [rocJPEG 1.6.0](https://github.com/ROCm/rocm-systems/tree/therock-7.14/projects/rocjpeg)                  |                                                                                              |                                                                                                                                          |                                                                                                          |                                                                                                                                          |
| Storage libraries                                                                                                                        | [hipFile 0.3.0](https://github.com/ROCm/rocm-systems/tree/therock-7.14/projects/hipFile)                  |                                                                                              |                                                                                                                                          |                                                                                                          |                                                                                                                                          |
| Runtimes and compilers                                                                                                                   | Runtimes and compilers                                                                                    | [HIP 7.14](https://github.com/ROCm/rocm-systems/tree/therock-7.14/projects/hip)              |                                                                                                                                          |                                                                                                          |                                                                                                                                          |
| [HIPIFY 7.14](https://github.com/ROCm/HIPIFY/tree/therock-7.14)                                                                          |                                                                                                           |                                                                                              |                                                                                                                                          |                                                                                                          |                                                                                                                                          |
| [LLVM 23.0.0](https://github.com/ROCm/llvm-project/tree/therock-7.14)                                                                    |                                                                                                           |                                                                                              |                                                                                                                                          |                                                                                                          |                                                                                                                                          |
| [ROCr Runtime 1.21.0](https://github.com/ROCm/rocm-systems/tree/therock-7.14/projects/rocr-runtime)                                      |                                                                                                           |                                                                                              |                                                                                                                                          |                                                                                                          |                                                                                                                                          |
| [SPIRV-LLVM-Translator 23.0.0](https://github.com/ROCm/SPIRV-LLVM-Translator/tree/therock-7.14)                                          |                                                                                                           |                                                                                              |                                                                                                                                          |                                                                                                          |                                                                                                                                          |
| Profiling and debugging tools                                                                                                            | Profiling and debugging tools                                                                             | Profiling and debugging tools                                                                | [ROCm Compute Profiler (rocprofiler-compute) 3.7.0](https://github.com/ROCm/rocm-systems/tree/therock-7.14/projects/rocprofiler-compute) | [ROCprofiler-SDK 1.3.2](https://github.com/ROCm/rocm-systems/tree/therock-7.14/projects/rocprofiler-sdk) | [ROCm Compute Profiler (rocprofiler-compute) 3.7.0](https://github.com/ROCm/rocm-systems/tree/therock-7.14/projects/rocprofiler-compute) |
| [ROCm Systems Profiler (rocprofiler-systems) 1.7.0](https://github.com/ROCm/rocm-systems/tree/therock-7.14/projects/rocprofiler-systems) |                                                                                                           |                                                                                              |                                                                                                                                          |                                                                                                          |                                                                                                                                          |
| [ROCm Systems Profiler (rocprofiler-systems) 1.7.0](https://github.com/ROCm/rocm-systems/tree/therock-7.14/projects/rocprofiler-systems) |                                                                                                           |                                                                                              |                                                                                                                                          |                                                                                                          |                                                                                                                                          |
| [ROCprofiler-SDK 1.3.2](https://github.com/ROCm/rocm-systems/tree/therock-7.14/projects/rocprofiler-sdk)                                 |                                                                                                           |                                                                                              |                                                                                                                                          |                                                                                                          |                                                                                                                                          |
| [ROCprofiler-SDK 1.3.2](https://github.com/ROCm/rocm-systems/tree/therock-7.14/projects/rocprofiler-sdk)                                 |                                                                                                           |                                                                                              |                                                                                                                                          |                                                                                                          |                                                                                                                                          |
| [ROCdbgapi 0.80.0](https://github.com/ROCm/rocm-systems/tree/therock-7.14/projects/rocdbgapi)                                            |                                                                                                           |                                                                                              |                                                                                                                                          |                                                                                                          |                                                                                                                                          |
| [ROCm Debugger (ROCgdb) 16.3](https://github.com/ROCm/ROCgdb/tree/therock-7.14)                                                          |                                                                                                           |                                                                                              |                                                                                                                                          |                                                                                                          |                                                                                                                                          |
| [ROCR Debug Agent 2.1.0](https://github.com/ROCm/rocm-systems/tree/therock-7.14/projects/rocr-debug-agent)                               |                                                                                                           |                                                                                              |                                                                                                                                          |                                                                                                          |                                                                                                                                          |
| Control and monitoring tools                                                                                                             | [AMD SMI 26.5.0](https://github.com/ROCm/rocm-systems/tree/therock-7.14/projects/amdsmi)                  |                                                                                              |                                                                                                                                          |                                                                                                          |                                                                                                                                          |
| [rocminfo 1.0.0](https://github.com/ROCm/rocm-systems/tree/therock-7.14/projects/rocminfo)                                               |                                                                                                           |                                                                                              |                                                                                                                                          |                                                                                                          |                                                                                                                                          |
| [ROCm Data Center Tool (RDC) 1.3.1](https://github.com/ROCm/rocm-systems/tree/therock-7.14/projects/rdc)                                 |                                                                                                           |                                                                                              |                                                                                                                                          |                                                                                                          |                                                                                                                                          |

| Component group                                                                                 | Component name                                                                                                |
|-------------------------------------------------------------------------------------------------|---------------------------------------------------------------------------------------------------------------|
| Math and compute libraries                                                                      | [Composable Kernel 1.2.0](https://github.com/ROCm/rocm-libraries/tree/therock-7.14/projects/composablekernel) |
| [hipBLAS 3.5.0](https://github.com/ROCm/rocm-libraries/tree/therock-7.14/projects/hipblas)      |                                                                                                               |
| [hipBLASLt 1.4.1](https://github.com/ROCm/rocm-libraries/tree/therock-7.14/projects/hipblaslt)  |                                                                                                               |
| [hipCUB 4.5.0](https://github.com/ROCm/rocm-libraries/tree/therock-7.14/projects/hipcub)        |                                                                                                               |
| [hipFFT 1.0.24](https://github.com/ROCm/rocm-libraries/tree/therock-7.14/projects/hipfft)       |                                                                                                               |
| [hipRAND 3.4.0](https://github.com/ROCm/rocm-libraries/tree/therock-7.14/projects/hiprand)      |                                                                                                               |
| [hipSOLVER 3.5.0](https://github.com/ROCm/rocm-libraries/tree/therock-7.14/projects/hipsolver)  |                                                                                                               |
| [hipSPARSE 4.6.0](https://github.com/ROCm/rocm-libraries/tree/therock-7.14/projects/hipsparse)  |                                                                                                               |
| [MIOpen 3.5.2](https://github.com/ROCm/rocm-libraries/tree/therock-7.14/projects/miopen)        |                                                                                                               |
| [rocBLAS 5.5.0](https://github.com/ROCm/rocm-libraries/tree/therock-7.14/projects/rocblas)      |                                                                                                               |
| [rocFFT 1.0.38](https://github.com/ROCm/rocm-libraries/tree/therock-7.14/projects/rocfft)       |                                                                                                               |
| [rocPRIM 4.5.0](https://github.com/ROCm/rocm-libraries/tree/therock-7.14/projects/rocprim)      |                                                                                                               |
| [rocRAND 4.5.0](https://github.com/ROCm/rocm-libraries/tree/therock-7.14/projects/rocrand)      |                                                                                                               |
| [rocSOLVER 3.35.0](https://github.com/ROCm/rocm-libraries/tree/therock-7.14/projects/rocsolver) |                                                                                                               |
| [rocSPARSE 4.7.0](https://github.com/ROCm/rocm-libraries/tree/therock-7.14/projects/rocsparse)  |                                                                                                               |
| [rocThrust 4.5.0](https://github.com/ROCm/rocm-libraries/tree/therock-7.14/projects/rocthrust)  |                                                                                                               |
| [rocWMMA 2.2.1](https://github.com/ROCm/rocm-libraries/tree/therock-7.14/projects/rocwmma)      |                                                                                                               |
| Support libraries                                                                               | [ROCm CMake](https://github.com/ROCm/rocm-cmake/tree/therock-7.14)                                            |
| Runtimes and compilers                                                                          | [HIP 7.14](https://github.com/ROCm/rocm-systems/tree/therock-7.14/projects/hip)                               |
| [HIPIFY 7.14](https://github.com/ROCm/HIPIFY/tree/therock-7.14)                                 |                                                                                                               |
| [LLVM 23.0.0](https://github.com/ROCm/llvm-project/tree/therock-7.14)                           |                                                                                                               |
| [SPIRV-LLVM-Translator 23.0.0](https://github.com/ROCm/SPIRV-LLVM-Translator/tree/therock-7.14) |                                                                                                               |
| Control and monitoring tools                                                                    | hipinfo                                                                                                       |

---

**gpu=mi355x gpu=mi350x gpu=mi325x gpu=mi300x gpu=mi210 — GPU virtualization support**

**gpu=mi355x gpu=mi350x gpu=mi325x gpu=mi300x gpu=mi210**

Supported SR-IOV configurations require the GPU-IOV Module (GIM) driver
9.1.0.K – see the [AMD Instinct Virtualization Driver documentation](https://instinct.docs.amd.com/projects/virt-drv/en/mainline-9.1.0.k/) to
get started.

| Hypervisor   | Virtualization technology                                                       | Virtualization driver                                                           | Host OS      | Guest OS     |
|--------------|---------------------------------------------------------------------------------|---------------------------------------------------------------------------------|--------------|--------------|
| KVM          | Passthrough                                                                     | —                                                                               | Ubuntu 24.04 | Ubuntu 24.04 |
| SR-IOV       | [GIM 9.1.0.K](https://github.com/amd/MxGPU-Virtualization/releases/tag/9.1.0.K) | Ubuntu 24.04                                                                    | Ubuntu 24.04 |              |
| RHEL 10.0    |                                                                                 |                                                                                 |              |              |
| RHEL 9.6     |                                                                                 |                                                                                 |              |              |
| ESXi         | SR-IOV                                                                          | —                                                                               | ESXi 9.1     | Ubuntu 24.04 |
| KVM          | Passthrough                                                                     | —                                                                               | Ubuntu 24.04 | Ubuntu 24.04 |
| SR-IOV       | [GIM 9.1.0.K](https://github.com/amd/MxGPU-Virtualization/releases/tag/9.1.0.K) | Ubuntu 24.04                                                                    | Ubuntu 24.04 |              |
| RHEL 9.6     |                                                                                 |                                                                                 |              |              |
| ESXi         | SR-IOV                                                                          | —                                                                               | ESXi 9.1     | Ubuntu 24.04 |
| KVM          | SR-IOV                                                                          | [GIM 9.1.0.K](https://github.com/amd/MxGPU-Virtualization/releases/tag/9.1.0.K) | Ubuntu 22.04 | Ubuntu 22.04 |
| KVM          | Passthrough                                                                     | —                                                                               | Ubuntu 22.04 | Ubuntu 22.04 |
| SR-IOV       | [GIM 9.1.0.K](https://github.com/amd/MxGPU-Virtualization/releases/tag/9.1.0.K) | Ubuntu 24.04                                                                    | Ubuntu 24.04 |              |
| Ubuntu 22.04 | Ubuntu 22.04                                                                    |                                                                                 |              |              |
| KVM          | Passthrough                                                                     | —                                                                               | RHEL 9.4     | Ubuntu 22.04 |
| SR-IOV       | [GIM 9.1.0.K](https://github.com/amd/MxGPU-Virtualization/releases/tag/9.1.0.K) | RHEL 9.4                                                                        | Ubuntu 22.04 |              |
| RHEL 9.4     |                                                                                 |                                                                                 |              |              |

See the [release notes](https://rocm.docs.amd.com/en/latest/about/release-notes.html#release-virtualization-support)
for the full list of supported configurations.

**gpu=ai-r9700s gpu=v710 — GPU virtualization support**

**gpu=ai-r9700s gpu=v710**

| Hypervisor   | Virtualization technology   | Virtualization driver                                                           | Host OS      | Guest OS     |
|--------------|-----------------------------|---------------------------------------------------------------------------------|--------------|--------------|
| KVM          | Passthrough                 | —                                                                               | Ubuntu 24.04 | Ubuntu 24.04 |
| KVM          | SR-IOV                      | [GIM 9.1.0.K](https://github.com/amd/MxGPU-Virtualization/releases/tag/9.1.0.K) | Ubuntu 24.04 | Ubuntu 24.04 |

See the [release notes](https://rocm.docs.amd.com/en/latest/about/release-notes.html#release-virtualization-support)
for the full list of supported configurations.

## AI ecosystem compatibility

ROCm 7.14.0 provides optimized support for popular deep learning
frameworks and AI inference engines. The following table lists supported
frameworks and libraries, their validated versions, and compatible Python
versions.

| Framework    | Supported versions     | Python versions                |                        |
|--------------|------------------------|--------------------------------|------------------------|
| PyTorch      | 2.12.0, 2.11.0, 2.10.0 | 2.12.0, 2.11.0                 | 3.14, 3.13, 3.12, 3.11 |
| PyTorch      | 2.12.0                 | 3.14, 3.13, 3.12, 3.11         |                        |
| JAX          | 0.10.0, 0.9.1          | 3.14, 3.13, 3.12, 3.11         |                        |
| vLLM         | 0.23                   | 3.14 (requires PyTorch 2.11.0) |                        |
| SGLang       | 0.5.13                 | 3.14 (requires PyTorch 2.11.0) |                        |
| MIGraphX     | 2.16                   | 3.12                           |                        |
| ONNX Runtime | 1.23.2                 | 3.12                           |                        |

---

Source: https://rocm.docs.amd.com/en/latest/install/rocm.html

# Install AMD ROCm 7.14.0

---

Use the following selector to choose your installation method for your
supported AMD GPU or APU and operating system. For system requirements and
support information, see the [Compatibility matrix](https://rocm.docs.amd.com/en/latest/compatibility/compatibility-matrix.html). To learn more about changes introduced
in ROCm 7.14.0, see the [Release notes](https://rocm.docs.amd.com/en/latest/about/release-notes.html).

#### NOTE
If your GPU is not listed, it might be community-enabled through TheRock
nightly builds. For more information, see [TheRock supported GPUs](https://github.com/ROCm/TheRock/blob/main/SUPPORTED_GPUS.md). For
installation guidance, see [TheRock releases](https://github.com/ROCm/TheRock/blob/main/RELEASES.md).

**fam=radeon fam=ryzen**

**fam=all**

**w=compute**

**fam=instinct fam=radeon fam=ryzen**

**w=compute**

**fam=instinct**

**fam=radeon**

**gfx=gfx1201 gfx=gfx1200**

**gfx=gfx1101**

**fam=ryzen**

**fam=all**

**w=graphics**

**fam=radeon fam=ryzen**

**os=windows**

**os=ubuntu**

**fam=instinct**

**fam=radeon fam=all**

**fam=ryzen**

**os=wsl**

**os=debian**

**os=rhel**

**os=oracle-linux**

**os=rocky-linux**

**os=sles**

**os=windows**

**w=compute**

**fam=all**

**fam=instinct fam=radeon fam=ryzen**

---

**i=pkgman**

**os=ubuntu**

This installation method uses Ubuntu’s native package manager `apt`
to install ROCm. This method suits standard system-wide installations
where ROCm packages should be tracked, updated, and removed through
system package workflows.

**os=debian**

This installation method uses Debian’s native package manager `apt`
to install ROCm. This method suits standard system-wide installations
where ROCm packages should be tracked, updated, and removed through
system package workflows.

**os=rhel**

This installation method uses RHEL’s native package manager `dnf` to
install ROCm. This method suits standard system-wide installations where
ROCm packages should be tracked, updated, and removed through system
package workflows.

**os=oracle-linux**

This installation method uses Oracle Linux’s native package manager
`dnf` to install ROCm. This method suits standard system-wide
installations where ROCm packages should be tracked, updated, and removed
through system package workflows.

**os=rocky-linux**

This installation method uses Rocky Linux’s native package manager
`dnf` to install ROCm. This method suits standard system-wide
installations where ROCm packages should be tracked, updated, and removed
through system package workflows.

**os=sles**

This installation method uses SLES’s native package manager `zypper`
to install ROCm. This method suits standard system-wide installations
where ROCm packages should be tracked, updated, and removed through
system package workflows.

**i=pip**

The pip installation method provides ROCm components as Python wheel
packages in a virtual environment. This method suits Python-focused
development workflows that use an isolated, per-project ROCm environment
managed with standard Python packaging tools.

**i=tar**

The tarball installation method provides ROCm as a self-contained
installation from a pre-built archive. This method suits controlled or
restricted environments requiring manual placement, updates, and removal
outside the system package manager.

**i=runfile**

The ROCm Runfile Installer can install ROCm and/or the AMD GPU Driver (amdgpu)
without using a native Linux package management system, making it ideal for
systems with policy constraints or restricted environments. Network access is
not needed for install as long as dependencies for ROCm and/or AMD GPU driver
(amdgpu) are met. A single installer supports all GFX architectures, automates
post-installation configuration, and offers an interactive command line TUI for
guided setup.

#### NOTE
For detailed installation options and configuration, see
[ROCm Runfile Installer](https://rocm.docs.amd.com/en/latest/install/rocm-runfile-installer.html).

**w=graphics**

**os=ubuntu os=rhel**

Use the `amdgpu-install` script to install ROCm, the AMD GPU driver,
graphics components, and other packages. It simplifies installation by
automating GPU-specific and distro-specific package selection. The script
also runs post-installation checks and installs an uninstallation script,
allowing you to remove the entire ROCm stack with a single command.

---

## Prerequisites

**os=ubuntu os=debian os=rhel os=oracle-linux os=rocky-linux os=sles**

Before installing ROCm 7.14.0, ensure your system meets
all prerequisites. This includes installing the required dependencies and
configuring permissions for GPU access. To confirm that your system is
supported, see the [Compatibility matrix](https://rocm.docs.amd.com/en/latest/compatibility/compatibility-matrix.html).

**os=windows**

Before installing ROCm 7.14.0, ensure your system meets
all prerequisites. To confirm that your system is supported, see the
[Compatibility matrix](https://rocm.docs.amd.com/en/latest/compatibility/compatibility-matrix.html).

**os=ubuntu os=debian os=rhel os=rocky-linux os=oracle-linux os=sles**

**i=pkgman i=pip i=tar i=runfile**

**os=windows — Prepare Windows for ROCm installation**

1. Remove any existing HIP SDK for Windows installations and other
   conflicting AMD graphics software. To uninstall the HIP SDK using the
   GUI, navigate to the following screen:
   * Control Panel > Programs > Uninstall a program
2. Disable the following Windows security features as they can interfere
   with ROCm functionality:
   * Turn off WDAG (Windows Defender Application Guard)
     * Control Panel > Programs > Programs and Features > Turn Windows
       features on or off > **Clear** “Microsoft Defender Application
       Guard”
   * Turn off SAC (Smart App Control)
     * Settings > Privacy & security > Windows Security > App & browser
       control > Smart App Control settings > **Off**

**os=wsl**

**ubuntu-ver=26.04 — Install WSL2 and Ubuntu 26.04**

Install WSL2 and Ubuntu 26.04 on your Windows system. See [How to install Linux on Windows
with WSL2 (Microsoft Learn)](https://learn.microsoft.com/en-us/windows/wsl/install) for instructions.

Complete the following instructions in your WSL2 environment.

**ubuntu-ver=24.04 — Install WSL2 and Ubuntu 24.04**

Install WSL2 and Ubuntu 24.04 on your Windows system. See [How to install Linux on Windows
with WSL2 (Microsoft Learn)](https://learn.microsoft.com/en-us/windows/wsl/install) for instructions.

Complete the following instructions in your WSL2 environment.

**ubuntu-ver=22.04 — Install WSL2 and Ubuntu 22.04**

Install WSL2 and Ubuntu 22.04 on your Windows system. See [How to install Linux on Windows
with WSL2 (Microsoft Learn)](https://learn.microsoft.com/en-us/windows/wsl/install) for instructions.

Complete the following instructions in your WSL2 environment.

**fam=ryzen**

**i=pkgman i=pip i=tar**

**os=ubuntu os=wsl**

**ubuntu-ver=24.04 — Install the OEM kernel**

Ryzen APUs (gfx1150, gfx1151, gfx1152, gfx1153, and gfx1103) require the OEM
kernel 6.14 for Ubuntu 24.04. Use the following command to install it
using `apt`.

```bash
sudo apt update && sudo apt install linux-oem-24.04c
```

Reboot your system after installing the OEM kernel.

**w=graphics**

**os=ubuntu**

**ubuntu-ver=24.04 — Install the OEM kernel**

Ryzen APUs require the OEM kernel 6.14 for Ubuntu 24.04. Use the
following command to install it using `apt`.

```bash
sudo apt update && sudo apt install linux-oem-24.04c
```

Reboot your system after installing the OEM kernel.

**os=rhel — Register your Red Hat Enterprise Linux system**

Register your Red Hat Enterprise Linux (RHEL) system to enable access to Red
Hat repositories and ensure you’re able to download and install packages.

Run the following command to register your system:

**rhel-ver=10.2 rhel-ver=10.0**

```bash
subscription-manager register --username <username> --password <password>
```

**rhel-ver=9.8 rhel-ver=9.6 rhel-ver=9.4 rhel-ver=8.10**

```bash
subscription-manager register --username <username> --password <password>
subscription-manager attach --auto
```

**os=sles — Register your SUSE Linux Enterprise Server system**

Register your SUSE Linux Enterprise Server (SLES) system to enable access to
SUSE repositories and ensure you’re able to download and install packages.

Run the following command to register your system:

```bash
sudo SUSEConnect -r <REGCODE>
```

**os=rhel — Update your system**

After registering your system, update RHEL to the latest packages. This is
particularly important for newer hardware on older versions of RHEL.

Run the following command to update your system:

**rhel-ver=10.2**

```bash
sudo dnf update --releasever=10.2 --exclude=\*release\*
```

**rhel-ver=10.0**

```bash
sudo dnf update --releasever=10.0 --exclude=\*release\*
```

**rhel-ver=9.8**

```bash
sudo dnf update --releasever=9.8 --exclude=\*release\*
```

**rhel-ver=9.6**

```bash
sudo dnf update --releasever=9.6 --exclude=\*release\*
```

**rhel-ver=9.4**

```bash
sudo dnf update --releasever=9.4 --exclude=\*release\*
```

**rhel-ver=8.10**

```bash
sudo dnf update --releasever=8.10 --exclude=\*release\*
```

**os=sles — Update your system**

After registering your system, update SLES to the latest available packages.
This is particularly important for newer hardware on older versions of SLES.

Run the following command to update your system:

```bash
sudo zypper update
```

**w=graphics**

**os=rhel — Add additional package repositories**

ROCm installation packages depend on packages that aren’t included in
the default package repositories. Use the following command to add the
necessary repositories.

**rhel-ver=10.2 rhel-ver=10.0**

```bash
wget https://dl.fedoraproject.org/pub/epel/epel-release-latest-10.noarch.rpm
sudo rpm -ivh epel-release-latest-10.noarch.rpm
sudo dnf config-manager --enable codeready-builder-for-rhel-10-x86_64-rpms
```

**rhel-ver=9.8 rhel-ver=9.6 rhel-ver=9.4**

```bash
wget https://dl.fedoraproject.org/pub/epel/epel-release-latest-9.noarch.rpm
sudo rpm -ivh epel-release-latest-9.noarch.rpm
```

```bash
sudo dnf config-manager --enable codeready-builder-for-rhel-9-x86_64-rpms
```

**i=pkgman**

**os=oracle-linux — Update your system**

Update Oracle Linux to the latest available packages.

Run the following command to update your system:

**oracle-linux-ver=10**

```bash
sudo dnf update --releasever=10.1 --exclude=\*release\*
```

**oracle-linux-ver=9**

```bash
sudo dnf update --releasever=9.7 --exclude=\*release\*
```

**oracle-linux-ver=8**

```bash
sudo dnf update --releasever=8.10 --exclude=\*release\*
```

**os=rhel — Add additional package repositories**

ROCm installation packages depend on packages that aren’t included in
the default package repositories. Use the following command to add the
necessary repositories.

**rhel-ver=10.2 rhel-ver=10.0**

**rhel-ver=9.8 rhel-ver=9.6 rhel-ver=9.4**

```bash
wget https://dl.fedoraproject.org/pub/epel/epel-release-latest-9.noarch.rpm
sudo rpm -ivh epel-release-latest-9.noarch.rpm
```

```bash
sudo dnf config-manager --enable codeready-builder-for-rhel-9-x86_64-rpms
```

**rhel-ver=8.10**

```bash
wget https://dl.fedoraproject.org/pub/epel/epel-release-latest-8.noarch.rpm
sudo rpm -ivh epel-release-latest-8.noarch.rpm
```

```bash
sudo dnf config-manager --enable codeready-builder-for-rhel-8-x86_64-rpms
```

**os=oracle-linux — Add additional package repositories**

ROCm installation packages depend on packages that aren’t included in
the default package repositories. Use the following command to add the
necessary repositories.

**oracle-linux-ver=10**

```bash
wget https://dl.fedoraproject.org/pub/epel/epel-release-latest-10.noarch.rpm
sudo rpm -ivh epel-release-latest-10.noarch.rpm
sudo crb enable
```

**oracle-linux-ver=9**

```bash
wget https://dl.fedoraproject.org/pub/epel/epel-release-latest-9.noarch.rpm
sudo rpm -ivh epel-release-latest-9.noarch.rpm
sudo crb enable
```

**oracle-linux-ver=8**

```bash
wget https://dl.fedoraproject.org/pub/epel/epel-release-latest-8.noarch.rpm
sudo rpm -ivh epel-release-latest-8.noarch.rpm
sudo crb enable
```

**os=rocky-linux — Add additional package repositories**

ROCm installation packages depend on packages that aren’t included in
the default package repositories. Use the following command to add the
necessary repositories.

**rocky-linux-ver=9**

```bash
wget https://dl.fedoraproject.org/pub/epel/epel-release-latest-9.noarch.rpm
sudo rpm -ivh epel-release-latest-9.noarch.rpm
sudo crb enable
```

**os=ubuntu os=debian os=rhel os=oracle-linux os=rocky-linux os=sles os=wsl**

**w=graphics — Install additional packages**

Some ROCm tools require the `libatomic` and `libquadmath` libraries to run correctly. Install
them using your distribution’s package manager.

**os=ubuntu**

```bash
sudo apt install libatomic1 libquadmath0
```

**os=rhel**

```bash
sudo dnf install libatomic libquadmath
```

**i=pkgman i=pip i=tar — Install additional packages**

Some ROCm tools require the `libatomic` and `libquadmath` libraries to run correctly. Install
them using your distribution’s package manager.

**os=ubuntu os=debian**

```bash
sudo apt install libatomic1 libquadmath0
```

**fam=all**

**ubuntu-ver=24.04**

**os=wsl**

To build the ROCDXG library for WSL2, you’ll need GCC 11.4 or later and
CMake 3.15 or later.

```bash
sudo apt install libatomic1 libquadmath0 gcc g++ cmake
```

**os=rhel os=oracle-linux os=rocky-linux**

```bash
sudo dnf install libatomic libquadmath
```

**os=sles**

```bash
sudo zypper install libatomic1 libquadmath0
```

**i=pip**

**os=ubuntu**

**ubuntu-ver=26.04 — Install Python**

Install a supported Python version. For example, to install Python
3.14, run the following command:

```bash
sudo apt install python3.14 python3.14-venv
```

**ubuntu-ver=24.04 — Install Python**

Install a supported Python version. For example, to install Python
3.12, run the following command:

```bash
sudo apt install python3.12 python3.12-venv
```

**ubuntu-ver=22.04 — Install Python**

Install a supported Python version. For example, to install Python
3.11, run the following command:

```bash
sudo apt install python3.11 python3.11-venv
```

**os=debian**

**debian-ver=13 — Install Python**

Install a supported Python version. For example, to install Python
3.13, run the following command:

```bash
sudo apt install python3.13 python3.13-venv
```

**debian-ver=12 — Install Python**

Install a supported Python version. For example, to install Python
3.11, run the following command:

```bash
sudo apt install python3.11 python3.11-venv
```

**os=rhel**

**rhel-ver=10.2 rhel-ver=10.0 — Install Python**

Install a supported Python version. For example, to install Python
3.12, run the following command:

```bash
sudo dnf install python3.12 python3.12-pip
```

**rhel-ver=9.8 rhel-ver=9.6 rhel-ver=9.4 rhel-ver=9 rhel-ver=8.10 — Install Python**

Install a supported Python version. For example, to install Python
3.11, run the following command:

```bash
sudo dnf install python3.11 python3.11-pip
```

**os=oracle-linux**

**oracle-linux-ver=10 — Install Python**

Install a supported Python version. For example, to install Python
3.12, run the following command:

```bash
sudo dnf install python3.12 python3.12-pip
```

**oracle-linux-ver=9 oracle-linux-ver=8 — Install Python**

Install a supported Python version. For example, to install Python
3.11, run the following command:

```bash
sudo dnf install python3.11 python3.11-pip
```

**os=rocky-linux — Install Python**

Install a supported Python version. For example, to install Python
3.11, run the following command:

```bash
sudo dnf install python3.11 python3.11-pip
```

**os=sles — Install Python**

**sles-ver=16.0**

Install a supported Python version. For example, to install Python 3.13,
run the following command:

```bash
sudo zypper install -y python313 python313-pip
```

**sles-ver=15.7**

Install a supported Python version. For example, to install Python 3.11,
run the following command:

```bash
sudo zypper install -y python311 python311-pip
```

**os=windows — Install Python**

Install a supported Python version: 3.11, 3.12, 3.13, or 3.14. See [Python
Releases for Windows](https://www.python.org/downloads/windows/) for
installation details.

**i=pkgman i=pip i=tar**

**os=ubuntu os=debian os=rhel os=oracle-linux os=rocky-linux os=sles os=wsl — Configure permissions for GPU access**

There are two primary methods for configuring GPU access for ROCm: group
membership or udev rules. Each method has its own advantages. The choice
depends on your specific requirements and system management preferences.

**Group membership**

By default, GPU access is controlled by membership in the `video` and
`render` Linux system groups. The `video` group traditionally handles
video device access, while the `render` group manages GPU rendering
through DRM render nodes.

```bash
# Add the current user to the render and video groups
sudo usermod -a -G render,video $LOGNAME
```

**udev rules**

udev rules are a flexible, system-wide approach for managing device
permissions, eliminating the need for user group management while
allowing granular GPU access. To enable them and grant GPU access to
all users, run the following command:

```bash
sudo tee /etc/udev/rules.d/70-amdgpu.rules << EOF
KERNEL=="kfd", GROUP="render", MODE="0666"
SUBSYSTEM=="drm", KERNEL=="renderD*", GROUP="render", MODE="0666"
EOF

sudo udevadm control --reload-rules
sudo udevadm trigger
```

#### NOTE
To apply all settings, reboot your system.

**w=graphics**

**os=ubuntu os=debian os=rhel os=oracle-linux os=rocky-linux os=sles — Configure permissions for GPU access**

**Group membership**

```bash
# Add the current user to the render and video groups
sudo usermod -a -G render,video $LOGNAME
```

**udev rules**

```bash
sudo tee /etc/udev/rules.d/70-amdgpu.rules << EOF
KERNEL=="kfd", GROUP="render", MODE="0666"
SUBSYSTEM=="drm", KERNEL=="renderD*", GROUP="render", MODE="0666"
EOF

sudo udevadm control --reload-rules
sudo udevadm trigger
```

#### NOTE
To apply all settings, reboot your system.

**i=runfile — Quick start**

Download and launch the interactive installer to set up ROCm and/or the AMD
GPU Driver with guided, step-by-step configuration.

```bash
curl -fsSLO https://repo.radeon.com/rocm/installer/rocm-runfile-installer/rocm-rel-7.14/rocm-installer-7.14.0-7.run && bash rocm-installer-7.14.0-7.run
```

**i=runfile — Configuration options**

The following command line options are used to customize the runfile
installer, including dependency handling and GPU access configuration. For
recommended usage, go to [Installation](#rocm-install).

**i=runfile — Dependencies**

The runfile installer controls dependency installation via the `deps=`
argument.

| Command         | Description                                     |
|-----------------|-------------------------------------------------|
| `deps=install`  | Installs all required packages                  |
| `deps=list`     | Lists all required packages                     |
| `deps=validate` | Validates which required packages are installed |

Specify the target after the dependency argument; for example: `deps=install rocm`.

#### NOTE
It is recommended to include `deps=install` if you’re not sure what
dependencies are installed on your system.

**i=runfile — GPU access**

There are two primary methods of configuring GPU access for ROCm: group
membership or udev rules. The choice depends on your specific
requirements and system management preferences.

The runfile installer sets GPU access at install time using the `gpu-access=` argument.

| Argument          | Method                                                                  |
|-------------------|-------------------------------------------------------------------------|
| `gpu-access=user` | Group membership (adds the current user to the render and video groups) |
| `gpu-access=all`  | udev rules (configures system-wide GPU access)                          |

---

## Installation

**os=ubuntu ubuntu-ver=24.04 ubuntu-ver=22.04 i=pkgman i=runfile**

#### NOTE
If you have ROCm 7.2.4 or older installed, please uninstall it before proceeding.

**os=rhel rhel-ver=10.0 rhel-ver=9.6 rhel-ver=9.4 rhel-ver=8.10 i=pkgman i=runfile**

#### NOTE
If you have ROCm 7.2.4 or older installed, please uninstall it before proceeding.

**os=sles sles-ver=15.7 i=pkgman i=runfile**

#### NOTE
If you have ROCm 7.2.4 or older installed, please uninstall it before proceeding.

**os=debian os=oracle-linux os=rocky-linux os=windows i=pkgman i=runfile**

#### NOTE
If you have ROCm 7.2.4 or older installed, please uninstall it before proceeding.

**os=windows**

Before getting started, make sure you’ve completed the [Prerequisites](#rocm-prerequisites).
For information about supported operating systems and compatible AMD devices,
see the [Compatibility matrix](https://rocm.docs.amd.com/en/latest/compatibility/compatibility-matrix.html).

#### NOTE
Do not replace or copy the ROCm compiler and runtime DLLs to System32 as
this can cause conflicts.

**i=runfile — Download the runfile installer**

Use the following command to download the ROCm Runfile Installer.

```bash
wget https://repo.radeon.com/rocm/installer/rocm-runfile-installer/rocm-rel-7.14/rocm-installer-7.14.0-7.run
```

**w=graphics**

**os=ubuntu os=rhel — Install the amdgpu-install script**

Use the following commands to download and install the `amdgpu-install` script.

**os=ubuntu**

**ubuntu-ver=26.04**

```bash
sudo apt update
wget https://repo.radeon.com/amdgpu-install/31.40.1/ubuntu/resolute/amdgpu-install_31.40.1.314001-1_all.deb
sudo apt install ./amdgpu-install_31.40.1.314001-1_all.deb
```

**ubuntu-ver=24.04**

```bash
sudo apt update
wget https://repo.radeon.com/amdgpu-install/31.40.1/ubuntu/noble/amdgpu-install_31.40.1.314001-1_all.deb
sudo apt install ./amdgpu-install_31.40.1.314001-1_all.deb
```

**ubuntu-ver=22.04**

```bash
sudo apt update
wget https://repo.radeon.com/amdgpu-install/31.40.1/ubuntu/jammy/amdgpu-install_31.40.1.314001-1_all.deb
sudo apt install ./amdgpu-install_31.40.1.314001-1_all.deb
```

**os=rhel**

**rhel-ver=10.2**

```bash
wget https://repo.radeon.com/amdgpu-install/31.40.1/rhel/10.2/amdgpu-install-31.40.1.314001-1.el10.noarch.rpm
sudo dnf install ./amdgpu-install-31.40.1.314001-1.el10.noarch.rpm
```

**rhel-ver=10.0**

```bash
wget https://repo.radeon.com/amdgpu-install/31.40.1/rhel/10.0/amdgpu-install-31.40.1.314001-1.el10.noarch.rpm
sudo dnf install ./amdgpu-install-31.40.1.314001-1.el10.noarch.rpm
```

**rhel-ver=9.8**

```bash
wget https://repo.radeon.com/amdgpu-install/31.40.1/rhel/9.8/amdgpu-install-31.40.1.314001-1.el9.noarch.rpm
sudo dnf install ./amdgpu-install-31.40.1.314001-1.el9.noarch.rpm
```

**rhel-ver=9.6**

```bash
wget https://repo.radeon.com/amdgpu-install/31.40.1/rhel/9.6/amdgpu-install-31.40.1.314001-1.el9.noarch.rpm
sudo dnf install ./amdgpu-install-31.40.1.314001-1.el9.noarch.rpm
```

**rhel-ver=9.4**

```bash
wget https://repo.radeon.com/amdgpu-install/31.40.1/rhel/9.4/amdgpu-install-31.40.1.314001-1.el9.noarch.rpm
sudo dnf install ./amdgpu-install-31.40.1.314001-1.el9.noarch.rpm
```

**rhel-ver=8.10**

```bash
wget https://repo.radeon.com/amdgpu-install/31.40.1/rhel/8.10/amdgpu-install-31.40.1.314001-1.el8.noarch.rpm
sudo dnf install ./amdgpu-install-31.40.1.314001-1.el8.noarch.rpm
```

**os=ubuntu os=debian os=rhel os=oracle-linux os=rocky-linux os=sles**

**i=pkgman i=pip i=tar**

**fam=all — Install the kernel driver**

For information about AMD GPU Driver (amdgpu) compatibility, see the
[Compatibility matrix](https://rocm.docs.amd.com/en/latest/compatibility/compatibility-matrix.html).

**os=ubuntu**

For Instinct and Radeon devices, install the AMD GPU Driver (amdgpu).
See [Ubuntu native installation](https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-31.40.1/install/detailed-install/package-manager/package-manager-ubuntu.html)
in the AMD Instinct Data Center GPU Documentation.

**ubuntu-ver=26.04**

Supported Ryzen APUs require the inbox kernel driver included with
Ubuntu 26.04.

**ubuntu-ver=24.04**

Supported Ryzen APUs require the inbox kernel driver included with
Ubuntu 24.04.4.

**os=debian**

For Instinct and Radeon devices, install the AMD GPU Driver (amdgpu).
See [Debian native installation](https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-31.40.1/install/detailed-install/package-manager/package-manager-debian.html)
in the AMD Instinct Data Center GPU Documentation.

**os=rhel**

For Instinct and Radeon devices, install the AMD GPU Driver (amdgpu).
See [RHEL native installation](https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-31.40.1/install/detailed-install/package-manager/package-manager-rhel.html)
in the AMD Instinct Data Center GPU Documentation.

**os=oracle-linux**

For Instinct and Radeon devices, install the AMD GPU Driver (amdgpu).
See [Oracle Linux native installation](https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-31.40.1/install/detailed-install/package-manager/package-manager-ol.html)
in the AMD Instinct Data Center GPU Documentation.

**os=rocky-linux**

For Instinct and Radeon devices, install the AMD GPU Driver (amdgpu).
See [Rocky Linux native installation](https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-31.40.1/install/detailed-install/package-manager/package-manager-rl.html)
in the AMD Instinct Data Center GPU Documentation.

**os=sles**

For Instinct and Radeon devices, install the AMD GPU Driver (amdgpu).
See [SLES native installation](https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-31.40.1/install/detailed-install/package-manager/package-manager-sles.html)
in the AMD Instinct Data Center GPU Documentation.

**fam=instinct fam=radeon — Install the kernel driver**

For information about AMD GPU Driver (amdgpu) compatibility, see the
[Compatibility matrix](https://rocm.docs.amd.com/en/latest/compatibility/compatibility-matrix.html).

**os=ubuntu**

For instructions on installing the AMD GPU Driver (amdgpu), see [Ubuntu native
installation](https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-31.40.1/install/detailed-install/package-manager/package-manager-ubuntu.html)
in the AMD Instinct Data Center GPU Documentation.

**os=debian**

For instructions on installing the AMD GPU Driver (amdgpu), see [Debian native
installation](https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-31.40.1/install/detailed-install/package-manager/package-manager-debian.html)
in the AMD Instinct Data Center GPU Documentation.

**os=rhel**

For instructions on installing the AMD GPU Driver (amdgpu), see [RHEL native
installation](https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-31.40.1/install/detailed-install/package-manager/package-manager-rhel.html)
in the AMD Instinct Data Center GPU Documentation.

**os=oracle-linux**

For instructions on installing the AMD GPU Driver (amdgpu), see [Oracle Linux native
installation](https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-31.40.1/install/detailed-install/package-manager/package-manager-ol.html)
in the AMD Instinct Data Center GPU Documentation.

**os=rocky-linux**

For instructions on installing the AMD GPU Driver (amdgpu), see [Rocky Linux native
installation](https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-31.40.1/install/detailed-install/package-manager/package-manager-rl.html)
in the AMD Instinct Data Center GPU Documentation.

**os=sles**

For instructions on installing the AMD GPU Driver (amdgpu), see [SLES
native installation](https://instinct.docs.amd.com/projects/amdgpu-docs/en/docs-31.40.1/install/detailed-install/package-manager/package-manager-sles.html)
in the AMD Instinct Data Center GPU Documentation.

**fam=ryzen**

**os=ubuntu — About the kernel driver**

**ubuntu-ver=26.04**

Supported Ryzen APUs require the inbox kernel driver included with
Ubuntu 26.04.

**ubuntu-ver=24.04**

Supported Ryzen APUs require the inbox kernel driver included with
Ubuntu 24.04.4.

**i=runfile — Install the kernel driver**

For information about AMD GPU Driver (amdgpu) compatibility, see the
[Compatibility matrix](https://rocm.docs.amd.com/en/latest/compatibility/compatibility-matrix.html).

```bash
bash rocm-installer-7.14.0-7.run deps=install amdgpu
```

#### NOTE
Reboot your system after installing the AMD GPU Driver.

**os=windows os=wsl — Install AMD Software: Adrenalin Edition**

For details and the download link, see [https://www.amd.com/en/resources/support-articles/release-notes/RN-RAD-WIN-26-6-4.html#Downloads](https://www.amd.com/en/resources/support-articles/release-notes/RN-RAD-WIN-26-6-4.html#Downloads).

### Install ROCm

Use the following instructions to install ROCm packages on your system.

**i=pkgman — Register ROCm repositories**

**os=ubuntu os=wsl**

Register the ROCm repository with your system’s package manager. This lets you install and update
ROCm packages using `apt`.

**ubuntu-ver=26.04**

**fam=all**

```bash
# Download and install GPG key
sudo mkdir --parents --mode=0755 /etc/apt/keyrings
wget https://repo.amd.com/rocm/packages-multi-arch/gpg/rocm.gpg -O - | \
    gpg --dearmor | sudo tee /etc/apt/keyrings/amdrocm.gpg > /dev/null

sudo tee /etc/apt/sources.list.d/rocm.list << EOF
deb [arch=amd64 signed-by=/etc/apt/keyrings/amdrocm.gpg] https://repo.amd.com/rocm/packages-multi-arch/ubuntu2604 stable main
EOF

sudo apt update
```

**fam=instinct fam=radeon fam=ryzen**

```bash
# Download and install GPG key
sudo mkdir --parents --mode=0755 /etc/apt/keyrings

# ROCm release signing key
wget https://repo.amd.com/rocm/packages-multi-arch/gpg/rocm.gpg -O - | \
    gpg --dearmor | sudo tee /etc/apt/keyrings/amdrocm.gpg > /dev/null

sudo tee /etc/apt/sources.list.d/rocm.list << EOF
deb [arch=amd64 signed-by=/etc/apt/keyrings/amdrocm.gpg] https://repo.amd.com/rocm/packages-multi-arch/ubuntu2604 stable main
EOF

sudo apt update
```

**ubuntu-ver=24.04**

**fam=all**

sudo tee /etc/apt/sources.list.d/rocm.list << EOF
deb [arch=amd64 signed-by=/etc/apt/keyrings/amdrocm.gpg] https://repo.amd.com/rocm/packages-multi-arch/ubuntu2404 stable main
EOF

sudo apt update
```

**fam=instinct fam=radeon fam=ryzen**

```bash
# Download and install GPG key
sudo mkdir --parents --mode=0755 /etc/apt/keyrings

# ROCm release signing key
wget https://repo.amd.com/rocm/packages-multi-arch/gpg/rocm.gpg -O - | \
    gpg --dearmor | sudo tee /etc/apt/keyrings/amdrocm.gpg > /dev/null

sudo tee /etc/apt/sources.list.d/rocm.list << EOF
deb [arch=amd64 signed-by=/etc/apt/keyrings/amdrocm.gpg] https://repo.amd.com/rocm/packages-multi-arch/ubuntu2404 stable main
EOF

sudo apt update
```

**ubuntu-ver=22.04**

**fam=all**

sudo tee /etc/apt/sources.list.d/rocm.list << EOF
deb [arch=amd64 signed-by=/etc/apt/keyrings/amdrocm.gpg] https://repo.amd.com/rocm/packages-multi-arch/ubuntu2204 stable main
EOF

sudo apt update
```

**fam=instinct fam=radeon fam=ryzen**

```bash
# Download and install GPG key
sudo mkdir --parents --mode=0755 /etc/apt/keyrings

# ROCm release signing key
wget https://repo.amd.com/rocm/packages-multi-arch/gpg/rocm.gpg -O - | \
    gpg --dearmor | sudo tee /etc/apt/keyrings/amdrocm.gpg > /dev/null

sudo tee /etc/apt/sources.list.d/rocm.list << EOF
deb [arch=amd64 signed-by=/etc/apt/keyrings/amdrocm.gpg] https://repo.amd.com/rocm/packages-multi-arch/ubuntu2204 stable main
EOF

sudo apt update
```

**os=debian**

Register the ROCm repository with your system’s package manager. This enables
you to install and update ROCm packages using `apt`.

**debian-ver=13**

**fam=all**

sudo tee /etc/apt/sources.list.d/rocm.list << EOF
deb [arch=amd64 signed-by=/etc/apt/keyrings/amdrocm.gpg] https://repo.amd.com/rocm/packages-multi-arch/debian13 stable main
EOF

sudo apt update
```

**fam=instinct fam=radeon fam=ryzen**

sudo tee /etc/apt/sources.list.d/rocm.list << EOF
deb [arch=amd64 signed-by=/etc/apt/keyrings/amdrocm.gpg] https://repo.amd.com/rocm/packages-multi-arch/debian13 stable main
EOF

sudo apt update
```

**debian-ver=12**

**fam=all**

sudo tee /etc/apt/sources.list.d/rocm.list << EOF
deb [arch=amd64 signed-by=/etc/apt/keyrings/amdrocm.gpg] https://repo.amd.com/rocm/packages-multi-arch/debian12 stable main
EOF

sudo apt update
```

**fam=instinct fam=radeon fam=ryzen**

sudo tee /etc/apt/sources.list.d/rocm.list << EOF
deb [arch=amd64 signed-by=/etc/apt/keyrings/amdrocm.gpg] https://repo.amd.com/rocm/packages-multi-arch/debian12 stable main
EOF

sudo apt update
```

**os=rhel**

Register the ROCm repository with your system’s package manager. This enables
you to install and update ROCm packages using `dnf`.

**rhel-ver=10.2 rhel-ver=10.0**

**fam=all**

```bash
sudo tee /etc/yum.repos.d/rocm.repo <<EOF
[rocm]
name=ROCm 7.14.0
baseurl=https://repo.amd.com/rocm/packages-multi-arch/rhel10/x86_64
enabled=1
gpgcheck=1
gpgkey=https://repo.amd.com/rocm/packages-multi-arch/gpg/rocm.gpg
priority=50
EOF

sudo dnf clean all
```

**fam=instinct fam=radeon fam=ryzen**

sudo dnf clean all
```

**rhel-ver=9.8 rhel-ver=9.6 rhel-ver=9.4**

**fam=all**

```bash
sudo tee /etc/yum.repos.d/rocm.repo <<EOF
[rocm]
name=ROCm 7.14.0
baseurl=https://repo.amd.com/rocm/packages-multi-arch/rhel9/x86_64
enabled=1
gpgcheck=1
gpgkey=https://repo.amd.com/rocm/packages-multi-arch/gpg/rocm.gpg
priority=50
EOF

sudo dnf clean all
```

**fam=instinct fam=radeon fam=ryzen**

sudo dnf clean all
```

**rhel-ver=8.10**

**fam=all**

```bash
sudo tee /etc/yum.repos.d/rocm.repo <<EOF
[rocm]
name=ROCm 7.14.0
baseurl=https://repo.amd.com/rocm/packages-multi-arch/rhel8/x86_64
enabled=1
gpgcheck=1
gpgkey=https://repo.amd.com/rocm/packages-multi-arch/gpg/rocm.gpg
priority=50
EOF

sudo dnf clean all
```

**fam=instinct fam=radeon fam=ryzen**

sudo dnf clean all
```

**os=oracle-linux**

Register the ROCm repository with your system’s package manager. This enables
you to install and update ROCm packages using `dnf`.

**oracle-linux-ver=10**

**fam=all**

sudo dnf clean all
```

**fam=instinct fam=radeon fam=ryzen**

sudo dnf clean all
```

**oracle-linux-ver=9**

**fam=all**

sudo dnf clean all
```

**fam=instinct fam=radeon fam=ryzen**

sudo dnf clean all
```

**oracle-linux-ver=8**

**fam=all**

sudo dnf clean all
```

**fam=instinct fam=radeon fam=ryzen**

sudo dnf clean all
```

**os=rocky-linux**

Register the ROCm repository with your system’s package manager. This enables
you to install and update ROCm packages using `dnf`.

**fam=all**

sudo dnf clean all
```

**fam=instinct fam=radeon fam=ryzen**

sudo dnf clean all
```

**os=sles**

Register the ROCm repository with your system’s package manager. This enables
you to install and update ROCm packages using `zypper`.

**sles-ver=16.0**

**fam=all**

```bash
sudo tee /etc/zypp/repos.d/rocm.repo <<EOF
[rocm]
name=ROCm 7.14.0
baseurl=https://repo.amd.com/rocm/packages-multi-arch/sles16/x86_64
enabled=1
gpgcheck=1
gpgkey=https://repo.amd.com/rocm/packages-multi-arch/gpg/rocm.gpg
priority=50
EOF

sudo zypper --gpg-auto-import-keys refresh
```

**fam=instinct fam=radeon fam=ryzen**

sudo zypper --gpg-auto-import-keys refresh
```

**sles-ver=15.7**

**fam=all**

```bash
sudo tee /etc/zypp/repos.d/rocm.repo <<EOF
[rocm]
name=ROCm 7.14.0
baseurl=https://repo.amd.com/rocm/packages-multi-arch/sles15/x86_64
enabled=1
gpgcheck=1
gpgkey=https://repo.amd.com/rocm/packages-multi-arch/gpg/rocm.gpg
priority=50
EOF

sudo zypper --gpg-auto-import-keys refresh
```

**fam=instinct fam=radeon fam=ryzen**

sudo zypper --gpg-auto-import-keys refresh
```

**i=pkgman — Install ROCm packages**

**os=ubuntu os=debian os=wsl**

Use `apt` to install the core ROCm packages. See [ROCm meta
packages](#rocm-install-meta-packages) for additional installation
options.

**fam=all**

```bash
sudo apt install amdrocm7.14
```

**gfx=gfx950**

```bash
sudo apt install amdrocm7.14-gfx950
```

**gfx=gfx942**

```bash
sudo apt install amdrocm7.14-gfx942
```

**gfx=gfx90a**

```bash
sudo apt install amdrocm7.14-gfx90a
```

**gfx=gfx908**

```bash
sudo apt install amdrocm7.14-gfx908
```

**gfx=gfx1201**

```bash
sudo apt install amdrocm7.14-gfx1201
```

**gfx=gfx1200**

```bash
sudo apt install amdrocm7.14-gfx1200
```

**gfx=gfx1100**

```bash
sudo apt install amdrocm7.14-gfx1100
```

**gfx=gfx1101**

```bash
sudo apt install amdrocm7.14-gfx1101
```

**gfx=gfx1102**

```bash
sudo apt install amdrocm7.14-gfx1102
```

**gfx=gfx1103**

```bash
sudo apt install amdrocm7.14-gfx1103
```

**gfx=gfx1030**

```bash
sudo apt install amdrocm7.14-gfx1030
```

**gfx=gfx1151**

```bash
sudo apt install amdrocm7.14-gfx1151
```

**gfx=gfx1150**

```bash
sudo apt install amdrocm7.14-gfx1150
```

**gfx=gfx1152**

```bash
sudo apt install amdrocm7.14-gfx1152
```

**gfx=gfx1153**

```bash
sudo apt install amdrocm7.14-gfx1153
```

**os=rhel os=oracle-linux os=rocky-linux**

Use `dnf` to install the core ROCm packages. See [ROCm meta
packages](#rocm-install-meta-packages) for additional installation
options.

**fam=all**

```bash
sudo dnf install amdrocm7.14
```

**gfx=gfx950**

```bash
sudo dnf install amdrocm7.14-gfx950
```

**gfx=gfx942**

```bash
sudo dnf install amdrocm7.14-gfx942
```

**gfx=gfx90a**

```bash
sudo dnf install amdrocm7.14-gfx90a
```

**gfx=gfx908**

```bash
sudo dnf install amdrocm7.14-gfx908
```

**gfx=gfx1201**

```bash
sudo dnf install amdrocm7.14-gfx1201
```

**gfx=gfx1200**

```bash
sudo dnf install amdrocm7.14-gfx1200
```

**gfx=gfx1100**

```bash
sudo dnf install amdrocm7.14-gfx1100
```

**gfx=gfx1101**

```bash
sudo dnf install amdrocm7.14-gfx1101
```

**gfx=gfx1102**

```bash
sudo dnf install amdrocm7.14-gfx1102
```

**gfx=gfx1103**

```bash
sudo dnf install amdrocm7.14-gfx1103
```

**gfx=gfx1030**

```bash
sudo dnf install amdrocm7.14-gfx1030
```

**gfx=gfx1151**

```bash
sudo dnf install amdrocm7.14-gfx1151
```

**gfx=gfx1150**

```bash
sudo dnf install amdrocm7.14-gfx1150
```

**gfx=gfx1152**

```bash
sudo dnf install amdrocm7.14-gfx1152
```

**gfx=gfx1153**

```bash
sudo dnf install amdrocm7.14-gfx1153
```

**os=sles**

Use `zypper` to install the core ROCm packages. See [ROCm meta
packages](#rocm-install-meta-packages) for additional installation
options.

**fam=all**

```bash
sudo zypper install amdrocm7.14
```

**gfx=gfx950**

```bash
sudo zypper install amdrocm7.14-gfx950
```

**gfx=gfx942**

```bash
sudo zypper install amdrocm7.14-gfx942
```

**gfx=gfx90a**

```bash
sudo zypper install amdrocm7.14-gfx90a
```

**gfx=gfx908**

```bash
sudo zypper install amdrocm7.14-gfx908
```

**i=pkgman — ROCm meta packages**

Meta packages group related components and dependencies together, allowing
you to install only what is necessary for your use case. The following table
describes available ROCm meta packages:

| Meta package name | Use case | Description | Contents | | | | | | | | | | | | | | | | |
|----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|---------------------------------------------------------|-----------------------------------------------------------------------|-----------------------------------------------------------------------------------------------|-----------------------------------------------------------------------------------------------|-----------------------------------------------------------------------------------------------|-----------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------|----------------------------------------------------------------------------------------|
| `amdrocm7.14-gfx950` | `amdrocm7.14-gfx942` | `amdrocm7.14-gfx90a` | `amdrocm7.14-gfx908` | `amdrocm7.14-gfx1201` | `amdrocm7.14-gfx1200` | `amdrocm7.14-gfx1100` | `amdrocm7.14-gfx1101` | `amdrocm7.14-gfx1102` | `amdrocm7.14-gfx1103` | `amdrocm7.14-gfx1030` | `amdrocm7.14-gfx1151` | `amdrocm7.14-gfx1150` | `amdrocm7.14-gfx1152` | `amdrocm7.14-gfx1153` | `amdrocm7.14` | ROCm Base | Core runtime environment. Install this to run ROCm applications. | Runtimes, libraries, system control and monitoring tools, and other essential components. | |
| **gfx=gfx950** `amdrocm-core-dev7.14-gfx950` **gfx=gfx942** `amdrocm-core-dev7.14-gfx942` **gfx=gfx90a** `amdrocm-core-dev7.14-gfx90a` **gfx=gfx908** `amdrocm-core-dev7.14-gfx908` **gfx=gfx1201** `amdrocm-core-dev7.14-gfx1201` **gfx=gfx1200** `amdrocm-core-dev7.14-gfx1200` **gfx=gfx1100** `amdrocm-core-dev7.14-gfx1100` **gfx=gfx1101** `amdrocm-core-dev7.14-gfx1101` **gfx=gfx1102** `amdrocm-core-dev7.14-gfx1102` **gfx=gfx1103** `amdrocm-core-dev7.14-gfx1103` **gfx=gfx1030** `amdrocm-core-dev7.14-gfx1030` **gfx=gfx1151** `amdrocm-core-dev7.14-gfx1151` **gfx=gfx1150** `amdrocm-core-dev7.14-gfx1150` **gfx=gfx1152** `amdrocm-core-dev7.14-gfx1152` **gfx=gfx1153** `amdrocm-core-dev7.14-gfx1153` **fam=all** `amdrocm-core-dev7.14` | **gfx=gfx950** `amdrocm-core-devel7.14-gfx950` **gfx=gfx942** `amdrocm-core-devel7.14-gfx942` **gfx=gfx90a** `amdrocm-core-devel7.14-gfx90a` **gfx=gfx908** `amdrocm-core-devel7.14-gfx908` **gfx=gfx1201** `amdrocm-core-devel7.14-gfx1201` **gfx=gfx1200** `amdrocm-core-devel7.14-gfx1200` **gfx=gfx1100** `amdrocm-core-devel7.14-gfx1100` **gfx=gfx1101** `amdrocm-core-devel7.14-gfx1101` **gfx=gfx1102** `amdrocm-core-devel7.14-gfx1102` **gfx=gfx1103** `amdrocm-core-devel7.14-gfx1103` **gfx=gfx1030** `amdrocm-core-devel7.14-gfx1030` **gfx=gfx1151** `amdrocm-core-devel7.14-gfx1151` **gfx=gfx1150** `amdrocm-core-devel7.14-gfx1150` **gfx=gfx1152** `amdrocm-core-devel7.14-gfx1152` **gfx=gfx1153** `amdrocm-core-devel7.14-gfx1153` **fam=all** `amdrocm-core-devel7.14` | ROCm Developer Essentials | Development environment. Install this to build ROCm applications. | `amdrocm7.14-gfx950` plus compilers, CMake configurations, static library files, and headers. | `amdrocm7.14-gfx942` plus compilers, CMake configurations, static library files, and headers. | `amdrocm7.14-gfx90a` plus compilers, CMake configurations, static library files, and headers. | `amdrocm7.14-gfx908` plus compilers, CMake configurations, static library files, and headers. | `amdrocm7.14-gfx1201` plus compilers, CMake configurations, static library files, and headers. | `amdrocm7.14-gfx1200` plus compilers, CMake configurations, static library files, and headers. | `amdrocm7.14-gfx1100` plus compilers, CMake configurations, static library files, and headers. | `amdrocm7.14-gfx1101` plus compilers, CMake configurations, static library files, and headers. | `amdrocm7.14-gfx1102` plus compilers, CMake configurations, static library files, and headers. | `amdrocm7.14-gfx1103` plus compilers, CMake configurations, static library files, and headers. | `amdrocm7.14-gfx1030` plus compilers, CMake configurations, static library files, and headers. | `amdrocm7.14-gfx1151` plus compilers, CMake configurations, static library files, and headers. | `amdrocm7.14-gfx1150` plus compilers, CMake configurations, static library files, and headers. | `amdrocm7.14-gfx1152` plus compilers, CMake configurations, static library files, and headers. | `amdrocm7.14-gfx1153` plus compilers, CMake configurations, static library files, and headers. | `amdrocm7.14` plus compilers, CMake configurations, static library files, and headers. |
| `amdrocm-developer-tools7.14` | ROCm Profiler | Install this to profile and optimize ROCm applications. | Profilers and related tools. | | | | | | | | | | | | | | | | |
| `amdrocm-opencl7.14` | ROCm OpenCL | Install this to run OpenCL applications on ROCm. | Components needed to run OpenCL. | | | | | | | | | | | | | | | | |
| `amdrocm-core-sdk7.14-gfx950` | `amdrocm-core-sdk7.14-gfx942` | `amdrocm-core-sdk7.14-gfx90a` | `amdrocm-core-sdk7.14-gfx908` | `amdrocm-core-sdk7.14-gfx1201` | `amdrocm-core-sdk7.14-gfx1200` | `amdrocm-core-sdk7.14-gfx1100` | `amdrocm-core-sdk7.14-gfx1101` | `amdrocm-core-sdk7.14-gfx1102` | `amdrocm-core-sdk7.14-gfx1103` | `amdrocm-core-sdk7.14-gfx1030` | `amdrocm-core-sdk7.14-gfx1151` | `amdrocm-core-sdk7.14-gfx1150` | `amdrocm-core-sdk7.14-gfx1152` | `amdrocm-core-sdk7.14-gfx1153` | `amdrocm-core-sdk7.14` | ROCm Full Suite | Install this if you need everything. | The complete ROCm Core SDK including runtimes, compilers, development tools, and dependencies. | |

**w=graphics**

**os=ubuntu os=rhel**

**fam=radeon**

#### WARNING
Before proceeding, please enable “Above 4G Decoding” in your BIOS settings.

Run the `amdgpu-install` script with the following `--usecase` arguments
to install ROCm, graphics and amdgpu driver packages.

```bash
sudo amdgpu-install --usecase=rocm,graphics --gfxversion=auto
```

**fam=ryzen**

Run the `amdgpu-install` script with the following `--usecase`
arguments to install ROCm packages. Ryzen APUs require the inbox
kernel driver included with Ubuntu – to skip installing the AMD GPU
driver, add `--no-dkms`.

```bash
sudo amdgpu-install --usecase=rocm --gfxversion=auto --no-dkms
```

**fam=all**

Run the `amdgpu-install` script with the following `--usecase` arguments
to install ROCm and graphics packages.

```bash
sudo amdgpu-install --usecase=rocm,graphics --gfxversion=all
```

Reboot your system after installing.

**i=pip — Set up your Python virtual environment**

Create and activate the Python virtual environment where you’ll install
ROCm packages.

**os=ubuntu**

**ubuntu-ver=26.04**

For example, to create and activate a Python 3.14 virtual environment,
run the following command:

```bash
python3.14 -m venv .venv
source .venv/bin/activate
```

**ubuntu-ver=24.04**

For example, to create and activate a Python 3.12 virtual environment,
run the following command:

```bash
python3.12 -m venv .venv
source .venv/bin/activate
```

**ubuntu-ver=22.04**

For example, to create and activate a Python 3.11 virtual environment,
run the following command:

```bash
python3.11 -m venv .venv
source .venv/bin/activate
```

**os=debian**

**debian-ver=13**

For example, to create and activate a Python 3.13 virtual environment,
run the following command:

```bash
python3.13 -m venv .venv
source .venv/bin/activate
```

**debian-ver=12**

For example, to create and activate a Python 3.11 virtual environment,
run the following command:

```bash
python3.11 -m venv .venv
source .venv/bin/activate
```

**os=rhel**

**rhel-ver=10.2 rhel-ver=10.0**

For example, to create and activate a Python 3.12 virtual environment,
run the following command:

```bash
python3.12 -m venv .venv
source .venv/bin/activate
```

**rhel-ver=9.8 rhel-ver=9.6 rhel-ver=9.4 rhel-ver=8.10**

For example, to create and activate a Python 3.11 virtual environment,
run the following command:

```bash
python3.11 -m venv .venv
source .venv/bin/activate
```

**os=oracle-linux**

**oracle-linux-ver=10**

For example, to create and activate a Python 3.12 virtual environment,
run the following command:

```bash
python3.12 -m venv .venv
source .venv/bin/activate
```

**oracle-linux-ver=9**

For example, to create and activate a Python 3.11 virtual environment,
run the following command:

```bash
python3.11 -m venv .venv
source .venv/bin/activate
```

**oracle-linux-ver=8**

For example, to create and activate a Python 3.11 virtual environment,
run the following command:

```bash
python3.11 -m venv .venv
source .venv/bin/activate
```

**os=rocky-linux**

For example, to create and activate a Python 3.11 virtual environment,
run the following command:

```bash
python3.11 -m venv .venv
source .venv/bin/activate
```

**os=sles**

**sles-ver=16.0**

For example, to create and activate a Python 3.13 virtual environment,
run the following command:

```bash
python3.13 -m venv .venv
source .venv/bin/activate
```

**sles-ver=15.7**

For example, to create and activate a Python 3.11 virtual environment,
run the following command:

```bash
python3.11 -m venv .venv
source .venv/bin/activate
```

**os=windows**

For example, to create and activate a Python 3.12 virtual environment,
run the following command:

```bat
py -3.12 -m venv .venv
.venv\Scripts\activate
```

**i=pip — Install ROCm wheel packages**

**fam=all**

Use pip to install the ROCm libraries and development tools.

Run the following command:

```bash
python -m pip install --index-url https://repo.amd.com/rocm/whl-multi-arch/ "rocm[libraries,device-all]==7.14.0"
```

The table below lists the available packages. Each bracketed name is an
optional *extra* of the `rocm` meta package — combine the ones you need
as a comma-separated list (for example, `rocm[libraries,devel]`).

| Package               | Contents                                                                                                | Use case                                              |                       |                        |                        |                        |                        |                        |                        |                        |                        |                        |                        |                    |                                                    |                                                                 |
|-----------------------|---------------------------------------------------------------------------------------------------------|-------------------------------------------------------|-----------------------|------------------------|------------------------|------------------------|------------------------|------------------------|------------------------|------------------------|------------------------|------------------------|------------------------|--------------------|----------------------------------------------------|-----------------------------------------------------------------|
| `rocm`                | Core SDK: runtime, HIP, compiler, utility tools, and profiling SDK (rocprofiler-sdk, rocprofv3, roctx). | Required by all ROCm users.                           |                       |                        |                        |                        |                        |                        |                        |                        |                        |                        |                        |                    |                                                    |                                                                 |
| `rocm[libraries]`     | Pre-built math and ML host libraries.                                                                   | Required for ML frameworks such as PyTorch and JAX.   |                       |                        |                        |                        |                        |                        |                        |                        |                        |                        |                        |                    |                                                    |                                                                 |
| `rocm[devel]`         | Compilers, CMake configuration, headers, and static libraries.                                          | Building ROCm applications.                           |                       |                        |                        |                        |                        |                        |                        |                        |                        |                        |                        |                    |                                                    |                                                                 |
| `rocm[device-gfx950]` | `rocm[device-gfx942]`                                                                                   | `rocm[device-gfx90a]`                                 | `rocm[device-gfx908]` | `rocm[device-gfx1201]` | `rocm[device-gfx1200]` | `rocm[device-gfx1100]` | `rocm[device-gfx1101]` | `rocm[device-gfx1102]` | `rocm[device-gfx1103]` | `rocm[device-gfx1030]` | `rocm[device-gfx1151]` | `rocm[device-gfx1150]` | `rocm[device-gfx1152]` | `rocm[device-all]` | Pre-compiled GPU kernels for the specified target. | Required to run GPU workloads; installed alongside `libraries`. |
| `rocm[profiler]`      | Profiling tools: rocprofiler-systems and rocprofiler-compute.                                           | Optional; analyzing and optimizing ROCm applications. |                       |                        |                        |                        |                        |                        |                        |                        |                        |                        |                        |                    |                                                    |                                                                 |

**gfx=gfx950**

Use pip to install the ROCm libraries and development tools for
your `gfx950` GPU.

Run the following command:

```bash
python -m pip install --index-url https://repo.amd.com/rocm/whl-multi-arch/ "rocm[libraries,device-gfx950]==7.14.0"
```

**gfx=gfx942**

Use pip to install the ROCm libraries and development tools for
your `gfx942` device.

Run the following command:

```bash
python -m pip install --index-url https://repo.amd.com/rocm/whl-multi-arch/ "rocm[libraries,device-gfx942]==7.14.0"
```

**gfx=gfx90a**

Use pip to install the ROCm libraries and development tools.

Run the following command:

```bash
python -m pip install --index-url https://repo.amd.com/rocm/whl-multi-arch/ "rocm[libraries,device-gfx90a]==7.14.0"
```

**gfx=gfx908**

Use pip to install the ROCm libraries and development tools.

Run the following command:

```bash
python -m pip install --index-url https://repo.amd.com/rocm/whl-multi-arch/ "rocm[libraries,device-gfx908]==7.14.0"
```

**gfx=gfx1201**

Use pip to install the ROCm libraries and development tools for
your `gfx1201` GPU.

Run the following command:

```bash
python -m pip install --index-url https://repo.amd.com/rocm/whl-multi-arch/ "rocm[libraries,device-gfx1201]==7.14.0"
```

**gfx=gfx1200**

Use pip to install the ROCm libraries and development tools for
your `gfx1200` GPU.

Run the following command:

```bash
python -m pip install --index-url https://repo.amd.com/rocm/whl-multi-arch/ "rocm[libraries,device-gfx1200]==7.14.0"
```

**gfx=gfx1100**

Use pip to install the ROCm libraries and development tools.

Run the following command:

```bash
python -m pip install --index-url https://repo.amd.com/rocm/whl-multi-arch/ "rocm[libraries,device-gfx1100]==7.14.0"
```

**gfx=gfx1101**

Use pip to install the ROCm libraries and development tools.

Run the following command:

```bash
python -m pip install --index-url https://repo.amd.com/rocm/whl-multi-arch/ "rocm[libraries,device-gfx1101]==7.14.0"
```

**gfx=gfx1102**

Use pip to install the ROCm libraries and development tools.

Run the following command:

```bash
python -m pip install --index-url https://repo.amd.com/rocm/whl-multi-arch/ "rocm[libraries,device-gfx1102]==7.14.0"
```

**gfx=gfx1103**

Use pip to install the ROCm libraries and development tools.

Run the following command:

```bash
python -m pip install --index-url https://repo.amd.com/rocm/whl-multi-arch/ "rocm[libraries,device-gfx1103]==7.14.0"
```

**gfx=gfx1030**

Use pip to install the ROCm libraries and development tools.

Run the following command:

```bash
python -m pip install --index-url https://repo.amd.com/rocm/whl-multi-arch/ "rocm[libraries,device-gfx1030]==7.14.0"
```

**gfx=gfx1151**

Use pip to install the ROCm libraries and development tools for
your `gfx1151` Ryzen APU.

Run the following command:

```bash
python -m pip install --index-url https://repo.amd.com/rocm/whl-multi-arch/ "rocm[libraries,device-gfx1151]==7.14.0"
```

**gfx=gfx1150**

Use pip to install the ROCm libraries and development tools for
your `gfx1150` Ryzen APU.

Run the following command:

```bash
python -m pip install --index-url https://repo.amd.com/rocm/whl-multi-arch/ "rocm[libraries,device-gfx1150]==7.14.0"
```

**gfx=gfx1152**

Use pip to install the ROCm libraries and development tools for
your `gfx1152` Ryzen APU.

Run the following command:

```bash
python -m pip install --index-url https://repo.amd.com/rocm/whl-multi-arch/ "rocm[libraries,device-gfx1152]==7.14.0"
```

**gfx=gfx1153**

Use pip to install the ROCm libraries and development tools for
your `gfx1153` Ryzen APU.

Run the following command:

```bash
python -m pip install --index-url https://repo.amd.com/rocm/whl-multi-arch/ "rocm[libraries,device-gfx1153]==7.14.0"
```

**i=tar — Create the installation directory**

**os=ubuntu os=debian os=rhel os=oracle-linux os=rocky-linux os=sles**

Run the following command in your desired location to create your
installation directory:

```bash
mkdir therock-tarball && cd therock-tarball
```

#### IMPORTANT
Subsequent commands assume you’re working with the `therock-tarball`
directory. If you choose a different directory name, adjust the
commands accordingly.

**os=windows**

Create the installation directory in `C:\TheRock\build`. For example,
use the following command in your command prompt:

```bat
mkdir C:\TheRock\build
```

#### IMPORTANT
Subsequent commands assume you’re working with the
`C:\TheRock\build` directory. If you choose a different directory
name, adjust the commands accordingly.

**i=tar — Download and unpack the tarball**

**os=ubuntu os=debian os=rhel os=oracle-linux os=rocky-linux os=sles**

**fam=all**

Use the following commands to download and untar the ROCm tarball.

```bash
wget https://repo.amd.com/rocm/tarball-multi-arch/therock-dist-linux-multiarch-7.14.0.tar.gz
mkdir install
tar -xf *.tar.gz -C install
```

**gfx=gfx950**

Use the following commands to download and untar the ROCm tarball for
your `gfx950` GPU.

```bash
wget https://repo.amd.com/rocm/tarball-multi-arch/therock-dist-linux-gfx950-dcgpu-7.14.0.tar.gz
mkdir install
tar -xf *.tar.gz -C install
```

**gfx=gfx942**

Use the following commands to download and untar the ROCm tarball for
your `gfx942` GPU.

```bash
wget https://repo.amd.com/rocm/tarball-multi-arch/therock-dist-linux-gfx94X-dcgpu-7.14.0.tar.gz
mkdir install
tar -xf *.tar.gz -C install
```

**gfx=gfx90a**

Use the following commands to download and untar the ROCm tarball.

```bash
wget https://repo.amd.com/rocm/tarball-multi-arch/therock-dist-linux-gfx90a-7.14.0.tar.gz
mkdir install
tar -xf *.tar.gz -C install
```

**gfx=gfx908**

Use the following commands to download and untar the ROCm tarball.

```bash
wget https://repo.amd.com/rocm/tarball-multi-arch/therock-dist-linux-gfx908-7.14.0.tar.gz
mkdir install
tar -xf *.tar.gz -C install
```

**gfx=gfx1201 gfx=gfx1200**

Use the following commands to download and untar the ROCm tarball.

```bash
wget https://repo.amd.com/rocm/tarball-multi-arch/therock-dist-linux-gfx120X-all-7.14.0.tar.gz
mkdir install
tar -xf *.tar.gz -C install
```

**gfx=gfx1100 gfx=gfx1101 gfx=gfx1102 gfx=gfx1103**

Use the following commands to download and untar the ROCm tarball.

```bash
wget https://repo.amd.com/rocm/tarball-multi-arch/therock-dist-linux-gfx110X-all-7.14.0.tar.gz
mkdir install
tar -xf *.tar.gz -C install
```

**gfx=gfx1030**

Use the following commands to download and untar the ROCm tarball.

```bash
wget https://repo.amd.com/rocm/tarball-multi-arch/therock-dist-linux-gfx103X-all-7.14.0.tar.gz
mkdir install
tar -xf *.tar.gz -C install
```

**gfx=gfx1151**

Use the following commands to download and untar the ROCm tarball for
your `gfx1151` GPU.

```bash
wget https://repo.amd.com/rocm/tarball-multi-arch/therock-dist-linux-gfx1151-7.14.0.tar.gz
mkdir install
tar -xf *.tar.gz -C install
```

**gfx=gfx1150**

Use the following commands to download and untar the ROCm tarball for
your `gfx1150` GPU.

```bash
wget https://repo.amd.com/rocm/tarball-multi-arch/therock-dist-linux-gfx1150-7.14.0.tar.gz
mkdir install
tar -xf *.tar.gz -C install
```

**gfx=gfx1152**

Use the following commands to download and untar the ROCm tarball for
your `gfx1152` GPU.

```bash
wget https://repo.amd.com/rocm/tarball-multi-arch/therock-dist-linux-gfx1152-7.14.0.tar.gz
mkdir install
tar -xf *.tar.gz -C install
```

**gfx=gfx1153**

Use the following commands to download and untar the ROCm tarball for
your `gfx1153` GPU.

```bash
wget https://repo.amd.com/rocm/tarball-multi-arch/therock-dist-linux-gfx1153-7.14.0.tar.gz
mkdir install
tar -xf *.tar.gz -C install
```

**os=windows**

Download the tarball and extract the contents to `C:\TheRock\build`.
Run the following commands in your command prompt:

**fam=all**

```bat
cd C:\TheRock
curl -o therock-dist-windows-multiarch-7.14.0.tar.gz https://repo.amd.com/rocm/tarball-multi-arch/therock-dist-windows-multiarch-7.14.0.tar.gz
tar -xzf therock-dist-windows-multiarch-7.14.0.tar.gz -C build --strip-components=1
```

- Download link: [therock-dist-windows-multiarch-7.14.0.tar.gz](https://repo.amd.com/rocm/tarball-multi-arch/therock-dist-windows-multiarch-7.14.0.tar.gz)

**gfx=gfx1201 gfx=gfx1200**

```bat
cd C:\TheRock
curl -o therock-dist-windows-gfx120X-all-7.14.0.tar.gz https://repo.amd.com/rocm/tarball-multi-arch/therock-dist-windows-gfx120X-all-7.14.0.tar.gz
tar -xzf therock-dist-windows-gfx120X-all-7.14.0.tar.gz -C build --strip-components=1
```

- Download link: [therock-dist-windows-gfx120X-all-7.14.0.tar.gz](https://repo.amd.com/rocm/tarball-multi-arch/therock-dist-windows-gfx120X-all-7.14.0.tar.gz)

**gfx=gfx1100 gfx=gfx1101 gfx=gfx1102 gfx=gfx1103**

```bat
cd C:\TheRock
curl -o therock-dist-windows-gfx110X-all-7.14.0.tar.gz https://repo.amd.com/rocm/tarball-multi-arch/therock-dist-windows-gfx110X-all-7.14.0.tar.gz
tar -xzf therock-dist-windows-gfx110X-all-7.14.0.tar.gz -C build --strip-components=1
```

- Download link: [therock-dist-windows-gfx110X-all-7.14.0.tar.gz](https://repo.amd.com/rocm/tarball-multi-arch/therock-dist-windows-gfx110X-all-7.14.0.tar.gz)

**gfx=gfx1030**

```bat
cd C:\TheRock
curl -o therock-dist-windows-gfx103X-all-7.14.0.tar.gz https://repo.amd.com/rocm/tarball-multi-arch/therock-dist-windows-gfx103X-all-7.14.0.tar.gz
tar -xzf therock-dist-windows-gfx103X-all-7.14.0.tar.gz -C build --strip-components=1
```

- Download link: [therock-dist-windows-gfx103X-all-7.14.0.tar.gz](https://repo.amd.com/rocm/tarball-multi-arch/therock-dist-windows-gfx103X-all-7.14.0.tar.gz)

**gfx=gfx1151**

```bat
cd C:\TheRock
curl -o therock-dist-windows-gfx1151-7.14.0.tar.gz https://repo.amd.com/rocm/tarball-multi-arch/therock-dist-windows-gfx1151-7.14.0.tar.gz
tar -xzf therock-dist-windows-gfx1151-7.14.0.tar.gz -C build --strip-components=1
```

- Download link: [therock-dist-windows-gfx1151-7.14.0.tar.gz](https://repo.amd.com/rocm/tarball-multi-arch/therock-dist-windows-gfx1151-7.14.0.tar.gz)

**gfx=gfx1150**

```bat
cd C:\TheRock
curl -o therock-dist-windows-gfx1150-7.14.0.tar.gz https://repo.amd.com/rocm/tarball-multi-arch/therock-dist-windows-gfx1150-7.14.0.tar.gz
tar -xzf therock-dist-windows-gfx1150-7.14.0.tar.gz -C build --strip-components=1
```

- Download link: [therock-dist-windows-gfx1150-7.14.0.tar.gz](https://repo.amd.com/rocm/tarball-multi-arch/therock-dist-windows-gfx1150-7.14.0.tar.gz)

**gfx=gfx1152**

```bat
cd C:\TheRock
curl -o therock-dist-windows-gfx1152-7.14.0.tar.gz https://repo.amd.com/rocm/tarball-multi-arch/therock-dist-windows-gfx1152-7.14.0.tar.gz
tar -xzf therock-dist-windows-gfx1152-7.14.0.tar.gz -C build --strip-components=1
```

- Download link: [therock-dist-windows-gfx1152-7.14.0.tar.gz](https://repo.amd.com/rocm/tarball-multi-arch/therock-dist-windows-gfx1152-7.14.0.tar.gz)

**gfx=gfx1153**

```bat
cd C:\TheRock
curl -o therock-dist-windows-gfx1153-7.14.0.tar.gz https://repo.amd.com/rocm/tarball-multi-arch/therock-dist-windows-gfx1153-7.14.0.tar.gz
tar -xzf therock-dist-windows-gfx1153-7.14.0.tar.gz -C build --strip-components=1
```

- Download link: [therock-dist-windows-gfx1153-7.14.0.tar.gz](https://repo.amd.com/rocm/tarball-multi-arch/therock-dist-windows-gfx1153-7.14.0.tar.gz)

**i=runfile**

Install the `core` ROCm components. See [ROCm meta components](#rocm-install-runfile-meta-components) for additional installation options.

**fam=all**

```bash
bash rocm-installer-7.14.0-7.run deps=install rocm gfx=all gpu-access=user
```

**gfx=gfx950**

```bash
bash rocm-installer-7.14.0-7.run deps=install rocm gfx=gfx950 gpu-access=user
```

**gfx=gfx942**

```bash
bash rocm-installer-7.14.0-7.run deps=install rocm gfx=gfx942 gpu-access=user
```

**gfx=gfx90a**

```bash
bash rocm-installer-7.14.0-7.run deps=install rocm gfx=gfx90a gpu-access=user
```

**gfx=gfx908**

```bash
bash rocm-installer-7.14.0-7.run deps=install rocm gfx=gfx908 gpu-access=user
```

**gfx=gfx1201**

```bash
bash rocm-installer-7.14.0-7.run deps=install rocm gfx=gfx1201 gpu-access=user
```

**gfx=gfx1200**

```bash
bash rocm-installer-7.14.0-7.run deps=install rocm gfx=gfx1200 gpu-access=user
```

**gfx=gfx1100**

```bash
bash rocm-installer-7.14.0-7.run deps=install rocm gfx=gfx1100 gpu-access=user
```

**gfx=gfx1101**

```bash
bash rocm-installer-7.14.0-7.run deps=install rocm gfx=gfx1101 gpu-access=user
```

**gfx=gfx1102**

```bash
bash rocm-installer-7.14.0-7.run deps=install rocm gfx=gfx1102 gpu-access=user
```

**gfx=gfx1103**

```bash
bash rocm-installer-7.14.0-7.run deps=install rocm gfx=gfx1103 gpu-access=user
```

**gfx=gfx1030**

```bash
bash rocm-installer-7.14.0-7.run deps=install rocm gfx=gfx1030 gpu-access=user
```

**gfx=gfx1151**

```bash
bash rocm-installer-7.14.0-7.run deps=install rocm gfx=gfx1151 gpu-access=user
```

**gfx=gfx1150**

```bash
bash rocm-installer-7.14.0-7.run deps=install rocm gfx=gfx1150 gpu-access=user
```

**gfx=gfx1152**

```bash
bash rocm-installer-7.14.0-7.run deps=install rocm gfx=gfx1152 gpu-access=user
```

**gfx=gfx1153**

```bash
bash rocm-installer-7.14.0-7.run deps=install rocm gfx=gfx1153 gpu-access=user
```

**i=runfile — ROCm meta components**

Meta components are similar to the meta packages used in the package manager
installation method. They group related components and dependencies
together, allowing you to install only what is necessary for your use case.
The following table describes available ROCm meta components:

| Meta component name   | Use case                  | Description                                                       | Contents                                                                                       |
|-----------------------|---------------------------|-------------------------------------------------------------------|------------------------------------------------------------------------------------------------|
| `core`                | ROCm Base                 | Core runtime environment. Install this to run ROCm applications.  | Runtimes, libraries, system control and monitoring tools, and other essential components.      |
| `core-dev`            | ROCm Developer Essentials | Development environment. Install this to build ROCm applications. | `core` plus compilers, CMake configurations, static library files, and headers.                |
| `dev-tools`           | ROCm Profiler             | Install this to profile and optimize ROCm applications.           | Profilers and related tools.                                                                   |
| `opencl`              | ROCm OpenCL               | Install this to run OpenCL applications on ROCm.                  | Components needed to run OpenCL.                                                               |
| `core-sdk`            | ROCm Full Suite           | Install this if you need everything.                              | The complete ROCm Core SDK including runtimes, compilers, development tools, and dependencies. |

The default installation uses the core meta component. To select other
components, add the `compo=` argument. For example, to install both `core` and
`core-dev`:

**gfx=gfx950**

```bash
bash rocm-installer-7.14.0-7.run deps=install rocm gfx=gfx950 compo=core,core-dev gpu-access=user
```

**gfx=gfx942**

```bash
bash rocm-installer-7.14.0-7.run deps=install rocm gfx=gfx942 compo=core,core-dev gpu-access=user
```

**gfx=gfx90a**

```bash
bash rocm-installer-7.14.0-7.run deps=install rocm gfx=gfx90a compo=core,core-dev gpu-access=user
```

**gfx=gfx908**

```bash
bash rocm-installer-7.14.0-7.run deps=install rocm gfx=gfx908 compo=core,core-dev gpu-access=user
```

**gfx=gfx1201**

```bash
bash rocm-installer-7.14.0-7.run deps=install rocm gfx=gfx1201 compo=core,core-dev gpu-access=user
```

**gfx=gfx1200**

```bash
bash rocm-installer-7.14.0-7.run deps=install rocm gfx=gfx1200 compo=core,core-dev gpu-access=user
```

**gfx=gfx1100**

```bash
bash rocm-installer-7.14.0-7.run deps=install rocm gfx=gfx1100 compo=core,core-dev gpu-access=user
```

**gfx=gfx1101**

```bash
bash rocm-installer-7.14.0-7.run deps=install rocm gfx=gfx1101 compo=core,core-dev gpu-access=user
```

**gfx=gfx1102**

```bash
bash rocm-installer-7.14.0-7.run deps=install rocm gfx=gfx1102 compo=core,core-dev gpu-access=user
```

**gfx=gfx1103**

```bash
bash rocm-installer-7.14.0-7.run deps=install rocm gfx=gfx1103 compo=core,core-dev gpu-access=user
```

**gfx=gfx1030**

```bash
bash rocm-installer-7.14.0-7.run deps=install rocm gfx=gfx1030 compo=core,core-dev gpu-access=user
```

**gfx=gfx1151**

```bash
bash rocm-installer-7.14.0-7.run deps=install rocm gfx=gfx1151 compo=core,core-dev gpu-access=user
```

**gfx=gfx1150**

```bash
bash rocm-installer-7.14.0-7.run deps=install rocm gfx=gfx1150 compo=core,core-dev gpu-access=user
```

**gfx=gfx1152**

```bash
bash rocm-installer-7.14.0-7.run deps=install rocm gfx=gfx1152 compo=core,core-dev gpu-access=user
```

**gfx=gfx1153**

```bash
bash rocm-installer-7.14.0-7.run deps=install rocm gfx=gfx1153 compo=core,core-dev gpu-access=user
```

**os=wsl — Install ROCDXG and AMD SMI for WSL**

**Package manager**

1. Download and install ROCDXG Library.
   ```bash
   wget https://github.com/ROCm/librocdxg/releases/download/v1.2.1/rocdxg-roct_1.2.1_amd64.deb
   sudo apt install ./rocdxg-roct_1.2.1_amd64.deb
   ```
2. Download and install AMD SMI Library for WSL.
   ```bash
   sudo apt install python3-pip python3-wheel python3-argcomplete
   wget https://github.com/ROCm/librocdxg/releases/download/v1.2.1/rocdxg-amd-smi-lib_1.2.1_amd64.deb
   sudo apt install ./rocdxg-amd-smi-lib_1.2.1_amd64.deb
   source /etc/profile.d/rocdxg-amd-smi-lib.sh
   ```

**Build from source**

1. In your host Windows environment, download and install the [Windows SDK](https://learn.microsoft.com/en-us/windows/apps/windows-sdk/) for
   Windows 11. Make sure you have the necessary permissions to access the
   Windows SDK installation files from your WSL2 environment.
2. In your WSL2 environment, clone the ROCDXG library.
   ```bash
   git clone https://github.com/ROCm/librocdxg.git
   ```
3. Set `WIN_SDK_PATH` to the Windows SDK include directory for your
   installed version.
   ```bash
   # Set the Windows SDK path (adjust version number if different)
   export WIN_SDK_PATH="/mnt/c/Program Files (x86)/Windows Kits/10/Include/10.0.28000.0"
   ```
4. Build the ROCDXG Library using CMake.
   ```bash
   cd librocdxg

mkdir -p build
   cd build
   cmake .. -DWIN_SDK="${WIN_SDK_PATH}/shared"
   make
   sudo make install
   ```
5. Build the AMD SMI Library for WSL using CMake.
   ```bash
   # Before proceeding, cd /path/to/librocdxg/
   cd amdsmi
   cmake -B build -DWIN_SDK="${WIN_SDK_PATH}/shared" .
   cmake --build build
   sudo cmake --install build
   source /etc/profile.d/rocdxg-amd-smi-lib.sh
   ```

See [librocdxg](https://github.com/ROCm/librocdxg) and
[amdsmi](https://github.com/ROCm/librocdxg/tree/develop/amdsmi) for
more information.

---

## Post-installation

After installing ROCm 7.14.0, complete these post-installation steps to
complete your system configuration and validate the installation.

**w=compute**

**os=ubuntu os=debian os=rhel os=oracle-linux os=rocky-linux os=sles**

**i=tar — Configure your environment**

Configure environment variables so that ROCm libraries and tools are
available either to all users on the system or only to your user account.

**System-wide setup**

Create a profile script so that all users inherit the ROCm
environment variables when they start a shell session. Make sure
you’re in the `therock-tarball` directory before proceeding.

```bash
# Configure ROCm PATH. Make sure you're in the therock-tarball directory before proceeding.
ROCM_INSTALL_PATH=$(pwd)/install
sudo tee /etc/profile.d/set-rocm-env.sh << EOF
export ROCM_PATH=$ROCM_INSTALL_PATH
export PATH=\$PATH:\$ROCM_PATH/bin
export LD_LIBRARY_PATH=\$ROCM_PATH/lib
EOF
sudo chmod +x /etc/profile.d/set-rocm-env.sh
source /etc/profile.d/set-rocm-env.sh
```

**User setup**

Configure the ROCm environment for your user by updating your shell
startup configuration file.

Use the following commands to update your shell configuration file
(`~/.bashrc` or `~/.profile`) and add ROCm to your PATH. Before proceeding, make sure you’re in the
`therock-tarball` directory so the install path resolves correctly.

**.bashrc**

```bash
# Configure ROCm PATH. Make sure you're in the therock-tarball directory before proceeding.
ROCM_INSTALL_PATH=$(pwd)/install
tee --append ~/.bashrc << EOF

# BEGIN ROCm environment configuration
export ROCM_PATH=$ROCM_INSTALL_PATH
export PATH=\$PATH:\$ROCM_PATH/bin
export LD_LIBRARY_PATH=\$ROCM_PATH/lib
# END ROCm environment configuration
EOF
source ~/.bashrc
```

**.profile**

```bash
# Configure ROCm PATH. Make sure you're in the therock-tarball directory before proceeding.
ROCM_INSTALL_PATH=$(pwd)/install
tee --append ~/.profile << EOF

**os=windows**

**i=tar — Configure your environment**

Configure environment variables so that ROCm libraries and tools are
available on your Windows system.

1. **Run command prompt as an administrator** and set the following environment variables.
   ```bat
   setx HIP_DEVICE_LIB_PATH "C:\TheRock\build\lib\llvm\amdgcn\bitcode" /M
   setx HIP_PATH "C:\TheRock\build" /M
   setx HIP_PLATFORM "amd" /M
   setx LLVM_PATH "C:\TheRock\build\lib\llvm" /M
   ```
2. Add the following paths into the PATH environment variable.
   ```bat
   setx PATH "%PATH%;C:\TheRock\build\bin;C:\TheRock\build\lib\llvm\bin" /M
   ```
3. Open a new command prompt window for the environment variables to take effect. Run `set`
   to see the list of active variables.
   ```bat
   set
   ```

**os=ubuntu os=debian os=rhel os=oracle-linux os=rocky-linux os=sles — Verify your installation**

Use the following ROCm tools to verify that ROCm is correctly
installed and that your AMD devices are visible to the system.

1. Use `rocminfo` to list detected AMD GPUs and confirm that the ROCm
   runtimes and drivers are correctly installed and loaded.
   ```bash
   rocminfo
   ```
2. Use the AMD SMI CLI `amd-smi` to validate system information.
   ```bash
   amd-smi version
   ```

**i=pip**

1. Inspect your installation in your Python environment and confirm that
   ROCm packages, including the `rocm-sdk` CLI, are available.
   ```bash
   pip freeze | grep rocm
   which rocm-sdk
   ls .venv/bin
   ```

**os=windows — Verify your installation**

Use the following ROCm tools to verify that ROCm is correctly
installed and that your AMD devices are visible to the system.

**i=pip**

1. Use `hipinfo` to list detected AMD GPUs and confirm that the ROCm
   runtimes and drivers are correctly installed and loaded.
   ```bash
   hipinfo
   ```
2. Inspect your installation in your Python environment and confirm that
   ROCm packages, including the `rocm-sdk` CLI, are available.
   ```bash
   pip freeze
   where rocm-sdk
   dir .venv\Scripts
   ```

**i=tar**

Use `hipinfo` to list detected AMD GPUs and confirm that the ROCm
runtimes and drivers are correctly installed and loaded.

```bash
hipinfo
```

**os=wsl — Verify your installation**

Use `rocminfo` to verify that ROCm is correctly
installed and that your AMD devices are visible to the system.

```bash
rocminfo
```

**os=ubuntu os=debian os=rhel os=oracle-linux os=rocky-linux os=sles**

**i=pip — Configure your environment**

#### NOTE
Follow this step only if you installed the `devel` package.

Initialize the ROCm SDK.

```bash
rocm-sdk init
```

**i=pip — Test your installation**

Run the following commands from your Python virtual environment to confirm
that the ROCm SDK is correctly configured and that basic checks complete
successfully.

```bash
rocm-sdk targets
rocm-sdk test
```

To learn more about the `rocm-sdk` tool and to see example
outputs, see [Using ROCm Python packages (TheRock)](https://github.com/ROCm/TheRock/blob/main/RELEASES.md#using-rocm-python-packages).

**os=windows**

**i=pip — Test your installation**

Run the following commands from your Python virtual environment to confirm
that the ROCm SDK is correctly configured and that basic checks complete
successfully.

```bash
rocm-sdk test
```

To learn more about the `rocm-sdk` tool and to see example
outputs, see [Using ROCm Python packages (TheRock)](https://github.com/ROCm/TheRock/blob/main/RELEASES.md#using-rocm-python-packages).

**i=pip**

#### NOTE
If you need to deactivate your Python virtual environment when finished, run:

```bash
deactivate
```

#### NOTE
**fam=all fam=instinct**

To install deep learning frameworks, including [PyTorch](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/frameworks/pytorch/install.html)
and [JAX](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/frameworks/jax/install.html),
and get started with AI training and inference, see the [AI Ecosystem](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/)
documentation portal.

**fam=radeon fam=ryzen**

To install deep learning frameworks, including [PyTorch](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/frameworks/pytorch/install.html),
and get started with AI training and inference, see the [AI Ecosystem](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/)
documentation portal.

**fam=all fam=instinct**

To learn about HPC libraries and applications, see
[ROCm HPC SDK](https://rocm.docs.amd.com/en/latest/components/hpc-sdk/index.html).

**fam=all fam=instinct fam=radeon**

To learn about ROCm Extras, which include supplementary tools for
benchmarking and validating, see [ROCm Extras](https://rocm.docs.amd.com/en/latest/components/extras.html).

---

## Uninstalling

**i=pkgman**

1. Use your package manager to remove the [installed packages](#rocm-install-rocm).

**os=ubuntu os=debian os=wsl**

**fam=all**
   ```bash
   sudo apt autoremove amdrocm7.14
   ```

**gfx=gfx950**
   ```bash
   sudo apt autoremove amdrocm7.14-gfx950
   ```

**gfx=gfx942**
   ```bash
   sudo apt autoremove amdrocm7.14-gfx942
   ```

**gfx=gfx90a**
   ```bash
   sudo apt autoremove amdrocm7.14-gfx90a
   ```

**gfx=gfx908**
   ```bash
   sudo apt autoremove amdrocm7.14-gfx908
   ```

**gfx=gfx1201**
   ```bash
   sudo apt autoremove amdrocm7.14-gfx1201
   ```

**gfx=gfx1200**
   ```bash
   sudo apt autoremove amdrocm7.14-gfx1200
   ```

**gfx=gfx1100**
   ```bash
   sudo apt autoremove amdrocm7.14-gfx1100
   ```

**gfx=gfx1101**
   ```bash
   sudo apt autoremove amdrocm7.14-gfx1101
   ```

**gfx=gfx1102**
   ```bash
   sudo apt autoremove amdrocm7.14-gfx1102
   ```

**gfx=gfx1103**
   ```bash
   sudo apt autoremove amdrocm7.14-gfx1103
   ```

**gfx=gfx1030**
   ```bash
   sudo apt autoremove amdrocm7.14-gfx1030
   ```

**gfx=gfx1151**
   ```bash
   sudo apt autoremove amdrocm7.14-gfx1151
   ```

**gfx=gfx1150**
   ```bash
   sudo apt autoremove amdrocm7.14-gfx1150
   ```

**gfx=gfx1152**
   ```bash
   sudo apt autoremove amdrocm7.14-gfx1152
   ```

**gfx=gfx1153**
   ```bash
   sudo apt autoremove amdrocm7.14-gfx1153
   ```

**os=rhel os=oracle-linux os=rocky-linux**

**fam=all**
   ```bash
   sudo dnf remove amdrocm7.14
   ```

**gfx=gfx950**
   ```bash
   sudo dnf remove amdrocm7.14-gfx950
   ```

**gfx=gfx942**
   ```bash
   sudo dnf remove amdrocm7.14-gfx94x
   ```

**gfx=gfx90a**
   ```bash
   sudo dnf remove amdrocm7.14-gfx90a
   ```

**gfx=gfx908**
   ```bash
   sudo dnf remove amdrocm7.14-gfx908
   ```

**gfx=gfx1201 gfx=gfx1200**
   ```bash
   sudo dnf remove amdrocm7.14-gfx120x
   ```

**gfx=gfx1100 gfx=gfx1101 gfx=gfx1102 gfx=gfx1103**
   ```bash
   sudo dnf remove amdrocm7.14-gfx110x
   ```

**gfx=gfx1030**
   ```bash
   sudo dnf remove amdrocm7.14-gfx103x
   ```

**gfx=gfx1151**
   ```bash
   sudo dnf remove amdrocm7.14-gfx1151
   ```

**gfx=gfx1150**
   ```bash
   sudo dnf remove amdrocm7.14-gfx1150
   ```

**gfx=gfx1152**
   ```bash
   sudo dnf remove amdrocm7.14-gfx1152
   ```

**gfx=gfx1153**
   ```bash
   sudo dnf remove amdrocm7.14-gfx1153
   ```

**os=sles**
   ```bash
   sudo zypper remove amdrocm*
   ```
2. Remove ROCm repositories.

**os=ubuntu os=debian os=wsl**
   ```bash
   # Remove ROCm repositories
   sudo rm /etc/apt/sources.list.d/rocm.list

# Clear the cache and clean the system
   sudo rm -rf /var/cache/apt/*
   sudo apt clean all
   sudo apt update
   ```

**os=rhel os=oracle-linux os=rocky-linux**
   ```bash
   # Remove ROCm repositories
   sudo rm /etc/yum.repos.d/rocm.repo*

# Clear the cache and clean the system
   sudo rm -rf /var/cache/dnf
   sudo dnf clean all
   ```

**os=sles**
   ```bash
   # Remove ROCm repositories
   sudo zypper removerepo "rocm"

# Clear the cache and clean the system
   sudo zypper clean --all
   sudo zypper refresh
   ```

**i=pip**

1. Clear the pip cache.

**os=ubuntu os=debian os=rhel os=oracle-linux os=rocky-linux os=sles**
   ```bash
   rm -rf ~/.cache/pip
   ```

**os=windows**
   ```bat
   pip cache purge
   ```
2. Remove your local Python virtual environment.

**os=ubuntu os=debian os=rhel os=oracle-linux os=rocky-linux os=sles**
   ```bash
   rm -rf .venv
   ```

**os=windows**
   ```bat
   rmdir /s /q .venv
   ```

**i=tar**

**os=ubuntu os=debian os=rhel os=oracle-linux os=rocky-linux os=sles**

1. To uninstall ROCm, remove your installation directory.

#### IMPORTANT
   The following command assumes you’re working with the
   `therock-tarball` directory. If you chose a different directory
   name when [installing ROCm](#rocm-install), adjust the command
   accordingly.

```bash
   sudo rm -rf therock-tarball
   ```
2. Remove your ROCm environment configuration from your system.

**System-wide**

If you opted for a [system-wide setup](#rocm-post-install-env) during the installation
   process, remove the ROCm environment variables.
   ```bash
   sudo rm -f /etc/profile.d/set-rocm-env.sh
   ```

**User**

If you opted for a [user-specific setup](#rocm-post-install-env) during the installation
   process, remove the ROCm environment configuration block from
   your shell configuration file (`~/.bashrc` or `~/.profile`).

**os=windows**

1. To uninstall ROCm, remove your installation directory.
   ```bat
   rmdir /s /q C:\TheRock
   ```

#### IMPORTANT
   This step assumes you’re working with the `C:\TheRock\build`
   directory. If you chose a different directory name when
   [installing ROCm](#rocm-install), adjust this step
   accordingly.
2. **Run command prompt as an administrator** and delete the following environment variables.
   ```bat
   setx HIP_DEVICE_LIB_PATH "" /M
   setx HIP_PATH "" /M
   setx HIP_PLATFORM "" /M
   setx LLVM_PATH "" /M
   ```

Remove the following paths from your PATH environment variable using your system settings GUI.
   Navigate to the following screen:
   * Control Panel > System and Security > Edit environment variables

Edit the PATH variable and delete the following paths:
   * `C:\TheRock\build\bin`
   * `C:\TheRock\build\lib\llvm\bin`
3. To uninstall the Adrenalin Driver, see [Uninstall AMD Software](https://www.amd.com/en/resources/support-articles/faqs/RSX2-UNINSTALL.html).

**i=runfile**

1. Use the following command to uninstall ROCm.

**fam=all**
   ```bash
   bash rocm-installer-7.14.0-7.run uninstall-rocm gfx=all
   ```

**gfx=gfx950**
   ```bash
   bash rocm-installer-7.14.0-7.run uninstall-rocm gfx=gfx950
   ```

**gfx=gfx942**
   ```bash
   bash rocm-installer-7.14.0-7.run uninstall-rocm gfx=gfx942
   ```

**gfx=gfx90a**
   ```bash
   bash rocm-installer-7.14.0-7.run uninstall-rocm gfx=gfx90a
   ```

**gfx=gfx908**
   ```bash
   bash rocm-installer-7.14.0-7.run uninstall-rocm gfx=gfx908
   ```

**gfx=gfx1201**
   ```bash
   bash rocm-installer-7.14.0-7.run uninstall-rocm gfx=gfx1201
   ```

**gfx=gfx1200**
   ```bash
   bash rocm-installer-7.14.0-7.run uninstall-rocm gfx=gfx1200
   ```

**gfx=gfx1100**
   ```bash
   bash rocm-installer-7.14.0-7.run uninstall-rocm gfx=gfx1100
   ```

**gfx=gfx1101**
   ```bash
   bash rocm-installer-7.14.0-7.run uninstall-rocm gfx=gfx1101
   ```

**gfx=gfx1102**
   ```bash
   bash rocm-installer-7.14.0-7.run uninstall-rocm gfx=gfx1102
   ```

**gfx=gfx1103**
   ```bash
   bash rocm-installer-7.14.0-7.run uninstall-rocm gfx=gfx1103
   ```

**gfx=gfx1030**
   ```bash
   bash rocm-installer-7.14.0-7.run uninstall-rocm gfx=gfx1030
   ```

**gfx=gfx1151**
   ```bash
   bash rocm-installer-7.14.0-7.run uninstall-rocm gfx=gfx1151
   ```

**gfx=gfx1150**
   ```bash
   bash rocm-installer-7.14.0-7.run uninstall-rocm gfx=gfx1150
   ```

**gfx=gfx1152**
   ```bash
   bash rocm-installer-7.14.0-7.run uninstall-rocm gfx=gfx1152
   ```

**gfx=gfx1153**
   ```bash
   bash rocm-installer-7.14.0-7.run uninstall-rocm gfx=gfx1153
   ```
2. Use the following command to uninstall the AMD GPU Driver (amdgpu).
   ```bash
   bash rocm-installer-7.14.0-7.run uninstall-amdgpu
   ```

**w=graphics**

**os=ubuntu os=rhel**

**fam=radeon**

1. Use `amdgpu-uninstall` to remove the [installed packages](#rocm-install-rocm).
   ```bash
   sudo amdgpu-uninstall
   ```

**fam=ryzen**

1. Use `amdgpu-uninstall` to remove the [installed packages](#rocm-install-rocm).
   ```bash
   sudo amdgpu-uninstall
   ```

**fam=all**

1. Use `amdgpu-uninstall` to remove the [installed packages](#rocm-install-rocm).
   ```bash
   sudo amdgpu-uninstall
   ```

1. Remove ROCm repositories.

**os=ubuntu**
   ```bash
   sudo apt purge amdgpu-install
   sudo apt autoremove

# Clear the cache and clean the system
   sudo rm -rf /var/cache/apt/*
   sudo apt clean all
   sudo apt update
   ```

**os=rhel**
   ```bash
   sudo dnf remove amdgpu-install

# Clear the cache and clean the system
   sudo rm -rf /var/cache/dnf
   sudo dnf clean all
   ```

---

Source: https://rocm.docs.amd.com/en/latest/install/redirect/_prerequisites.html

# Install AMD ROCm 7.14.0

## Redirecting…

#### IMPORTANT
If you’re not redirected automatically, go to [Prerequisites](https://rocm.docs.amd.com/en/latest/install/rocm.html#rocm-prerequisites).

---

Source: https://rocm.docs.amd.com/en/latest/install/redirect/_install.html

# Install AMD ROCm 7.14.0

## Redirecting…

#### IMPORTANT
If you’re not redirected automatically, go to [Installation](https://rocm.docs.amd.com/en/latest/install/rocm.html#rocm-install).

---

Source: https://rocm.docs.amd.com/en/latest/install/redirect/_post-install.html

# Install AMD ROCm 7.14.0

## Redirecting…

#### IMPORTANT
If you’re not redirected automatically, go to [Post-installation](https://rocm.docs.amd.com/en/latest/install/rocm.html#rocm-post-install).

---

Source: https://rocm.docs.amd.com/en/latest/install/redirect/_uninstall.html

# Install AMD ROCm 7.14.0

## Redirecting…

#### IMPORTANT
If you’re not redirected automatically, go to [Uninstalling](https://rocm.docs.amd.com/en/latest/install/rocm.html#rocm-uninstall).

---

Source: https://rocm.docs.amd.com/en/latest/install/rocm-runfile-installer.html

# ROCm Runfile Installer

The ROCm Runfile Installer installs ROCm, the AMD GPU Driver,
or a combination of the two on a system with or without
network or internet access. Unlike all other installation methods, the ROCm Runfile Installer
can install ROCm and the AMD GPU Driver without using a native Linux package management system.

The key advantage of using the ROCm Runfile Installer is its offline installation support.
Many system environments have network or internet access restrictions, making installation via
normal package management difficult. Furthermore, some installation environments might also have general
restrictions on package management usage. Therefore, the ROCm Runfile Installer lets you perform
a completely self-contained ROCm software installation.

The ROCm Runfile Installer includes these features:

* An optional easy-to-use user interface for configuring the installation
* An optional command line interface for the installation
* Offline ROCm and AMD GPU Driver installation (requires the prior installation of dependencies)
* Packageless ROCm and AMD GPU Driver install without native package management
* A single self-contained installer for all supported Linux distributions
* Multi-architecture GPU support with selective installation and uninstall
* Flexible component selection (core, development tools, SDK, OpenCL)
* Graphics support option for Mesa/OpenGL workloads
* Configurable installation location for the ROCm install

## Prerequisites

The ROCm Runfile Installer requires the following configuration:

* Installation of dependency requirements for the ROCm runtime
* Installation of dependency requirements for the AMD GPU Driver (optional)
* Sufficient storage space for the installation (100 GB of free space recommended)
* A supported Linux distribution

## Dependency requirements

The ROCm components contained within the ROCm Runfile installer have a specific set of libraries, frameworks, and other elements
that must be pre-installed on the system before you can use ROCm after the installation. Similarly, the
inclusion of the AMD GPU Driver as part of the installation also requires specific libraries that must be pre-installed. ROCm Runfile installer users
must pre-install the list of required first-level dependencies.

To install the pre-install dependencies, use one of two methods:

* Manual installation
* The ROCm Runfile Installer

### Manual installation

You can determine the dependent packages from the ROCm
Runfile Installer **Pre-Install Configuration Settings** menu
in the GUI or from the command line by using the `deps=list rocm` argument
for ROCm or the `deps=list amdgpu` argument for the AMD GPU Driver. This list, which indicates all packages
required for the ROCm runtime or the AMD GPU Driver, is displayed and saved to a
file named `deps_list.txt` in the root `rocm-installer` directory. The required libraries, frameworks, and other components
within the required packages must be on the system when running ROCm or installing the AMD GPU Driver.
Users can manually install the required packages in the list using any method.

The list of packages that is displayed and saved is not directly usable by the package manager.
Each line indicates the name of the package or versioning required to meet the dependency requirement.
Some dependencies might show multiple packages separated by the `|` symbol, meaning that any package from the list satisfies
the package dependency. For example, the `libgcc-dev` dependency might be `libgcc-5-dev | libgcc-7-dev`, meaning
that either `libgcc-5-dev` or `libgcc-7-dev` can be installed to satisfy the corresponding dependency.

System administrators might prefer a manual installation process when deploying ROCm across a multi-node
cluster environment, where a base operating system image
is prepared and applied. This base OS image might have the dependency requirements pre-installed.

### ROCm Runfile Installer

For single-system environments, users can choose to have the ROCm Runfile Installer automatically install the
dependency requirements as part of the pre-installation stage for ROCm or the AMD GPU Driver. Any missing dependency requirements can
be installed using the **Install Dependencies** option of the **Pre-Install Configuration Settings** menu
in the GUI or from the command line using the `deps=install rocm` or `deps=install amdgpu` argument.

#### NOTE
The ROCm Runfile Installer requires a network or internet connection to install the dependency requirements.

## Supported Linux distributions

The ROCm Runfile Installer supports the following Linux distributions and versions:

* Ubuntu: 22.04, 24.04, 26.04
* RHEL: 8.10, 9.4, 9.6, 9.8, 10.0, 10.2
* SLES: 15.7, 16.0
* Debian: 12, 13
* Oracle Linux: 8.10, 9.7, 10.1
* Rocky Linux: 9.7

#### IMPORTANT
The new ROCm runfile installer is a universal installer that supports all listed distributions from a single
`.run` file. There is no need to download distribution-specific builds as was required in previous versions.

## Getting started

The ROCm Runfile Installer is distributed as a self-extracting `.run` file.
To install ROCm, launch the installer from any directory on the system.

### Downloading the ROCm Runfile Installer

Download the ROCm Runfile Installer from [repo.radeon.com](https://repo.radeon.com/) using the following command:

```shell
wget https://repo.radeon.com/rocm/installer/rocm-runfile-installer/rocm-rel-<rocm-version>/<installer-file>
```

Substitute values specific to your installation for the following placeholders:

```shell
<rocm-version> = ROCm version number for the installer (for example, 7.14)
<installer-file> = The installer .run file
```

For example, to download ROCm 7.14 of the ROCm Runfile Installer:

```shell
wget https://repo.radeon.com/rocm/installer/rocm-runfile-installer/rocm-rel-7.14/rocm-installer-7.14.0-7.run
```

### Running the ROCm Runfile Installer

After downloading the ROCm Runfile Installer, run it from a terminal using
[GUI install](#gui-install) or [Command line install](#command-line-install). See the sections below for more details.

You can obtain help, version, architecture, or component information using the following installer `.run` file argument options:

```shell
bash rocm-installer-7.14.0-7.run help
bash rocm-installer-7.14.0-7.run version
bash rocm-installer-7.14.0-7.run buildinfo
bash rocm-installer-7.14.0-7.run gfx=list
bash rocm-installer-7.14.0-7.run compo=list
```

The `help`, `version`, `buildinfo`, `gfx=list`, and `compo=list` commands run without extracting the installer contents.
For all other argument options, or if no arguments are specified, the installer `.run` file self-extracts
to the current working directory where the `.run` file is executing.
The self-extraction process creates a new directory named `rocm-installer` containing the content
and tools required for the installation. The `rocm-installer` directory also includes a `logs`
directory for recording the installation process. For more information, see [Log files](#log-files) below.

#### NOTE
The installer self-extraction process might take a significant amount of time due to the size of the installer content and the decompression process.

## Install methods

The ROCm Runfile Installer provides two methods for running the ROCm installation:

* [GUI install](#gui-install): The GUI installation includes a visual interface for configuring the installation,
  letting you specify the pre- and post-installation requirements. In addition, the GUI provides
  feedback and guidance for setting up the installation. This method is recommended for new
  and intermediate installer users.
* [Command line install](#command-line-install): The command line interface installation method provides a direct terminal-based approach for
  configuring and running the installation. This method is recommended for more advanced installer users and automation scenarios.

## GUI install

Launch the GUI-based installation of the ROCm Runfile Installer from the terminal command line without arguments as follows:

```shell
bash rocm-installer-7.14.0-7.run
```

### GUI

Use the Runfile Installer GUI to configure the installation, from the pre- to post-install options.

Starting from the **Main** menu, the user interface contains multiple menus and sub-menus for each stage
of the installation process.

#### Main menu

The **Main** menu is the installation starting point.

![The Main user interface menu for the ROCm Runfile Installer](data/how-to/rocm-runfile-main-menu-1.png)

#### Pre-Install Configuration Settings menu

The **Pre-Install Configuration Settings** menu is an optional menu used to configure pre-installation
requirements before installation. The pre-installation settings relate to the dependent libraries
and packages required by the ROCm runtime or the AMD GPU Driver.

![The Pre-Install Configuration user interface menu for the ROCm Runfile Installer](data/how-to/rocm-runfile-preinstall-menu-2.png)
* **File**

**File** displays the location of the `deps_list.txt` file. This file is based
  on the combination of selected dependencies using the **ROCm**, **Driver**, and **Graphics** checkboxes and either
  **Display Dependencies** or **Validate Dependencies**. If **Display Dependencies** is selected, `deps_list.txt`
  includes a list of all required dependencies. If **Validate Dependencies** is selected,
  `deps_list.txt` contains only the missing dependencies on the system that still require installation.
  The **File** field is initially blank until **Display Dependencies** or **Validate Dependencies** is selected.
  Use the **Install Dependencies** to install the required dependencies on the system.
* **ROCm** / **Driver**

The **ROCm** and **Driver** checkboxes are used to select which dependencies you want to display, validate, or install.
* **Graphics**

The **Graphics** checkbox enables mixed graphics and compute support for Mesa/OpenGL workloads.
  This is an option for non-Instinct GPUs such as AMD Radeon or Ryzen GPUs/APUs. When checked, the installer includes the
  `amdgpu-lib` package.
* **Display Dependencies**

**Display Dependencies** lists all required (Debian or RPM) packages that must be pre-installed on the system.
  These packages are required by ROCm or the AMD GPU Driver being installed by a particular ROCm Runfile Installer version.

#### NOTE
  The required packages are listed in the `deps_list.txt` file and can be installed separately from the ROCm Runfile Installer.
* **Validate Dependencies**

**Validate Dependencies** verifies which required packages are currently installed on the system where
  ROCm or the AMD GPU Driver is being installed. It displays which packages from the required packages list are missing.

#### NOTE
  The missing packages are listed in the `deps_list.txt` file.
* **Install Dependencies**

If the installer is running on the system where ROCm or the AMD GPU Driver will be installed, you can choose to install
  any missing dependencies using the **Install Dependencies** option.

#### NOTE
  **Install Dependencies** is only intended for a system with network or internet access. For offline installation
  using the ROCm Runfile Installer, you must manually install the required packages.
  For more details, see the [Dependency requirements](#dependency-requirements) section.

#### ROCm Options menu

The **ROCm Options** menu can include or exclude ROCm from the installation and provides access to device/architecture and component selection.

![The ROCm Options user interface menu for the ROCm Runfile Installer](data/how-to/rocm-runfile-rocm-menu-3.png)
* **Install ROCm**

This field indicates whether to include ROCm components in the installation. If this field is set to `yes`,
  ROCm installation is enabled and the Runfile Installer searches for any existing ROCm installations at the currently set
  **ROCm Install Path**. If a previous ROCm installation is detected, the **Uninstall ROCm** field becomes selectable.
* **ROCm Device**

This submenu allows you to select which GPU architectures to install. You can choose specific architectures
  (gfx942, gfx1100, gfx950, and so on), multiple architectures, or all available architectures. The installer will list the
  currently auto-detected GPU and associated architecture by default.
  ![The ROCm Device user interface menu for the ROCm Runfile Installer](data/how-to/rocm-runfile-rocm-device-3a.png)
* **ROCm Components**

This submenu allows you to select which ROCm component categories to install:
  * **core**: Core ROCm components
  * **core-dev**: Core development components
  * **dev-tools**: Developer tools
  * **core-sdk**: Core SDK components
  * **opencl**: OpenCL runtime

![The ROCm Components user interface menu for the ROCm Runfile Installer](data/how-to/rocm-runfile-rocm-components-3b.png)
* **ROCm Install Path**

If **Install ROCm** is set to `yes`, the **ROCm Install Path** field can set the full path to
  the directory where ROCm will be installed. The default location is `/opt`, which results in a `/opt/rocm/core-x.y.z` installation.

When **ROCm Install Path** is set to a new path, the installer validates the new directory location.
  If the directory exists, the ROCm installation can proceed. If an invalid location is specified,
  the ROCm installation will not be allowed.
* **Uninstall ROCm**

This field is only available if previous Runfile ROCm installations are discovered on the system
  where ROCm is being installed, based on the currently selected **ROCm Install Path** location.
  The installer only lists previous installation locations that are on the current install path.
  If any installation locations are present on this path, they can be selected for uninstall.

The installer indicates the type of installation as follows:
  * **P** (Package manager): Package manager installation that matches the ROCm version for the Runfile Installer. In this case, uninstall is not allowed.
  * **C** (Runfile conflict): Runfile installation that matches the ROCm version for the Runfile Installer. The conflicting installation can be uninstalled.
  * **R** (Runfile): Runfile installation that differs from the ROCm version for the Runfile Installer. The installation can be uninstalled.

**Uninstall ROCm** is for a single ROCm instance at a time and is only available for Runfile installs.
  Package manager installs of ROCm cannot be uninstalled by the Runfile installer and must be uninstalled manually
  using the package management application.

#### NOTE
  For multi-architecture installs from the current **ROCm Install Path**, the GUI interface will uninstall all architectures.
  To uninstall specific GPU architectures, use the command line interface for uninstall.

#### Driver Options menu

The **Driver Options** menu can include or exclude the AMD GPU Driver from the installation.

![The Driver Options user interface menu for the ROCm Runfile Installer](data/how-to/rocm-runfile-driver-menu-4.png)
* **Install Driver**

This field indicates whether to include the AMD GPU Driver in the installation. If this field is set to `yes`,
  AMD GPU Driver installation is enabled. When this field is enabled, the system is searched for any existing
  AMD GPU Driver installations. If a previous AMD GPU Driver installation is detected, the **Uninstall Driver** field becomes selectable.

#### NOTE
  The AMD GPU Driver is installed for the currently running Linux kernel version. To install the AMD GPU Driver for a different kernel,
  reboot to the specific Linux kernel and reinstall the AMD GPU Driver using the runfile.
* **Start on install**

If the **Start on install** option is enabled, the Runfile installer uses `modprobe` to automatically launch
  the AMD GPU Driver after installation. If a pre-existing AMD GPU Driver is already loaded on the system, the new driver will not start.
  This option can be useful for installing the driver on a system where the GPU device is newer and not yet natively
  supported as part of the upstream GPU driver for the Linux distribution.
* **Uninstall Driver**

This field is only available if a previous Runfile install of the AMD GPU Driver is discovered on the install system.
  If a Runfile installation of the AMD GPU Driver is detected, select **Uninstall Driver** to remove it.
  Package manager installs of the AMD GPU Driver cannot be uninstalled by the Runfile installer and must be
  uninstalled manually using the package management application.

#### Post-Install Options menu

Use the **Post-Install Options** menu to optionally enable additional setup and configuration items
after the initial ROCm install.

![The Post-Install Options user interface menu for the ROCm Runfile Installer](data/how-to/rocm-runfile-postinstall-menu-5.png)
* **Set GPU access permissions**

This section sets the GPU access permissions after the ROCm installation.
  Typically, any ROCm component using the GPU and requiring access to GPU resources
  needs to set the access permission. For ROCm, GPU access is controlled by membership in
  the `video` and `render` groups. Membership and access to GPU resources can be set using one of two methods.
  * **Add video,render group**

If the **Add video,render group** field is enabled, the current `$USER` is added to the groups
    and will be granted GPU access.
  * **Add udev rule**

If the **Add udev rule** field is enabled, GPU access is granted to all users on the system.

#### NOTE
  It’s **recommended** that you enable one of the GPU access options for using ROCm.

Only one method for adding GPU access can be selected.
* **Post ROCm setup**

When this option is enabled, the ROCm post-install setup is performed. This includes configuring
  symbolic links and other system requirements for using ROCm and the ROCm runtime. This option is enabled by default for ROCm installations.

#### NOTE
  It’s recommended that you enable the **Post ROCm setup** to guarantee proper
  functioning of the ROCm components and applications. However, advanced users
  who understand the ROCm setup might want to disable this option so they can control the
  post-installation ROCm setup for their specific environment.

### Using the GUI

Start the ROCm Runfile Installer user interface from the terminal and launch the **Main** menu.
While navigating through the user interface menus, use the **Done** option to return to the previous menu.
Some menus have a **Help** option to display more information about the elements within the current menu.

Follow these steps to install ROCm and the AMD GPU Driver:

1. **(Optional)** Install dependencies:
   1. Enter the **Pre-Install Configuration** menu.
   2. Select the **ROCm** checkbox, **Driver** checkbox, or both to specify the dependency type.
   3. **(Optional)** Select the **Graphics** checkbox if you need graphics support for Mesa/OpenGL workloads.
   4. If using the installer to install missing required dependencies, select **Install Dependencies**.

To manually install the required dependencies, select **Display Dependencies** to list
      all the dependencies or **Validate Dependencies** to only list the missing dependencies on the
      current system. Quit the installer using **DONE->F1** and separately install the required dependencies,
      which will be listed in the `deps_list.txt` file at the **File** location.
      After completing this task, restart the ROCm Runfile Installer and proceed to step 2.
2. Set the ROCm options:
   1. Enter the **ROCm Options** menu.
   2. Set **Install ROCm** to **yes** to include ROCm components in the installation.
   3. **(Optional)** Enter the **ROCm Device** submenu to select GPU architectures. By default, the installer auto-detects your GPU.
   4. Enter the **ROCm Components** submenu to select component categories.
   5. **(Optional)** Select **Uninstall ROCm** to uninstall a previous Runfile installation or specific architectures.
   6. Leave the **ROCm Install Path** field set to the default location (`/opt`) to install ROCm to `/opt/rocm/core-x.y.z`
      or set the install location to a valid existing directory.
3. Set the AMD GPU Driver options:
   1. Enter the **Driver Options** menu.
   2. Set **Install Driver** to **yes** to include the AMD GPU Driver in the installation.
   3. **(Optional)** Select **Start on install** to load the driver after installation.
   4. **(Optional)** Select **Uninstall Driver** to uninstall a previous Runfile installation.
4. Set the post-install options:
   1. Set the method of enabling permission for GPU access to **Add video,render group** or **Add udev rule**.
   2. Set the **Post ROCm setup** to **yes** to include configuration of ROCm post-installation, which is recommended and enabled by default. Set to **no** to exclude the post-installation.

## Command line install

The command line install interface can be used as an alternative to the menu-based ROCm Runfile Installer GUI
to reduce user interaction during the installation and enable automation scenarios.

Run the ROCm Runfile Installer from the terminal command line as follows:

```shell
bash rocm-installer-7.14.0-7.run <options>
```

The `<options>` parameter can be set to these options:

* User help/information
 * `help`: Displays information on how to use the ROCm Runfile Installer.
 * `version`: Displays the current version of the ROCm Runfile Installer.
 * `buildinfo`: Display ROCm build information (theRock commit, GitHub run ID, and so on).
* Runfile options
 * `noexec`: Disable all installer execution. Extract the `.run` file content only.
 * `noexec-cleanup`: Disable cleanup after installer execution. Keep all `.run` extracted and runtime files.
* Dependencies
 * `deps=<arg> <compo>`:
 * `<arg>`:
 * `list <compo>`: Lists the required dependencies for the install `<compo>`.
 * `validate <compo>`: Validates which required dependencies are installed or not installed for `<compo>`.
 * `install-only <compo>`: Installs the required dependencies only for `<compo>`.
 * `install <compo>`: Installs with the required dependencies for `<compo>`.
 * `<compo>`: Install component (`rocm`/`amdgpu`/`rocm amdgpu`).
* Install
 * `rocm`: Enable ROCm components install.
 * `amdgpu`: Enable AMD GPU Driver install.
 * `force`: Force the ROCm and AMD GPU Driver install.
 * `graphics`: Enable graphics support by installing `amdgpu-lib` for Mesa/OpenGL.
 * `target=<directory>`: The target directory path for the ROCm components install (default: `/opt`).
 * `gfx=<arch>`: GPU architecture to install.
 * `<arch>`: Architecture specification (for example, `gfx942`, `gfx1100`, `gfx950`).

Special values:
 * `gfx=all`: Install all available architectures.
 * `gfx=list`: Show available architectures in the installer.
 * `gfx=list-installed`: Show currently installed architectures (checks `/opt` by default; use with `target=<path>` to check a different location).
 * `gfx=show`: Detect and display GPU information.
 * Multiple architectures: `gfx=gfx942,gfx950` (comma-separated).
 * If not specified, auto-detects GPU and installs matching architecture.
 * `compo=<component_list>`: Comma-separated list of ROCm components to install.
 * Available components: `core`, `core-dev`, `dev-tools`, `core-sdk`, `opencl`.
 * Default: `core` (if not specified).
 * `compo=list`: Show available component categories.
 * Examples: `compo=core`, `compo=core,dev-tools`, `compo=core-sdk`.
* Post-install
 * `postrocm`: Run the post-installation ROCm configuration. By default, post-install runs automatically with `rocm` installation.
 Use this to run post-install after installing with `nopostrocm`.
 * `nopostrocm`: Disable post ROCm installation configuration. Use this flag to skip post-install processing.
 * `amdgpu-start`: Start the AMD GPU Driver after the install.
 * `gpu-access=<access_type>`
 * `<access_type>`:
 * `user`: Adds the current user to the `render,video` group for GPU access.
 * `all`: Grants GPU access to all users on the system using `udev` rules.
* Uninstall
 * `uninstall-rocm (target=<directory>)`: Uninstall ROCm.
 * (`target=<directory>`): Optional target directory for the ROCm uninstall (default: `/opt/rocm/core-x.y.z`).
 * Multi-architecture selective uninstall:
 * `gfx=<arch>`: Uninstall specific architecture (for example, `gfx=gfx1100`).
 * `gfx=all`: Uninstall all architectures.
 * Base components remain until last architecture is removed.
 * `uninstall-amdgpu`: Uninstall the AMD GPU Driver.
* Information/Debug
 * `findrocm`: Search for a ROCm installation.
 * `manifest`: List the version of all ROCm components included in the installer.
 * `manifest=<gfx>`: List components for specific GFX architecture (for example, `manifest=gfx1100`, `manifest=base`).
 * `prompt`: Run the installer with user prompts.
 * `assumeyes`: Automatically answer yes to all prompts (useful for automation/scripting).
 * `verbose`: Run the installer with verbose logging.

#### NOTE
The installer can be used with multiple `<options>` combinations to enable specific stages of the
ROCm install process (pre-installation and post-installation). The exception is any option that uses the
keyword `-only` will apply that option only and no others. Some
informational options are also single-option commands.

### Command line interface

The command line interface for the ROCm Runfile Installer is based on the
`<options>` list provided to the installer `.run` file.

#### Basic installation examples

This example demonstrates how to perform a basic ROCm installation with auto-detected GPU architecture:

```shell
bash rocm-installer-7.14.0-7.run rocm
```

This example demonstrates how to perform a typical ROCm installation with dependencies installed, specific GPU architecture,
and post-install configuration:

```shell
bash rocm-installer-7.14.0-7.run deps=install gfx=gfx942 rocm
```

This example demonstrates how to install ROCm to a custom location with GPU access configuration:

```shell
bash rocm-installer-7.14.0-7.run deps=install target="$HOME/myrocm" gfx=gfx942 rocm gpu-access=all
```

#### Multi-architecture installation examples

Install multiple GPU architectures:

```shell
bash rocm-installer-7.14.0-7.run gfx=gfx942,gfx950,gfx1100 rocm
```

Install all available architectures:

```shell
bash rocm-installer-7.14.0-7.run gfx=all rocm
```

#### Component selection examples

Install core components only (default):

```shell
bash rocm-installer-7.14.0-7.run gfx=gfx942 rocm
```

Install core SDK with dependencies:

```shell
bash rocm-installer-7.14.0-7.run deps=install compo=core-sdk gfx=gfx942 rocm
```

Install multiple components:

```shell
bash rocm-installer-7.14.0-7.run compo=core,dev-tools gfx=gfx942 rocm
```

#### Graphics support example

Install with graphics support for Mesa/OpenGL:

```shell
bash rocm-installer-7.14.0-7.run deps=install gfx=gfx942 rocm graphics
```

#### AMD GPU Driver installation examples

Install AMD GPU Driver only:

```shell
bash rocm-installer-7.14.0-7.run amdgpu
```

Install AMD GPU Driver with dependencies:

```shell
bash rocm-installer-7.14.0-7.run deps=install amdgpu
```

#### Combined installation examples

Install both ROCm and AMD GPU Driver:

```shell
bash rocm-installer-7.14.0-7.run deps=install gfx=gfx942 rocm amdgpu gpu-access=all
```

Install to custom location with all components:

```shell
bash rocm-installer-7.14.0-7.run deps=install target="$HOME/myrocm" gfx=gfx942 rocm amdgpu gpu-access=all
```

#### Runfile options

The ROCm Runfile Installer is a self-extracting `.run` file with a few options for controlling the extraction process.
When the installer `.run` file starts execution, the extraction process begins with a checksum validation to verify
the integrity of the `.run` file. If there are no errors, it begins extracting and decompressing the package contents.
The contents are output to a new directory named `rocm-installer`, located in the current working directory.

Usually, when extraction and decompression are complete, execution begins automatically by either starting up the GUI
(if no arguments are provided) or executing the `rocm-installer.sh` script with all command line arguments.
The `rocm-installer.sh` script is in the extracted `rocm-installer` directory.

When the GUI exits or the command line completes execution, the Runfile installer will automatically clean up the
`rocm-installer` directory and delete all content except for the log files and the `deps_list.txt` file.

In some cases, you might want to execute multiple commands from the same extraction and save the time required
to verify the checksum and extract and decompress the package contents of the `.run` file.
Two command line options let you disable the `.run` cleanup process: `noexec` and `noexec-cleanup`.

* `noexec`

The `noexec` option is a single command line argument that lets the checksum and extraction process complete
  and then exits without starting the GUI or executing the `rocm-installer.sh` script.
  All content will be maintained after the exit. You can then use the `rocm-installer.sh` script directly from the
  command line without specifying the `.run` file name.

For example, extract the `.run` file and then use `rocm-installer.sh` instead of `rocm-installer-7.14.0-7.run`
  to install ROCm and the AMD GPU Driver separately:
  ```shell
  bash rocm-installer-7.14.0-7.run noexec
  cd rocm-installer
  bash rocm-installer.sh gfx=gfx942 rocm
  bash rocm-installer.sh amdgpu
  ```

#### NOTE
 If the `noexec` option is used, all other `<options>` on the command line will be ignored.
* `noexec-cleanup`

The `noexec-cleanup` option disables the cleanup process after the GUI or command line interface exits.
  Unlike the `noexec` option, all command line arguments are processed as normal,
  but no content is deleted upon exit or completion. At this point, you can switch to using the
  `rocm-installer.sh` script within the `rocm-installer` directory to avoid re-extracting the contents.

#### Dependency options

Like the GUI **Pre-Install Configuration** menu, the command line interface can list,
validate, and install the required dependencies. At the command line, add the `deps=<arg> <compo>` option to the
list of `<options>` for the `.run` file. The `<compo>` option can include any combination of the
`rocm`, `amdgpu`, and `graphics` tags.

The `graphics` option is intended for non-Instinct GPUs such as AMD Radeon or Ryzen GPUs/APUs to enable mixed
graphics and compute support for Mesa/OpenGL workloads.

* `deps=list <rocm/amdgpu/graphics>`

This dependency option lists all the dependencies required for the `.run` file-based ROCm installation,
  AMDGPU installation, graphics support, or any combination. It lists all required (Debian or RPM) packages that require
  pre-installation on the system. The additional `rocm`, `amdgpu`, or `graphics` parameter is a requirement for the
  `deps=list` option that instructs the installer to list only the dependencies required by ROCm, the AMD GPU Driver,
  graphics support, or any combination.

Running `deps=list` causes the installer to quit after listing the dependencies. `deps=list`
  can combine `rocm`, `amdgpu`, and `graphics` and output the combined list of required dependencies.

Use `deps=list <rocm/amdgpu/graphics>` as a single `<options>` parameter:
 ```shell
 bash rocm-installer-7.14.0-7.run deps=list rocm
 bash rocm-installer-7.14.0-7.run deps=list amdgpu
 bash rocm-installer-7.14.0-7.run deps=list graphics
 bash rocm-installer-7.14.0-7.run deps=list rocm amdgpu graphics
 ```

#### NOTE
 The list of required packages can be installed separately from the Runfile Installer.
* `deps=validate <rocm/amdgpu/graphics>`

This dependency option verifies whether any of the *required* ROCm, AMD GPU Driver, or graphics support packages in the
  dependency list are already installed on the system running the command.
  The output is a list of any missing dependency packages that require installation.

Running `deps=validate` causes the installer to quit after listing the missing dependencies.
  `deps=validate` can combine `rocm`, `amdgpu`, and `graphics` and output the combined list of missing dependencies.

Use `deps=validate <rocm/amdgpu/graphics>` as a single `<options>` parameter:
 ```shell
 bash rocm-installer-7.14.0-7.run deps=validate rocm
 bash rocm-installer-7.14.0-7.run deps=validate amdgpu
 bash rocm-installer-7.14.0-7.run deps=validate graphics
 bash rocm-installer-7.14.0-7.run deps=validate rocm amdgpu graphics
 ```

#### NOTE
  The list of missing packages can be installed separately from the Runfile Installer.
* `deps=install`

This dependency option validates and installs any required packages in the dependency list for
 ROCm, the AMD GPU Driver, graphics support, or any combination that are missing on the system running the command.
 This dependency option is not a single `<options>` parameter and can be added to a list of other options for the installer.
 The `deps=install` option expects at least one of the `rocm`, `amdgpu`, or `graphics` `<options>` parameters
 to also be present in the list to enable the pre-installation of required dependencies before the Runfile Installer
 installs ROCm, the AMD GPU Driver, or graphics support.

For example, to install the dependencies and ROCm, the command line is as follows:
  ```shell
  bash rocm-installer-7.14.0-7.run deps=install rocm
  ```

To install the dependencies and the AMD GPU Driver, the command line is as follows:
  ```shell
  bash rocm-installer-7.14.0-7.run deps=install amdgpu
  ```

To install the dependencies for graphics support, the command line is as follows:
  ```shell
  bash rocm-installer-7.14.0-7.run deps=install graphics
  ```

To install the dependencies for ROCm, AMD GPU Driver, and graphics support, the command line is as follows:
  ```shell
  bash rocm-installer-7.14.0-7.run deps=install rocm amdgpu graphics
  ```

#### NOTE
  The installer can be set to only install dependencies and then quit. In this case, add `-only` to the `deps=install` option:
  ```shell
  bash rocm-installer-7.14.0-7.run deps=install-only rocm
  bash rocm-installer-7.14.0-7.run deps=install-only amdgpu
  bash rocm-installer-7.14.0-7.run deps=install-only graphics
  bash rocm-installer-7.14.0-7.run deps=install-only rocm amdgpu graphics
  ```

#### Install options

Like the GUI **ROCm Options** menu, the ROCm Runfile Installer can be configured to install ROCm with specific
architectures, components, and installation locations.

* `rocm`

At the command line, add the `rocm` option to enable ROCm installation.

#### NOTE
 This option must be in the list of `.run <options>` for ROCm component installation.

For example, to install ROCm with auto-detected GPU architecture:
 ```shell
 bash rocm-installer-7.14.0-7.run rocm
 ```
* `target=<directory>`

This install option is used to set the target directory where ROCm will be installed.
  The `target` option is only used as an option for ROCm installation and is not required for an AMD GPU Driver install.

When `target=<directory>` is not specified in the `<options>` list, the installer uses the default installation path of `/opt`.
 This results in a ROCm installation at `/opt/rocm/core-x.y.z`.

You can change the default location and set the ROCm component installation directory using
 the `target=<directory>` option. The `<directory>` argument must be a valid and absolute path to
 a directory on the system executing the ROCm Runfile Installer.

For example, to install ROCm to the default `/opt/rocm/core-x.y.z` location (explicit):
  ```shell
  bash rocm-installer-7.14.0-7.run target="/opt" gfx=gfx942 rocm
  ```

To install ROCm to a directory called `amd/myrocm` in the `$USER` directory:
 ```shell
 bash rocm-installer-7.14.0-7.run target="$HOME/myrocm" gfx=gfx942 rocm
 ```
* `gfx=<arch>`

This install option specifies the GPU architecture to install. The installer includes architecture-specific components
  for different AMD GPU families.

Single architecture installation:
  ```shell
  bash rocm-installer-7.14.0-7.run gfx=gfx942 rocm
  bash rocm-installer-7.14.0-7.run gfx=gfx1100 rocm
  bash rocm-installer-7.14.0-7.run gfx=gfx950 rocm
  ```

Multiple architecture installation (comma-separated):
  ```shell
  bash rocm-installer-7.14.0-7.run gfx=gfx942,gfx950,gfx1100 rocm
  ```

Install all available architectures:
  ```shell
  bash rocm-installer-7.14.0-7.run gfx=all rocm
  ```

Auto-detect GPU (default if `gfx=` not specified):
  ```shell
  bash rocm-installer-7.14.0-7.run rocm
  ```

Architecture information commands:
 ```shell
 bash rocm-installer-7.14.0-7.run gfx=show # Detect and display GPU info
 bash rocm-installer-7.14.0-7.run gfx=list # List available architectures
 bash rocm-installer-7.14.0-7.run gfx=list-installed # Show installed architectures
 ```
* `compo=<component_list>`

This install option specifies which ROCm component categories to install. Components can be specified individually
  or as comma-separated lists.

Available components:
  * `core`: Core ROCm components (default)
  * `core-dev`: Core development components
  * `dev-tools`: Developer tools
  * `core-sdk`: Core SDK components
  * `opencl`: OpenCL runtime

Single component installation:
  ```shell
  bash rocm-installer-7.14.0-7.run compo=core gfx=gfx942 rocm
  bash rocm-installer-7.14.0-7.run compo=core-sdk gfx=gfx942 rocm
  ```

Multiple component installation (comma-separated):
  ```shell
  bash rocm-installer-7.14.0-7.run compo=core,dev-tools gfx=gfx942 rocm
  bash rocm-installer-7.14.0-7.run compo=core-sdk,opencl gfx=gfx942 rocm
  ```

List available components:
  ```shell
  bash rocm-installer-7.14.0-7.run compo=list
  ```

If `compo=` is not specified, the installer defaults to installing the `core` component.
* `graphics`

This install option enables graphics support for Mesa/OpenGL workloads. When specified, the installer includes
  the `amdgpu-lib` package.
  ```shell
  bash rocm-installer-7.14.0-7.run gfx=gfx942 rocm graphics
  bash rocm-installer-7.14.0-7.run deps=install gfx=gfx942 rocm graphics
  ```
* `amdgpu`

At the command line, add the `amdgpu` option to enable AMD GPU Driver installation.

#### NOTE
 This option must be in the list of `.run <options>` for ROCm component installation.
 The AMD GPU Driver is installed for the currently running Linux kernel version. To install the AMD GPU Driver for a different kernel,
 reboot to the specific Linux kernel and reinstall the AMD GPU Driver using the runfile.

For example, to install the AMD GPU Driver with no other options:
  ```shell
  bash rocm-installer-7.14.0-7.run amdgpu
  ```

#### NOTE
 Both `rocm` and `amdgpu` can be combined in the `<options>` list to install both components.
 For this case, the command line is as follows:
 ```shell
 bash rocm-installer-7.14.0-7.run gfx=gfx942 rocm amdgpu
 ```
* `force`

The `force` install option can be added to the `<options>` list for the case of multiple
 pre-existing Runfile installs of ROCm. Add this option to disable any installer prompts that ask
 for confirmation to continue with the ROCm install if there are currently one or more Runfile
 ROCm installations already on the system.

#### Post-install options

Similar to the GUI **Post-Install Options** menu, the post-install options can configure the ROCm Runfile Installer
to apply additional post-installation options after completing the installation.
At the command line, add one or more of the post-installation options to the `<options>` list for the `.run` file.

* `postrocm`

This post-install option applies any post-installation configuration settings for ROCm following installation on the target system.
  The post-installation configuration includes any symbolic link creation, library configuration, and script execution
  required to use the ROCm runtime.

By default, post-install runs automatically when you use the `rocm` install option. You only need to explicitly use
  `postrocm` in these scenarios:
  * You installed with `nopostrocm` and now want to run post-install separately.
  * You want to run post-install on an existing installation.

To run post-install separately after installing with `nopostrocm`:
  ```shell
  bash rocm-installer-7.14.0-7.run gfx=gfx942 rocm nopostrocm
  bash rocm-installer-7.14.0-7.run postrocm
  ```

The post-install takes into account the install location of ROCm, which you may specify with the addition of the `target=` option.
  By default, if `target=` is not provided, the default location of `/opt` will be used to apply the post-installation:
  ```shell
  bash rocm-installer-7.14.0-7.run postrocm
  ```

For installation locations other than `/opt`, use the `target=` option:
  ```shell
  bash rocm-installer-7.14.0-7.run target=/custom/path postrocm
  ```

ROCm installations using the Runfile installer can specify not only the location, but also the GPU architecture (`gfx=`)
  and ROCm components (`compo=`). Post-installation takes this into account and will auto-detect the architectures and
  components installed at a location if not provided with the `postrocm` option. However, if a user needs to specify
  either architecture or components, then `gfx=` and `compo=` options may be used with the `postrocm` option.

Run post-install with a specific architecture:
  ```shell
  bash rocm-installer-7.14.0-7.run gfx=gfx942 postrocm
  ```

Run post-install with a specific component:
  ```shell
  bash rocm-installer-7.14.0-7.run compo=core-sdk postrocm
  ```

#### NOTE
  Adding the `postrocm` option (or relying on the default behavior with `rocm`) is highly recommended to guarantee
  proper functioning of the ROCm components and applications.
* `nopostrocm`

This post-install option disables the automatic post-install configuration that normally runs when you use the
  `rocm` install option. Use this when you want to install ROCm without running the post-install scripts.
  ```shell
  bash rocm-installer-7.14.0-7.run gfx=gfx942 rocm nopostrocm
  ```

You can run the post-install later using the `postrocm` option (see above).
* `gpu-access=<access_type>`

This post-install option sets the GPU resource access permissions. ROCm runtime libraries and applications
 might need access to the GPU. This requires setting the access permission to the `video` and `render` groups
 using the `<access_type>`.

If the ROCm installation is for a single user, then set the `<access_type>` for the `gpu-access` option to `user`.

For example, to add the current user (`$USER`) to the `video,render` group for GPU access
  for a ROCm installation, the command line is as follows:
  ```shell
  bash rocm-installer-7.14.0-7.run gfx=gfx942 rocm gpu-access=user
  ```

In cases where a system administrator is installing ROCm for multiple users, they might want to enable
 GPU access permission for all users. For this case, set the `<access_type>` for the `gpu-access` option to `all`:
 ```shell
 bash rocm-installer-7.14.0-7.run gfx=gfx942 rocm gpu-access=all
 ```

#### NOTE
 Adding the `gpu-access` option to the `<options>` list is recommended for using ROCm.
 A typical ROCm installation includes one of the `gpu-access` types.

#### Uninstall options

These options configure the ROCm Runfile Installer to uninstall a previous ROCm or AMD GPU Driver installation.

* `uninstall-rocm (target=<directory>) (gfx=<arch>)`

This option configures the ROCm Runfile Installer to uninstall a previous ROCm installation.
  For multi-architecture installations, you can selectively uninstall specific architectures or all architectures.
  * Uninstall with auto-detection of architecture:

To uninstall ROCm from the default location (`/opt/rocm/core-x.y.z`):
    ```shell
    bash rocm-installer-7.14.0-7.run uninstall-rocm
    ```

To uninstall ROCm from a custom location:
    ```shell
    bash rocm-installer-7.14.0-7.run target="$HOME/myrocm/rocm-x.y.z" uninstall-rocm
    ```
  * Uninstall with architecture selection:

Selective uninstall of one or more GPU
    architectures installed to a location. Using `gfx=`, you can specify the architecture or architectures to uninstall.
    Only gfx-specific components will be removed for multi-installs, with common components remaining until the final
    architecture is uninstalled.

To uninstall a specific architecture:
    ```shell
    bash rocm-installer-7.14.0-7.run gfx=gfx1100 uninstall-rocm
    bash rocm-installer-7.14.0-7.run target="$HOME/myrocm" gfx=gfx942 uninstall-rocm
    ```

To uninstall all architectures:
    ```shell
    bash rocm-installer-7.14.0-7.run gfx=all uninstall-rocm
    ```

#### NOTE
  The `uninstall-rocm` option can only remove ROCm if it was installed using the ROCm Runfile Installer.
  Traditional package-based ROCm installs will not be removed.
* `uninstall-amdgpu`

This option configures the ROCm Runfile Installer to uninstall a previous AMD GPU Driver installation.

To uninstall the AMD GPU Driver from the system, use the following command line:
  ```shell
  bash rocm-installer-7.14.0-7.run uninstall-amdgpu
  ```

#### NOTE
  The `uninstall-amdgpu` option can only remove the AMD GPU Driver if it was installed using the ROCm Runfile Installer.
  Traditional package-based AMD GPU Driver installs will not be removed.

#### Information and debug options

The ROCm Runfile Installer command line interface includes options for information output or debugging.

* `findrocm`

This information option searches the install system for any existing installations of ROCm.
  Any install locations are output to the terminal and the type of installation is indicated (Runfile or Package manager).

Use `findrocm` as a standalone `<options>` parameter:
 ```shell
 bash rocm-installer-7.14.0-7.run findrocm
 ```
* `manifest`

This information option lists the version of each ROCm component included in the installer.
  The `manifest` option causes the Runfile Installer to quit after listing the ROCm components.

Use `manifest` as a standalone `<options>` parameter:
 ```shell
 bash rocm-installer-7.14.0-7.run manifest
 ```

To list components for a specific architecture:
  ```shell
  bash rocm-installer-7.14.0-7.run manifest=gfx942
  bash rocm-installer-7.14.0-7.run manifest=gfx1100
  bash rocm-installer-7.14.0-7.run manifest=base
  ```
* `buildinfo`

This information option displays theRock build information, including commit hash, GitHub run ID, build date,
  and other build metadata.

Use `buildinfo` as a standalone `<options>` parameter:
 ```shell
 bash rocm-installer-7.14.0-7.run buildinfo
 ```
* `prompt`

This debug option enables user prompts in the installer. At specific, critical points of the installation process,
  the installer halts execution and prompts the user to either continue with the install or exit.
* `assumeyes`

This automation option automatically answers yes to all prompts. This is useful for automation and scripting scenarios
  where user interaction is not desired.
  ```shell
  bash rocm-installer-7.14.0-7.run deps=install gfx=gfx942 rocm assumeyes
  ```
* `verbose`

This debug option enables verbose logging during ROCm Runfile Installer execution.

For example, to install ROCm with user prompts and verbose logging:
  ```shell
  bash rocm-installer-7.14.0-7.run target="/opt" gfx=gfx942 rocm prompt verbose
  ```

## Log files

By default, the ROCm Runfile Installer GUI and command line interfaces record the output execution in log files.
These log files are created in the `rocm-installer/logs` directory.

The `rocm-installer` directory is created when the ROCm Runfile Installer self-extracts
to the current working directory where it is being executed.

## Troubleshooting

The following are common issues and solutions for the ROCm Runfile Installer.

* **Issue:** Installer fails with “missing dependencies” error

**Solution:** Use `deps=validate rocm` to see which dependencies are missing, then either:
  * Install them manually using your package manager
  * Use `deps=install rocm` to have the installer install them (requires internet connection)
* **Issue:** GPU not detected or wrong architecture selected

**Solution:** Use `gfx=show` to see detected GPUs, then manually specify the correct architecture:
  ```shell
  bash rocm-installer-7.14.0-7.run gfx=gfx942 rocm
  ```
* **Issue:** Post-install configuration was skipped

**Solution:** Run post-install separately:
  ```shell
  bash rocm-installer-7.14.0-7.run postrocm
  ```
* **Issue:** Cannot access GPU after installation

**Solution:** Ensure GPU access permissions are set. Either:
  * Add your user to the video/render groups: `bash rocm-installer-7.14.0-7.run gpu-access=user`
  * Grant access to all users: `bash rocm-installer-7.14.0-7.run gpu-access=all`
  * Log out and log back in for group membership changes to take effect
* **Issue:** Multi-architecture installation conflicts

**Solution:**
  * Check installed architectures: `bash rocm-installer-7.14.0-7.run gfx=list-installed`
  * Remove specific architecture if needed: `bash rocm-installer-7.14.0-7.run gfx=gfx1100 uninstall-rocm`
  * Reinstall with correct architecture selection

---

Source: https://rocm.docs.amd.com/en/latest/install/build-from-source.html

# Build the ROCm Core SDK from source

You can build the ROCm Core SDK from source using the open-source unified build
system [TheRock](https://github.com/ROCm/TheRock). To learn more about the
motivation and architecture behind this system, see [ROCm Technology Preview:
ROCm Core SDK and TheRock Build System](https://rocm.blogs.amd.com/software-tools-optimization/therock/README.html).

This page consists mainly of references to [TheRock’s README](https://github.com/ROCm/TheRock?tab=readme-ov-file#building-from-source)
and [supporting development manuals](https://github.com/ROCm/TheRock/blob/main/README.md#development-manuals),
which provide up-to-date build instructions and guidance for supported
platforms. See [TheRock Development Guide](https://github.com/ROCm/TheRock/blob/main/docs/development/development_guide.md#therock-development-guide)
to learn about the overall build architecture.

#### NOTE
Building from source is recommended only if you need custom builds or are
contributing to ROCm development.
For most users, installing from official AMD releases is faster and easier.
See [Install AMD ROCm 7.14.0](https://rocm.docs.amd.com/en/latest/install/rocm.html) for installation instructions.

## Prerequisites

A successful build depends on a correctly configured environment. Before you
begin, ensure your system meets all hardware and software requirements. Review the
following resources:

* [Environment setup guide](https://github.com/ROCm/TheRock/blob/main/docs/environment_setup_guide.md)
* [ROCm 7.14.0 compatibility matrix](https://rocm.docs.amd.com/en/latest/compatibility/compatibility-matrix.html)

## High-level build process

For an overview of the build architecture, start with TheRock’s [development guide](https://github.com/ROCm/TheRock/blob/main/docs/development/development_guide.md).
While specific commands vary by platform, the general workflow for building
from source involves these stages:

1. Clone the repository and install build dependencies. See [Platform-specific setup](#build-from-src-plat-setup).
2. Configure the build: Use CMake feature flags to configure the build. This
   step allows you to target specific platforms, build subsets of ROCm Core SDK
   components, and toggle component features. See [Build configuration](https://github.com/ROCm/TheRock/blob/main/README.md#build-configuration)
   for optional and required build flags.
3. Execute the build: Run the build command to compile the source code. This
   can be a time- and resource-intensive process. See [CMake build usage](https://github.com/ROCm/TheRock/blob/main/README.md#cmake-build-usage).
4. After a successful build, the outputs are available for use in downstream
   tasks. To learn more about build outputs, see the relevant [TheRock
   documentation](https://github.com/ROCm/TheRock/blob/main/docs/development/artifacts.md).
   Common post-build tasks include:
   * Using `build/dist/rocm`: When the build completes, you should have a
     build of ROCm in the `build/dist/rocm/` directory. See [Using installed
     tarballs](https://github.com/ROCm/TheRock/blob/main/RELEASES.md#using-installed-tarballs)
     for more information.
   * Building Python packages: Prepare the build artifacts for distribution as
     Python packages. See [Building Python packages](https://github.com/ROCm/TheRock/blob/main/docs/packaging/python_packaging.md#building-packages).
   * Building PyTorch: Build a compatible PyTorch version against ROCm wheels.
     See the [PyTorch build instructions](https://github.com/ROCm/TheRock/tree/main/external-builds/pytorch#build-instructions).

## Platform-specific setup

### ManyLinux

On Linux, it’s recommended to build with ManyLinux to produce binaries that are
portable across Ubuntu and other Linux distributions. To learn more about what
a ROCm ManyLinux build entails, see [ManyLinux builds](https://github.com/ROCm/TheRock/blob/main/docs/design/manylinux_builds.md).
Refer to [ManyLinux x86_64](https://github.com/ROCm/TheRock/blob/main/docs/environment_setup_guide.md#manylinux-x86-64)
in the environment setup guide.

### Ubuntu 24.04

TheRock provides detailed instructions and scripts for preparing an Ubuntu
24.04 system, including installing necessary packages via apt.
Refer to [Setup — Ubuntu 24.04](https://github.com/ROCm/TheRock?tab=readme-ov-file#setup---ubuntu-2404)
in the TheRock repository for guidance.

### Windows 11

For setup instructions on Windows 11 using Visual Studio 2022, see
[Setup — Windows 11](https://github.com/ROCm/TheRock?tab=readme-ov-file#setup---windows-11-vs-2022)
in the TheRock repository.

#### NOTE
For details on supported configurations, known issues, and other
Windows-specific considerations, review the [Windows support](https://github.com/ROCm/TheRock/blob/main/docs/development/windows_support.md)
documentation.

---

Source: https://rocm.docs.amd.com/en/latest/components/core.html

# ROCm Core SDK components

The ROCm Core SDK is the foundation of the ROCm software stack. It provides the
libraries, runtimes, compilers, and tools needed to develop and run GPU-accelerated
applications on AMD hardware.

## Math and compute libraries

A comprehensive set of GPU-accelerated math libraries covering dense and sparse
linear algebra, FFTs, random number generation, and more.

* Libraries prefixed with `roc*` are native, high-performance
  implementations written in HIP specifically for AMD GPUs.
* Libraries prefixed with `hip*` are portable wrappers that implement
  NVIDIA CUDA-equivalent APIs, allowing CUDA applications to be ported to AMD GPUs
  with minimal code changes.

Libraries include:

* [Composable Kernel](https://rocm.docs.amd.com/projects/composable_kernel/en/docs-7.14.0) – Provides a programming model for writing performance critical kernels for machine learning workloads across multiple architectures.
* [hipBLAS](https://rocm.docs.amd.com/projects/hipBLAS/en/docs-7.14.0) – BLAS-marshalling library that supports rocBLAS and cuBLAS backends.
* [hipBLASLt](https://rocm.docs.amd.com/projects/hipBLASLt/en/docs-7.14.0) – Provides general matrix-matrix operations with a flexible API and extends functionalities beyond traditional BLAS library.
* [hipCUB](https://rocm.docs.amd.com/projects/hipCUB/en/docs-7.14.0) – Thin header-only wrapper library on top of rocPRIM or CUB that allows project porting using the CUB library to the HIP layer.
* [hipFFT](https://rocm.docs.amd.com/projects/hipFFT/en/docs-7.14.0) – Fast Fourier Transforms (FFT)-marshalling library that supports rocFFT or cuFFT backends.
* [hipRAND](https://rocm.docs.amd.com/projects/hipRAND/en/docs-7.14.0) – Ports CUDA applications that use the cuRAND library into the HIP layer.
* [hipSOLVER](https://rocm.docs.amd.com/projects/hipSOLVER/en/docs-7.14.0) – LAPACK-marshalling library that supports rocSOLVER and cuSOLVER backends.
* [hipSPARSE](https://rocm.docs.amd.com/projects/hipSPARSE/en/docs-7.14.0) – SPARSE-marshalling library that supports rocSPARSE and cuSPARSE backends.
* [MIOpen](https://rocm.docs.amd.com/projects/MIOpen/en/docs-7.14.0) – An open source deep-learning library.
* [rocBLAS](https://rocm.docs.amd.com/projects/rocBLAS/en/docs-7.14.0) – BLAS implementation (in the HIP programming language) on the ROCm runtime and toolchains.
* [rocFFT](https://rocm.docs.amd.com/projects/rocFFT/en/docs-7.14.0) – Software library for computing Fast Fourier Transforms (FFTs) written in HIP.
* [rocPRIM](https://rocm.docs.amd.com/projects/rocPRIM/en/docs-7.14.0) – Header-only library for HIP parallel primitives.
* [rocRAND](https://rocm.docs.amd.com/projects/rocRAND/en/docs-7.14.0) – Provides functions that generate pseudorandom and quasirandom numbers.
* [rocSOLVER](https://rocm.docs.amd.com/projects/rocSOLVER/en/docs-7.14.0) – An implementation of LAPACK routines on ROCm software, implemented in the HIP programming language and optimized for AMD’s latest discrete GPUs.
* [rocSPARSE](https://rocm.docs.amd.com/projects/rocSPARSE/en/docs-7.14.0) – Exposes a common interface that provides BLAS for sparse computation implemented on ROCm runtime and toolchains (in the HIP programming language).
* [rocThrust](https://rocm.docs.amd.com/projects/rocThrust/en/docs-7.14.0) – Parallel algorithm library.
* [rocWMMA](https://rocm.docs.amd.com/projects/rocWMMA/en/docs-7.14.0) – C++ library for accelerating mixed-precision matrix multiply-accumulate (MMA) operations.

## Communication libraries

* [RCCL](https://rocm.docs.amd.com/projects/rccl/en/docs-7.14.0) – Standalone library that provides multi-GPU and multi-node collective communication primitives.
* [rocSHMEM](https://rocm.docs.amd.com/projects/rocSHMEM/en/docs-7.14.0) – An intra-kernel networking library that provides GPU-centric networking througithub_repo an OpenSHMEM-like interface.

## Runtime and compilers

* [HIP](https://rocm.docs.amd.com/projects/HIP/en/docs-7.14.0) – A C++ runtime API and kernel programming language designed for AMD GPUs. By providing an interface closely aligned with NVIDIA CUDA, HIP allows developers to write portable applications and efficiently migrate existing CUDA code to AMD platforms.
* [HIPIFY](https://rocm.docs.amd.com/projects/HIPIFY/en/docs-7.14.0) – Translates CUDA source code into portable HIP C++.
* [LLVM](https://rocm.docs.amd.com/projects/llvm-project/en/docs-7.14.0) – AMD’s LLVM-based compiler infrastructure, including the ROCm device compiler (amdclang), which compiles HIP and OpenCL code for AMD GPUs.

## Profiling and debugging tools

* [ROCdbgapi](https://rocm.docs.amd.com/projects/ROCdbgapi/en/docs-7.14.0) – ROCm debugger API library.
* [ROCgdb](https://rocm.docs.amd.com/projects/ROCgdb/en/docs-7.14.0) – Source-level debugger for Linux, based on the GNU Debugger (GDB).
* [ROCm Compute Profiler](https://rocm.docs.amd.com/projects/rocprofiler-compute/en/docs-7.14.0) – Kernel-level profiling for machine learning and higithub_repo performance computing (HPC) workloads.
* [ROCm Systems Profiler](https://rocm.docs.amd.com/projects/rocprofiler-systems/en/docs-7.14.0) – Comprehensive profiling and tracing of applications running on the CPU or the CPU and GPU.
* [ROCprofiler-SDK](https://rocm.docs.amd.com/projects/rocprofiler-sdk/en/docs-7.14.0) – Toolkit for developing analysis tools for profiling and tracing GPU compute applications.
* [ROCR Debug Agent](https://rocm.docs.amd.com/projects/rocr_debug_agent/en/docs-7.14.0) – Prints the state of all AMD GPU wavefronts that caused a queue error by sending a SIGQUIT signal to the process while the program is running.

## Control and monitoring tools

* [AMD SMI](https://rocm.docs.amd.com/projects/amdsmi/en/docs-7.14.0) – System management interface to control AMD GPU settings, monitor performance, and retrieve device and process information.
* [ROCm Data Center Tool](https://rocm.docs.amd.com/projects/rdc/en/docs-7.14.0) – Simplifies administration and addresses key infrastructure challenges in AMD GPUs in cluster and data-center environments.
* [rocminfo](https://rocm.docs.amd.com/projects/rocminfo/en/docs-7.14.0) – Reports system information.

## Media libraries

* [rocDecode](https://rocm.docs.amd.com/projects/rocDecode/en/docs-7.14.0) – Higithub_repo-performance SDK for access to video decoding features on AMD GPUs.
* [rocJPEG](https://rocm.docs.amd.com/projects/rocJPEG/en/docs-7.14.0) – Library for decoding JPEG images on AMD GPUs.

## Storage

* [hipFile](https://rocm.docs.amd.com/projects/hipFile/en/docs-7.14.0) – AMD’s Infinity Storage library that provides direct-to-GPU I/O for the ROCm platform.

---

Source: https://rocm.docs.amd.com/en/latest/components/math-and-compute-libs.html

# ROCm math and compute libraries

ROCm math and compute libraries provide GPU-accelerated implementations of
common numerical operations including dense and sparse linear algebra, FFTs,
random number generation, and deep learning primitives.

Libraries prefixed with `roc*` are native, high-performance implementations
written in HIP specifically for AMD GPUs. Libraries prefixed with `hip*` are
portable wrappers that implement NVIDIA CUDA-equivalent APIs, allowing CUDA applications
to be ported to AMD GPUs with minimal code changes.

* [Composable Kernel](https://rocm.docs.amd.com/projects/composable_kernel/en/latest/index.html) – Provides a programming
  model for writing performance critical kernels for machine learning workloads
  across multiple architectures.
* [hipBLAS](https://rocm.docs.amd.com/projects/hipBLAS/en/latest/index.html) – BLAS-marshalling library that supports
  rocBLAS and cuBLAS backends.
* [hipBLASLt](https://rocm.docs.amd.com/projects/hipBLASLt/en/latest/index.html) – Provides general matrix-matrix
  operations with a flexible API and extends functionalities beyond traditional
  BLAS library.
* [hipCUB](https://rocm.docs.amd.com/projects/hipCUB/en/latest/index.html) – Thin header-only wrapper library on top of
  rocPRIM or CUB that allows project porting using the CUB library to the HIP
  layer.
* [hipFFT](https://rocm.docs.amd.com/projects/hipFFT/en/latest/index.html) – Fast Fourier transforms (FFT)-marshalling
  library that supports rocFFT or cuFFT backends.
* [hipRAND](https://rocm.docs.amd.com/projects/hipRAND/en/latest/index.html) – Ports CUDA applications that use the
  cuRAND library into the HIP layer.
* [hipSOLVER](https://rocm.docs.amd.com/projects/hipSOLVER/en/latest/index.html) – An LAPACK-marshalling library that
  supports rocSOLVER and cuSOLVER backends.
* [hipSPARSE](https://rocm.docs.amd.com/projects/hipSPARSE/en/latest/index.html) – SPARSE-marshalling library that
  supports rocSPARSE and cuSPARSE backends.
* [hipSPARSELt](https://rocm.docs.amd.com/projects/hipSPARSELt/en/latest/index.html) – SPARSE-marshalling library with
  multiple supported backends.
* [MIOpen](https://rocm.docs.amd.com/projects/MIOpen/en/latest/index.html) – An open source deep-learning library.
* [rocBLAS](https://rocm.docs.amd.com/projects/rocBLAS/en/latest/index.html) – BLAS implementation (in the HIP
  programming language) on the ROCm runtime and toolchains.
* [rocFFT](https://rocm.docs.amd.com/projects/rocFFT/en/latest/index.html) – Software library for computing fast Fourier
  transforms (FFTs) written in HIP.
* [rocRAND](https://rocm.docs.amd.com/projects/rocRAND/en/latest/index.html) – Provides functions that generate
  pseudorandom and quasirandom numbers.
* [rocSOLVER](https://rocm.docs.amd.com/projects/rocSOLVER/en/latest/index.html) – An implementation of LAPACK routines
  on ROCm software, implemented in the HIP programming language and optimized
  for AMD’s latest discrete GPUs.
* [rocSPARSE](https://rocm.docs.amd.com/projects/rocSPARSE/en/latest/index.html) – Exposes a common interface that
  provides BLAS for sparse computation implemented on ROCm runtime and
  toolchains (in the HIP programming language).
* [rocPRIM](https://rocm.docs.amd.com/projects/rocPRIM/en/latest/index.html) – Header-only library for HIP parallel
  primitives.
* [rocThrust](https://rocm.docs.amd.com/projects/rocThrust/en/latest/index.html) – Parallel algorithm library.
* [rocWMMA](https://rocm.docs.amd.com/projects/rocWMMA/en/latest/index.html) – C++ library for accelerating
  mixed-precision matrix multiply-accumulate (MMA) operations.

---

Source: https://rocm.docs.amd.com/en/latest/components/communication-libs.html

# ROCm communication libraries

ROCm communication libraries provide high-performance multi-GPU and multi-node
communication primitives for deep learning and HPC workloads.

* [RCCL](https://rocm.docs.amd.com/projects/rccl/en/latest/index.html) – Standalone library that provides multi-GPU and
  multi-node collective communication primitives.
* [rocSHMEM](https://rocm.docs.amd.com/projects/rocSHMEM/en/latest/index.html) – An intra-kernel networking library that
  provides GPU-centric networking through an OpenSHMEM-like interface.

---

Source: https://rocm.docs.amd.com/en/latest/components/media-libs.html

# ROCm media libraries

ROCm media libraries provide GPU-accelerated video decoding and image
processing capabilities for AMD GPUs.

* [rocDecode](https://rocm.docs.amd.com/projects/rocDecode/en/latest/index.html) – High-performance SDK for access to
  video decoding features on AMD GPUs.
* [rocJPEG](https://rocm.docs.amd.com/projects/rocJPEG/en/latest/index.html) – Library for decoding JPG images on AMD
  GPUs.

---

Source: https://rocm.docs.amd.com/en/latest/components/storage-libs.html

# ROCm storage libraries

AMD Infinity Storage (AIS) Libraries provide optimized communication and data transfer between storage and AMD GPUs for AI/ML and HPC workloads.

* [hipFile](https://rocm.docs.amd.com/projects/hipFile/en/latest/index.html) – A runtime library for direct-to-GPU I/O operations.

---

Source: https://rocm.docs.amd.com/en/latest/components/runtimes-and-compilers.html

# ROCm runtimes and compilers

ROCm runtimes and compilers provide the core execution environment and
programming tools for GPU application development on AMD hardware.

* [HIP](https://rocm.docs.amd.com/projects/HIP/en/latest/index.html) – A C++ runtime API and kernel programming language
  designed for AMD GPUs. By providing an interface closely aligned with NVIDIA
  CUDA, HIP allows developers to write portable applications and efficiently
  migrate existing CUDA code to AMD platforms.
* [HIPIFY](https://rocm.docs.amd.com/projects/HIPIFY/en/latest/index.html) – Translates CUDA source code into portable
  HIP C++.
* [LLVM](https://rocm.docs.amd.com/projects/llvm-project/en/latest/index.html) – AMD’s LLVM-based compiler
  infrastructure, including the ROCm device compiler (`amdclang`), which
  compiles HIP and OpenCL code for AMD GPUs.

---

Source: https://rocm.docs.amd.com/en/latest/components/profilers-and-debuggers.html

# ROCm profiling and debugging tools

ROCm profiling and debugging tools help you measure GPU application performance,
identify bottlenecks, and diagnose execution faults.

* [ROCm Compute Profiler](https://rocm.docs.amd.com/projects/rocprofiler-compute/en/latest/index.html)
  (rocprofiler-compute) – Kernel-level profiling for machine learning and
  high performance computing (HPC) workloads.
* [ROCm Systems Profiler](https://rocm.docs.amd.com/projects/rocprofiler-systems/en/latest/index.html)
  (rocprofiler-systems) – Comprehensive profiling and tracing of applications
  running on the CPU or the CPU and GPU.
* [ROCprofiler-SDK](https://rocm.docs.amd.com/projects/rocprofiler-sdk/en/latest/index.html) – Toolkit for developing
  analysis tools for profiling and tracing GPU compute applications.
* [ROCdbgapi](https://rocm.docs.amd.com/projects/ROCdbgapi/en/latest/index.html) – ROCm debugger API library.
* [ROCm Debugger](https://rocm.docs.amd.com/projects/ROCgdb/en/latest/index.html) (ROCgdb) – Source-level debugger for
  Linux, based on the GNU Debugger (GDB).
* [ROCR Debug Agent](https://rocm.docs.amd.com/projects/rocr_debug_agent/en/latest/index.html) – Prints the state of all
  AMD GPU wavefronts that caused a queue error by sending a SIGQUIT signal to
  the process while the program is running.

#### NOTE
[ROCprof Compute Viewer](https://rocm.docs.amd.com/projects/rocprof-compute-viewer/en/latest/) is a tool for visualizing and analyzing GPU thread trace data collected using [rocprofv3](https://rocm.docs.amd.com/projects/rocprofiler-sdk/en/latest/index.html). Note that ROCprof Compute Viewer is in an early access state. Running production workloads is not recommended.

---

Source: https://rocm.docs.amd.com/en/latest/components/control-and-monitoring-tools.html

# ROCm system control and monitoring tools

ROCm control and monitoring tools provide interfaces for inspecting and
managing AMD GPU hardware state in workstation and data center environments.

* [AMD SMI](https://rocm.docs.amd.com/projects/amdsmi/en/latest/index.html) – System management interface to control AMD
  GPU settings, monitor performance, and retrieve device and process
  information.
* [ROCm Data Center Tool](https://rocm.docs.amd.com/projects/rdc/en/latest/index.html) (RDC) – Simplifies administration
  and addresses key infrastructure challenges in AMD GPUs in cluster and
  data-center environments.
* [rocminfo](https://rocm.docs.amd.com/projects/rocminfo/en/latest/index.html) – Reports system information.

---

Source: https://rocm.docs.amd.com/en/latest/components/hpc-sdk/index.html

# AMD ROCm HPC SDK

The AMD ROCm™ open-source software stack is optimized to extract high-performance
computing (HPC) workload performance from AMD Instinct™ GPUs while maintaining
compatibility with industry software frameworks.

ROCm enhances support and access for developers by providing streamlined and improved
tools that significantly increase productivity. Being open-source, ROCm fosters
innovation, differentiation, and collaboration within the developer community, making it
a powerful and accessible solution for leveraging the full potential of AMD GPUs’
capabilities in diverse computational applications.

Some of the most popular HPC frameworks are part of the ROCm HPC SDK, including those
that help parallelize operations across multiple GPUs and servers, handle memory
hierarchies, and solve linear systems. The HPC application catalog provides a vast set of platform-compatible HPC applications,
including those for astrophysics, climate and weather, computational chemistry,
computational fluid dynamics, earth science, genomics, geophysics, molecular dynamics,
and physics computing.

HPC SDK includes [hipTensor](https://github.com/ROCm/rocm-libraries/tree/develop/projects/hiptensor) and [rocALUTION](https://github.com/ROCm/rocm-libraries/tree/develop/projects/rocalution) packaged as part of the installation.

---

Source: https://rocm.docs.amd.com/en/latest/components/hpc-sdk/install.html

# Install ROCm HPC SDK

AMD ROCm HPC SDK provides high-performance computing libraries and tools for
AMD Instinct GPUs. This guide walks you through installing the HPC SDK
alongside ROCm installation on a supported Linux distribution.

The ROCm for HPC applications and containers run on a standard ROCm
installation. See the [Compatibility matrix](https://rocm.docs.amd.com/en/latest/compatibility/compatibility-matrix.html#compat-matrix) for details
on supported hardware and operating systems.

The HPC application containers are published through
[AMD InfinityHub-CI](https://github.com/amd/InfinityHub-CI). Each container
provides parameters to specify source code branches and release versions of ROCm,
OpenMPI, UCX, and Ubuntu.

**w=compute**

**fam=instinct fam=radeon fam=ryzen**

**fam=instinct**

**fam=all**

---

Before installing the HPC SDK, make sure your system meets the ROCm hardware,
software, and driver requirements. For instructions, see [Install AMD ROCm](https://rocm.docs.amd.com/en/latest/install/rocm.html#rocm-install-selector). Use the
selector panel on that page to view instructions appropriate for your system
environment.

## Install HPC SDK

**i=tar**

The standard ROCm tarball installation includes the HPC SDK. No
additional steps are required. For details on ROCm tarball installation,
refer to [Install AMD ROCm](https://rocm.docs.amd.com/en/latest/install/rocm.html#rocm-install) and select Tarball
installation method from installation environment selector.

**i=pkgman**

1. [Install ROCm](https://rocm.docs.amd.com/en/latest/install/rocm.html). Remember to complete the [ROCm
   installation prerequisites](https://rocm.docs.amd.com/en/latest/install/rocm.html#rocm-prerequisites) to install dependencies
   and configure GPU access permissions.

**fam=all**

1. Use the following command to install HPC SDK for all GPU architectures:

**os=ubuntu os=debian**
   ```bash
   sudo apt install amdrocm-hpc7.14 amdrocm-hpc-sdk7.14
   ```

**os=rhel os=rocky-linux os=oracle-linux**
   ```bash
   sudo dnf install amdrocm-hpc7.14 amdrocm-hpc-sdk7.14
   ```

**os=sles**
   ```bash
   sudo zypper install amdrocm-hpc7.14 amdrocm-hpc-sdk7.14
   ```

**gfx=gfx950**

1. Use the following command to install HPC SDK for your `gfx950` GPU:

**os=ubuntu os=debian**
   ```bash
   sudo apt install amdrocm-hpc7.14-gfx950 amdrocm-hpc-sdk7.14-gfx950
   ```

**os=rhel os=rocky-linux os=oracle-linux**
   ```bash
   sudo dnf install amdrocm-hpc7.14-gfx950 amdrocm-hpc-sdk7.14-gfx950
   ```

**os=sles**
   ```bash
   sudo zypper install amdrocm-hpc7.14-gfx950 amdrocm-hpc-sdk7.14-gfx950
   ```

**gfx=gfx942**

1. Use the following command to install HPC SDK for your `gfx942` GPU:

**os=ubuntu os=debian**
   ```bash
   sudo apt install amdrocm-hpc7.14-gfx942 amdrocm-hpc-sdk7.14-gfx942
   ```

**os=rhel os=rocky-linux os=oracle-linux**
   ```bash
   sudo dnf install amdrocm-hpc7.14-gfx942 amdrocm-hpc-sdk7.14-gfx942
   ```

**os=sles**
   ```bash
   sudo zypper install amdrocm-hpc7.14-gfx942 amdrocm-hpc-sdk7.14-gfx942
   ```

**gfx=gfx90a**

1. Use the following command to install HPC SDK for your `gfx90a` GPU:

**os=ubuntu os=debian**
   ```bash
   sudo apt install amdrocm-hpc7.14-gfx90a amdrocm-hpc-sdk7.14-gfx90a
   ```

**os=rhel os=rocky-linux os=oracle-linux**
   ```bash
   sudo dnf install amdrocm-hpc7.14-gfx90a amdrocm-hpc-sdk7.14-gfx90a
   ```

**os=sles**
   ```bash
   sudo zypper install amdrocm-hpc7.14-gfx90a amdrocm-hpc-sdk7.14-gfx90a
   ```

**gfx=gfx908**

1. Use the following command to install HPC SDK for your `gfx908` GPU:

**os=ubuntu os=debian**
   ```bash
   sudo apt install amdrocm-hpc7.14-gfx908 amdrocm-hpc-sdk7.14-gfx908
   ```

**os=rhel os=rocky-linux os=oracle-linux**
   ```bash
   sudo dnf install amdrocm-hpc7.14-gfx908 amdrocm-hpc-sdk7.14-gfx908
   ```

**os=sles**
   ```bash
   sudo zypper install amdrocm-hpc7.14-gfx908 amdrocm-hpc-sdk7.14-gfx908
   ```

**gfx=gfx1200**

1. Use the following command to install HPC SDK for your `gfx1200` GPU:

**os=ubuntu os=debian**
   ```bash
   sudo apt install amdrocm-hpc7.14-gfx1200 amdrocm-hpc-sdk7.14-gfx1200
   ```

**os=rhel os=rocky-linux os=oracle-linux**
   ```bash
   sudo dnf install amdrocm-hpc7.14-gfx1200 amdrocm-hpc-sdk7.14-gfx1200
   ```

**os=sles**
   ```bash
   sudo zypper install amdrocm-hpc7.14-gfx1200 amdrocm-hpc-sdk7.14-gfx1200
   ```

**gfx=gfx1201**

1. Use the following command to install HPC SDK for your `gfx1201` GPU:

**os=ubuntu os=debian**
   ```bash
   sudo apt install amdrocm-hpc7.14-gfx1201 amdrocm-hpc-sdk7.14-gfx1201
   ```

**os=rhel os=rocky-linux os=oracle-linux**
   ```bash
   sudo dnf install amdrocm-hpc7.14-gfx1201 amdrocm-hpc-sdk7.14-gfx1201
   ```

**os=sles**
   ```bash
   sudo zypper install amdrocm-hpc7.14-gfx1201 amdrocm-hpc-sdk7.14-gfx1201
   ```

**gfx=gfx1100**

1. Use the following command to install HPC SDK for your `gfx1100` GPU:

**os=ubuntu os=debian**
   ```bash
   sudo apt install amdrocm-hpc7.14-gfx1100 amdrocm-hpc-sdk7.14-gfx1100
   ```

**os=rhel os=rocky-linux os=oracle-linux**
   ```bash
   sudo dnf install amdrocm-hpc7.14-gfx1100 amdrocm-hpc-sdk7.14-gfx1100
   ```

**os=sles**
   ```bash
   sudo zypper install amdrocm-hpc7.14-gfx1100 amdrocm-hpc-sdk7.14-gfx1100
   ```

**gfx=gfx1101**

1. Use the following command to install HPC SDK for your `gfx1101` GPU:

**os=ubuntu os=debian**
   ```bash
   sudo apt install amdrocm-hpc7.14-gfx1101 amdrocm-hpc-sdk7.14-gfx1101
   ```

**os=rhel os=rocky-linux os=oracle-linux**
   ```bash
   sudo dnf install amdrocm-hpc7.14-gfx1101 amdrocm-hpc-sdk7.14-gfx1101
   ```

**os=sles**
   ```bash
   sudo zypper install amdrocm-hpc7.14-gfx1101 amdrocm-hpc-sdk7.14-gfx1101
   ```

**gfx=gfx1102**

1. Use the following command to install HPC SDK for your `gfx1102` GPU:

**os=ubuntu os=debian**
   ```bash
   sudo apt install amdrocm-hpc7.14-gfx1102 amdrocm-hpc-sdk7.14-gfx1102
   ```

**os=rhel os=rocky-linux os=oracle-linux**
   ```bash
   sudo dnf install amdrocm-hpc7.14-gfx1102 amdrocm-hpc-sdk7.14-gfx1102
   ```

**os=sles**
   ```bash
   sudo zypper install amdrocm-hpc7.14-gfx1102 amdrocm-hpc-sdk7.14-gfx1102
   ```

**gfx=gfx1103**

1. Use the following command to install HPC SDK for your `gfx1103` GPU:

**os=ubuntu os=debian**
   ```bash
   sudo apt install amdrocm-hpc7.14-gfx1103 amdrocm-hpc-sdk7.14-gfx1103
   ```

**os=rhel os=rocky-linux os=oracle-linux**
   ```bash
   sudo dnf install amdrocm-hpc7.14-gfx1103 amdrocm-hpc-sdk7.14-gfx1103
   ```

**os=sles**
   ```bash
   sudo zypper install amdrocm-hpc7.14-gfx1103 amdrocm-hpc-sdk7.14-gfx1103
   ```

**gfx=gfx1030**

1. Use the following command to install HPC SDK for your `gfx1030` GPU:

**os=ubuntu os=debian**
   ```bash
   sudo apt install amdrocm-hpc7.14-gfx1030 amdrocm-hpc-sdk7.14-gfx1030
   ```

**os=rhel os=rocky-linux os=oracle-linux**
   ```bash
   sudo dnf install amdrocm-hpc7.14-gfx1030 amdrocm-hpc-sdk7.14-gfx1030
   ```

**os=sles**
   ```bash
   sudo zypper install amdrocm-hpc7.14-gfx1030 amdrocm-hpc-sdk7.14-gfx1030
   ```

**gfx=gfx1151**

1. Use the following command to install HPC SDK for your `gfx1151` GPU:

**os=ubuntu os=debian**
   ```bash
   sudo apt install amdrocm-hpc7.14-gfx1151 amdrocm-hpc-sdk7.14-gfx1151
   ```

**os=rhel os=rocky-linux os=oracle-linux**
   ```bash
   sudo dnf install amdrocm-hpc7.14-gfx1151 amdrocm-hpc-sdk7.14-gfx1151
   ```

**os=sles**
   ```bash
   sudo zypper install amdrocm-hpc7.14-gfx1151 amdrocm-hpc-sdk7.14-gfx1151
   ```

**gfx=gfx1150**

1. Use the following command to install HPC SDK for your `gfx1150` GPU:

**os=ubuntu os=debian**
   ```bash
   sudo apt install amdrocm-hpc7.14-gfx1150 amdrocm-hpc-sdk7.14-gfx1150
   ```

**os=rhel os=rocky-linux os=oracle-linux**
   ```bash
   sudo dnf install amdrocm-hpc7.14-gfx1150 amdrocm-hpc-sdk7.14-gfx1150
   ```

**os=sles**
   ```bash
   sudo zypper install amdrocm-hpc7.14-gfx1150 amdrocm-hpc-sdk7.14-gfx1150
   ```

**gfx=gfx1152**

1. Use the following command to install HPC SDK for your `gfx1152` GPU:

**os=ubuntu os=debian**
   ```bash
   sudo apt install amdrocm-hpc7.14-gfx1152 amdrocm-hpc-sdk7.14-gfx1152
   ```

**os=rhel os=rocky-linux os=oracle-linux**
   ```bash
   sudo dnf install amdrocm-hpc7.14-gfx1152 amdrocm-hpc-sdk7.14-gfx1152
   ```

**os=sles**
   ```bash
   sudo zypper install amdrocm-hpc7.14-gfx1152 amdrocm-hpc-sdk7.14-gfx1152
   ```

**gfx=gfx1153**

1. Use the following command to install HPC SDK for your `gfx1153` GPU:

**os=ubuntu os=debian**
   ```bash
   sudo apt install amdrocm-hpc7.14-gfx1153 amdrocm-hpc-sdk7.14-gfx1153
   ```

**os=rhel os=rocky-linux os=oracle-linux**
   ```bash
   sudo dnf install amdrocm-hpc7.14-gfx1153 amdrocm-hpc-sdk7.14-gfx1153
   ```

**os=sles**
   ```bash
   sudo zypper install amdrocm-hpc7.14-gfx1153 amdrocm-hpc-sdk7.14-gfx1153
   ```

## Uninstall HPC SDK

**i=tar**

The standard ROCm uninstallation process can be followed to uninstall HPC SDK. No additional
steps are required to remove the HPC SDK separately. Refer to [Uninstalling](https://rocm.docs.amd.com/en/latest/install/rocm.html#rocm-uninstall)
section and select Tarball from the installation environment selector.

**i=pkgman**

**fam=all**

Use the following command to uninstall HPC SDK for all GPU architectures:

**os=ubuntu os=debian**

```bash
sudo apt autoremove amdrocm-hpc7.14 amdrocm-hpc-sdk7.14
```

**os=rhel os=rocky-linux os=oracle-linux**

```bash
sudo dnf remove amdrocm-hpc7.14 amdrocm-hpc-sdk7.14
```

**os=sles**

```bash
sudo zypper remove amdrocm-hpc7.14 amdrocm-hpc-sdk7.14
```

**gfx=gfx950**

Use the following command to uninstall HPC SDK for your `gfx950` GPU:

**os=ubuntu os=debian**

```bash
sudo apt autoremove amdrocm-hpc7.14-gfx950 amdrocm-hpc-sdk7.14-gfx950
```

**os=rhel os=rocky-linux os=oracle-linux**

```bash
sudo dnf remove amdrocm-hpc7.14-gfx950 amdrocm-hpc-sdk7.14-gfx950
```

**os=sles**

```bash
sudo zypper remove amdrocm-hpc7.14-gfx950 amdrocm-hpc-sdk7.14-gfx950
```

**gfx=gfx942**

Use the following command to uninstall HPC SDK for your `gfx942` GPU:

**os=ubuntu os=debian**

```bash
sudo apt autoremove amdrocm-hpc7.14-gfx942 amdrocm-hpc-sdk7.14-gfx942
```

**os=rhel os=rocky-linux os=oracle-linux**

```bash
sudo dnf remove amdrocm-hpc7.14-gfx942 amdrocm-hpc-sdk7.14-gfx942
```

**os=sles**

```bash
sudo zypper remove amdrocm-hpc7.14-gfx942 amdrocm-hpc-sdk7.14-gfx942
```

**gfx=gfx90a**

Use the following command to uninstall HPC SDK for your `gfx90a` GPU:

**os=ubuntu os=debian**

```bash
sudo apt autoremove amdrocm-hpc7.14-gfx90a amdrocm-hpc-sdk7.14-gfx90a
```

**os=rhel os=rocky-linux os=oracle-linux**

```bash
sudo dnf remove amdrocm-hpc7.14-gfx90a amdrocm-hpc-sdk7.14-gfx90a
```

**os=sles**

```bash
sudo zypper remove amdrocm-hpc7.14-gfx90a amdrocm-hpc-sdk7.14-gfx90a
```

**gfx=gfx908**

Use the following command to uninstall HPC SDK for your `gfx908` GPU:

**os=ubuntu os=debian**

```bash
sudo apt autoremove amdrocm-hpc7.14-gfx908 amdrocm-hpc-sdk7.14-gfx908
```

**os=rhel os=rocky-linux os=oracle-linux**

```bash
sudo dnf remove amdrocm-hpc7.14-gfx908 amdrocm-hpc-sdk7.14-gfx908
```

**os=sles**

```bash
sudo zypper remove amdrocm-hpc7.14-gfx908 amdrocm-hpc-sdk7.14-gfx908
```

**gfx=gfx1200**

Use the following command to uninstall HPC SDK for your `gfx1200` GPU:

**os=ubuntu os=debian**

```bash
sudo apt autoremove amdrocm-hpc7.14-gfx1200 amdrocm-hpc-sdk7.14-gfx1200
```

**os=rhel os=rocky-linux os=oracle-linux**

```bash
sudo dnf remove amdrocm-hpc7.14-gfx1200 amdrocm-hpc-sdk7.14-gfx1200
```

**os=sles**

```bash
sudo zypper remove amdrocm-hpc7.14-gfx1200 amdrocm-hpc-sdk7.14-gfx1200
```

**gfx=gfx1201**

Use the following command to uninstall HPC SDK for your `gfx1201` GPU:

**os=ubuntu os=debian**

```bash
sudo apt autoremove amdrocm-hpc7.14-gfx1201 amdrocm-hpc-sdk7.14-gfx1201
```

**os=rhel os=rocky-linux os=oracle-linux**

```bash
sudo dnf remove amdrocm-hpc7.14-gfx1201 amdrocm-hpc-sdk7.14-gfx1201
```

**os=sles**

```bash
sudo zypper remove amdrocm-hpc7.14-gfx1201 amdrocm-hpc-sdk7.14-gfx1201
```

**gfx=gfx1100**

Use the following command to uninstall HPC SDK for your `gfx1100` GPU:

**os=ubuntu os=debian**

```bash
sudo apt autoremove amdrocm-hpc7.14-gfx1100 amdrocm-hpc-sdk7.14-gfx1100
```

**os=rhel os=rocky-linux os=oracle-linux**

```bash
sudo dnf remove amdrocm-hpc7.14-gfx1100 amdrocm-hpc-sdk7.14-gfx1100
```

**os=sles**

```bash
sudo zypper remove amdrocm-hpc7.14-gfx1100 amdrocm-hpc-sdk7.14-gfx1100
```

**gfx=gfx1101**

Use the following command to uninstall HPC SDK for your `gfx1101` GPU:

**os=ubuntu os=debian**

```bash
sudo apt autoremove amdrocm-hpc7.14-gfx1101 amdrocm-hpc-sdk7.14-gfx1101
```

**os=rhel os=rocky-linux os=oracle-linux**

```bash
sudo dnf remove amdrocm-hpc7.14-gfx1101 amdrocm-hpc-sdk7.14-gfx1101
```

**os=sles**

```bash
sudo zypper remove amdrocm-hpc7.14-gfx1101 amdrocm-hpc-sdk7.14-gfx1101
```

**gfx=gfx1102**

Use the following command to uninstall HPC SDK for your `gfx1102` GPU:

**os=ubuntu os=debian**

```bash
sudo apt autoremove amdrocm-hpc7.14-gfx1102 amdrocm-hpc-sdk7.14-gfx1102
```

**os=rhel os=rocky-linux os=oracle-linux**

```bash
sudo dnf remove amdrocm-hpc7.14-gfx1102 amdrocm-hpc-sdk7.14-gfx1102
```

**os=sles**

```bash
sudo zypper remove amdrocm-hpc7.14-gfx1102 amdrocm-hpc-sdk7.14-gfx1102
```

**gfx=gfx1103**

Use the following command to uninstall HPC SDK for your `gfx1103` GPU:

**os=ubuntu os=debian**

```bash
sudo apt autoremove amdrocm-hpc7.14-gfx1103 amdrocm-hpc-sdk7.14-gfx1103
```

**os=rhel os=rocky-linux os=oracle-linux**

```bash
sudo dnf remove amdrocm-hpc7.14-gfx1103 amdrocm-hpc-sdk7.14-gfx1103
```

**os=sles**

```bash
sudo zypper remove amdrocm-hpc7.14-gfx1103 amdrocm-hpc-sdk7.14-gfx1103
```

**gfx=gfx1030**

Use the following command to uninstall HPC SDK for your `gfx1030` GPU:

**os=ubuntu os=debian**

```bash
sudo apt autoremove amdrocm-hpc7.14-gfx1030 amdrocm-hpc-sdk7.14-gfx1030
```

**os=rhel os=rocky-linux os=oracle-linux**

```bash
sudo dnf remove amdrocm-hpc7.14-gfx1030 amdrocm-hpc-sdk7.14-gfx1030
```

**os=sles**

```bash
sudo zypper remove amdrocm-hpc7.14-gfx1030 amdrocm-hpc-sdk7.14-gfx1030
```

**gfx=gfx1151**

Use the following command to uninstall HPC SDK for your `gfx1151` GPU:

**os=ubuntu os=debian**

```bash
sudo apt autoremove amdrocm-hpc7.14-gfx1151 amdrocm-hpc-sdk7.14-gfx1151
```

**os=rhel os=rocky-linux os=oracle-linux**

```bash
sudo dnf remove amdrocm-hpc7.14-gfx1151 amdrocm-hpc-sdk7.14-gfx1151
```

**os=sles**

```bash
sudo zypper remove amdrocm-hpc7.14-gfx1151 amdrocm-hpc-sdk7.14-gfx1151
```

**gfx=gfx1150**

Use the following command to uninstall HPC SDK for your `gfx1150` GPU:

**os=ubuntu os=debian**

```bash
sudo apt autoremove amdrocm-hpc7.14-gfx1150 amdrocm-hpc-sdk7.14-gfx1150
```

**os=rhel os=rocky-linux os=oracle-linux**

```bash
sudo dnf remove amdrocm-hpc7.14-gfx1150 amdrocm-hpc-sdk7.14-gfx1150
```

**os=sles**

```bash
sudo zypper remove amdrocm-hpc7.14-gfx1150 amdrocm-hpc-sdk7.14-gfx1150
```

**gfx=gfx1152**

Use the following command to uninstall HPC SDK for your `gfx1152` GPU:

**os=ubuntu os=debian**

```bash
sudo apt autoremove amdrocm-hpc7.14-gfx1152 amdrocm-hpc-sdk7.14-gfx1152
```

**os=rhel os=rocky-linux os=oracle-linux**

```bash
sudo dnf remove amdrocm-hpc7.14-gfx1152 amdrocm-hpc-sdk7.14-gfx1152
```

**os=sles**

```bash
sudo zypper remove amdrocm-hpc7.14-gfx1152 amdrocm-hpc-sdk7.14-gfx1152
```

**gfx=gfx1153**

Use the following command to uninstall HPC SDK for your `gfx1153` GPU:

**os=ubuntu os=debian**

```bash
sudo apt autoremove amdrocm-hpc7.14-gfx1153 amdrocm-hpc-sdk7.14-gfx1153
```

**os=rhel os=rocky-linux os=oracle-linux**

```bash
sudo dnf remove amdrocm-hpc7.14-gfx1153 amdrocm-hpc-sdk7.14-gfx1153
```

**os=sles**

```bash
sudo zypper remove amdrocm-hpc7.14-gfx1153 amdrocm-hpc-sdk7.14-gfx1153
```

---

Source: https://rocm.docs.amd.com/en/latest/components/extras.html

# ROCm Extra components

* [ROCm Validation Suite](https://rocm.docs.amd.com/projects/ROCmValidationSuite/en/docs-1.5.122/index.html) (RVS) – A suite of tests for validating ROCm installations
 and AMD GPU hardware. Includes tests for GPU functionality, memory, power behavior, and
 peer-to-peer communication, helping diagnose installation issues and hardware faults.
 * [TransferBench](https://rocm.docs.amd.com/projects/TransferBench/en/latest/index.html) – A utility for benchmarking simultaneous memory transfers between user-specified devices (CPUs, GPUs, and NICs). This component is part of the ROCmValidationSuite (RVS) and is installed with it.
* More coming soon.

---

Source: https://rocm.docs.amd.com/en/latest/ai-ecosystem.html

# AMD ROCm AI ecosystem

The [AMD ROCm AI ecosystem documentation
portal](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/index.html)
includes guides covering deep framework installation and setup, large-scale
model training, LLM and diffusion inference serving, and AI workload
performance optimization on AMD GPUs. The ROCm AI ecosystem lives on top of the
[ROCm Core SDK](https://rocm.docs.amd.com/en/latest/about/what-is-rocm.html), which provides the underlying GPU
runtimes (HIP), compilers, and math libraries.

Install PyTorch and JAX on AMD GPUs. Includes hardware-specific instructions
for AMD Instinct and Radeon GPUs and Ryzen APUs across Linux and Windows using
pip.

- [Install PyTorch](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/frameworks/pytorch/install.html)
- [Install JAX](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/frameworks/jax/install.html)

Scale model training across multiple AMD GPUs using PyTorch distributed primitives
(DDP, RPC, collective communication) for large models that exceed single-GPU memory.

- [Scale model training](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/training/scale-model-training.html)

Serve LLMs and generative AI models using high-performance inference frameworks.
Covers single-node and distributed multi-GPU deployments.

- [vLLM](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/inference/vllm.html)
- [SGLang](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/inference/sglang.html)
- [MIGraphX](https://rocm.docs.amd.com/projects/AMDMIGraphX)
- [ONNX Runtime](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/inference/onnxruntime.html)
- [xDiT](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/inference/xdit.html)
- [ComfyUI](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/inference/comfy.html)

Multi-node prefill-decode disaggregated serving over RDMA networking
using MoRI (Modular RDMA Interface) on MI355X clusters.

- [vLLM with MoRI recipe](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/inference/distributed/vllm-mori-recipe.html)
- [SGLang with MoRI recipe](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/inference/distributed/sglang-mori-recipe.html)

Improve throughput, latency, and memory efficiency for AI workloads on AMD Instinct GPUs.

- [Workload optimization](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/optimization/workload-optimization.html)
- [vLLM V1 performance](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/optimization/vllm-v1-optimization.html)
- [Model quantization](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/optimization/model-quantization.html)
- [Model acceleration libraries](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/optimization/model-acceleration-libs.html)
- [Triton kernels](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/optimization/optimize-triton-kernels.html)
- [Composable Kernel](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/optimization/optimize-with-composable-kernel.html)

Hands-on guides and recipes for building AI applications on AMD hardware.

- [AI Playbooks](https://developer.amd.com/playbooks)
- [AI Developer Hub](https://rocm.docs.amd.com/projects/ai-developer-hub)

---

Source: https://rocm.docs.amd.com/en/latest/reference/gpu-specs.html

# AMD GPU specifications

The following tables provide an overview of the hardware specifications for AMD Instinct™ GPUs, AMD Radeon™ PRO and Radeon GPUs, and AMD Ryzen™ APUs.

For more information about ROCm hardware compatibility, see the ROCm [Compatibility matrix](https://rocm.docs.amd.com/en/latest/compatibility/compatibility-matrix.html).

#### NOTE
* [AMD GPU architectures](https://rocm.docs.amd.com/en/latest/reference/gpu-arch/index.html) – microarchitecture
  details, ISA references, and white papers for each GPU generation.
* [AMD GPU system optimization](https://rocm.docs.amd.com/en/latest/reference/system-optimization/index.html) – system
  setup and tuning guides for AMD Instinct, Radeon, and Ryzen hardware.
* [Data types and precision support](https://rocm.docs.amd.com/en/latest/reference/precision-support.html) – supported
  floating-point and integer data types across GPU architectures.
* [AMD GPU atomics operation support](https://rocm.docs.amd.com/en/latest/reference/gpu-atomics-operation.html) –
  atomics operation support by GPU architecture and memory scope.

**AMD Instinct GPUs**

| Name        | Architecture   | LLVM target name   |   VRAM (GiB) | Compute Units     |   Wavefront Size |   LDS (KiB) |   L3 Cache (MiB) | L2 Cache (MiB)   |   L1 Vector Cache (KiB) | L1 Scalar Cache (KiB)   | L1 Instruction Cache (KiB)   | VGPR File (KiB)          |   SGPR File (KiB) |   GFXIP Major version |   GFXIP Minor version |
|-------------|----------------|--------------------|--------------|-------------------|------------------|-------------|------------------|------------------|-------------------------|-------------------------|------------------------------|--------------------------|-------------------|-----------------------|-----------------------|
| MI355X      | CDNA4          | gfx950             |          288 | 256 (32 per XCD)  |               64 |         160 |              256 | 32 (4 per XCD)   |                      32 | 16 per 2 CUs            | 64 per 2 CUs                 | 512                      |              12.5 |                     9 |                     5 |
| MI350X      | CDNA4          | gfx950             |          288 | 256 (32 per XCD)  |               64 |         160 |              256 | 32 (4 per XCD)   |                      32 | 16 per 2 CUs            | 64 per 2 CUs                 | 512                      |              12.5 |                     9 |                     5 |
| MI325X      | CDNA3          | gfx942             |          256 | 304 (38 per XCD)  |               64 |          64 |              256 | 32 (4 per XCD)   |                      32 | 16 per 2 CUs            | 64 per 2 CUs                 | 512                      |              12.5 |                     9 |                     4 |
| MI300X      | CDNA3          | gfx942             |          192 | 304 (38 per XCD)  |               64 |          64 |              256 | 32 (4 per XCD)   |                      32 | 16 per 2 CUs            | 64 per 2 CUs                 | 512                      |              12.5 |                     9 |                     4 |
| MI300A      | CDNA3          | gfx942             |          128 | 228 (38 per XCD)  |               64 |          64 |              256 | 24 (4 per XCD)   |                      32 | 16 per 2 CUs            | 64 per 2 CUs                 | 512                      |              12.5 |                     9 |                     4 |
| MI250X      | CDNA2          | gfx90a             |          128 | 220 (110 per GCD) |               64 |          64 |                  | 16 (8 per GCD)   |                      16 | 16 per 2 CUs            | 32 per 2 CUs                 | 512                      |              12.5 |                     9 |                     0 |
| MI250       | CDNA2          | gfx90a             |          128 | 208 (104 per GCD) |               64 |          64 |                  | 16 (8 per GCD)   |                      16 | 16 per 2 CUs            | 32 per 2 CUs                 | 512                      |              12.5 |                     9 |                     0 |
| MI210       | CDNA2          | gfx90a             |           64 | 104               |               64 |          64 |                  | 8                |                      16 | 16 per 2 CUs            | 32 per 2 CUs                 | 512                      |              12.5 |                     9 |                     0 |
| MI100       | CDNA           | gfx908             |           32 | 120               |               64 |          64 |                  | 8                |                      16 | 16 per 3 CUs            | 32 per 3 CUs                 | 256 VGPR and 256 AccVGPR |              12.5 |                     9 |                     0 |
| MI60        | GCN5.1         | gfx906             |           32 | 64                |               64 |          64 |                  | 4                |                      16 | 16 per 3 CUs            | 32 per 3 CUs                 | 256                      |              12.5 |                     9 |                     0 |
| MI50 (32GB) | GCN5.1         | gfx906             |           32 | 60                |               64 |          64 |                  | 4                |                      16 | 16 per 3 CUs            | 32 per 3 CUs                 | 256                      |              12.5 |                     9 |                     0 |
| MI50 (16GB) | GCN5.1         | gfx906             |           16 | 60                |               64 |          64 |                  | 4                |                      16 | 16 per 3 CUs            | 32 per 3 CUs                 | 256                      |              12.5 |                     9 |                     0 |
| MI25        | GCN5.0         | gfx900             |           16 | 64                |               64 |          64 |                  | 4                |                      16 | 16 per 3 CUs            | 32 per 3 CUs                 | 256                      |              12.5 |                     9 |                     0 |
| MI8         | GCN3.0         | gfx803             |            4 | 64                |               64 |          64 |                  | 2                |                      16 | 16 per 4 CUs            | 32 per 4 CUs                 | 256                      |              12.5 |                     8 |                     0 |
| MI6         | GCN4.0         | gfx803             |           16 | 36                |               64 |          64 |                  | 2                |                      16 | 16 per 4 CUs            | 32 per 4 CUs                 | 256                      |              12.5 |                     8 |                     0 |

**AMD Radeon PRO GPUs**

| Name                       | Architecture   | LLVM target name   |   VRAM (GiB) |   Compute Units | Wavefront Size   |   LDS (KiB) |   Infinity Cache (MiB) |   L2 Cache (MiB) | Graphics L1 Cache (KiB)   |   L0 Vector Cache (KiB) |   L0 Scalar Cache (KiB) |   L0 Instruction Cache (KiB) |   VGPR File (KiB) |   SGPR File (KiB) |   GFXIP Major version |   GFXIP Minor version |
|----------------------------|----------------|--------------------|--------------|-----------------|------------------|-------------|------------------------|------------------|---------------------------|-------------------------|-------------------------|------------------------------|-------------------|-------------------|-----------------------|-----------------------|
| Radeon AI PRO R9700S       | RDNA4          | gfx1201            |           32 |              64 | 32 or 64         |         128 |                     64 |                8 | N/A                       |                      32 |                      16 |                           32 |               768 |                32 |                    12 |                     0 |
| Radeon AI PRO R9700        | RDNA4          | gfx1201            |           32 |              64 | 32 or 64         |         128 |                     64 |                8 | N/A                       |                      32 |                      16 |                           32 |               768 |                32 |                    12 |                     0 |
| Radeon AI PRO R9600D       | RDNA4          | gfx1201            |           32 |              48 | 32 or 64         |         128 |                     48 |                8 | N/A                       |                      32 |                      16 |                           32 |               768 |                32 |                    12 |                     0 |
| Radeon PRO V710            | RDNA3          | gfx1101            |           28 |              54 | 32 or 64         |         128 |                     56 |                4 | 256                       |                      32 |                      16 |                           32 |               768 |                32 |                    11 |                     0 |
| Radeon PRO W7900 Dual Slot | RDNA3          | gfx1100            |           48 |              96 | 32 or 64         |         128 |                     96 |                6 | 256                       |                      32 |                      16 |                           32 |               768 |                32 |                    11 |                     0 |
| Radeon PRO W7900           | RDNA3          | gfx1100            |           48 |              96 | 32 or 64         |         128 |                     96 |                6 | 256                       |                      32 |                      16 |                           32 |               768 |                32 |                    11 |                     0 |
| Radeon PRO W7800 48GB      | RDNA3          | gfx1100            |           48 |              70 | 32 or 64         |         128 |                     96 |                6 | 256                       |                      32 |                      16 |                           32 |               768 |                32 |                    11 |                     0 |
| Radeon PRO W7800           | RDNA3          | gfx1100            |           32 |              70 | 32 or 64         |         128 |                     64 |                6 | 256                       |                      32 |                      16 |                           32 |               768 |                32 |                    11 |                     0 |
| Radeon PRO W7700           | RDNA3          | gfx1101            |           16 |              48 | 32 or 64         |         128 |                     64 |                4 | 256                       |                      32 |                      16 |                           32 |               768 |                32 |                    11 |                     0 |

**AMD Radeon GPUs**

| Name                 | Architecture   | LLVM target name   |   VRAM (GiB) |   Compute Units | Wavefront Size   |   LDS (KiB) |   Infinity Cache (MiB) |   L2 Cache (MiB) | Graphics L1 Cache (KiB)   |   L0 Vector Cache (KiB) |   L0 Scalar Cache (KiB) |   L0 Instruction Cache (KiB) |   VGPR File (KiB) |   SGPR File (KiB) |   GFXIP Major version |   GFXIP Minor version |
|----------------------|----------------|--------------------|--------------|-----------------|------------------|-------------|------------------------|------------------|---------------------------|-------------------------|-------------------------|------------------------------|-------------------|-------------------|-----------------------|-----------------------|
| Radeon RX 9070 XT    | RDNA4          | gfx1201            |           16 |              64 | 32 or 64         |         128 |                     64 |                8 | N/A                       |                      32 |                      16 |                           32 |               768 |                32 |                    12 |                     0 |
| Radeon RX 9070 GRE   | RDNA4          | gfx1201            |           16 |              48 | 32 or 64         |         128 |                     48 |                6 | N/A                       |                      32 |                      16 |                           32 |               768 |                32 |                    12 |                     0 |
| Radeon RX 9070       | RDNA4          | gfx1201            |           16 |              56 | 32 or 64         |         128 |                     64 |                8 | N/A                       |                      32 |                      16 |                           32 |               768 |                32 |                    12 |                     0 |
| Radeon RX 9060 XT LP | RDNA4          | gfx1200            |           16 |              32 | 32 or 64         |         128 |                     32 |                4 | N/A                       |                      32 |                      16 |                           32 |               768 |                32 |                    12 |                     0 |
| Radeon RX 9060 XT    | RDNA4          | gfx1200            |           16 |              32 | 32 or 64         |         128 |                     32 |                4 | N/A                       |                      32 |                      16 |                           32 |               768 |                32 |                    12 |                     0 |
| Radeon RX 9060       | RDNA4          | gfx1200            |            8 |              28 | 32 or 64         |         128 |                     32 |                4 | N/A                       |                      32 |                      16 |                           32 |               768 |                32 |                    12 |                     0 |
| Radeon RX 7900 XTX   | RDNA3          | gfx1100            |           24 |              96 | 32 or 64         |         128 |                     96 |                6 | 256                       |                      32 |                      16 |                           32 |               768 |                32 |                    11 |                     0 |
| Radeon RX 7900 XT    | RDNA3          | gfx1100            |           20 |              84 | 32 or 64         |         128 |                     80 |                6 | 256                       |                      32 |                      16 |                           32 |               768 |                32 |                    11 |                     0 |
| Radeon RX 7900 GRE   | RDNA3          | gfx1100            |           16 |              80 | 32 or 64         |         128 |                     64 |                6 | 256                       |                      32 |                      16 |                           32 |               768 |                32 |                    11 |                     0 |
| Radeon RX 7800 XT    | RDNA3          | gfx1101            |           16 |              60 | 32 or 64         |         128 |                     64 |                4 | 256                       |                      32 |                      16 |                           32 |               768 |                32 |                    11 |                     0 |
| Radeon RX 7700       | RDNA3          | gfx1101            |           16 |              40 | 32 or 64         |         128 |                     64 |                4 | 256                       |                      32 |                      16 |                           32 |               768 |                32 |                    11 |                     0 |
| Radeon RX 7700 XT    | RDNA3          | gfx1101            |           12 |              54 | 32 or 64         |         128 |                     48 |                4 | 256                       |                      32 |                      16 |                           32 |               768 |                32 |                    11 |                     0 |
| Radeon RX 7600       | RDNA3          | gfx1102            |            8 |              32 | 32 or 64         |         128 |                     32 |                2 | 256                       |                      32 |                      16 |                           32 |               512 |                32 |                    11 |                     0 |

**AMD Ryzen APUs**

| Name                      | Graphics model   | Architecture   | LLVM target name   | VRAM (GiB)         |   Compute Units | Wavefront Size   |   LDS (KiB) | Infinity Cache (MiB)   |   L2 Cache (MiB) |   Graphics L1 Cache (KiB) |   L0 Vector Cache (KiB) |   L0 Scalar Cache (KiB) |   L0 Instruction Cache (KiB) |   VGPR File (KiB) |   SGPR File (KiB) |   GFXIP Major version |   GFXIP Minor version |
|---------------------------|------------------|----------------|--------------------|--------------------|-----------------|------------------|-------------|------------------------|------------------|---------------------------|-------------------------|-------------------------|------------------------------|-------------------|-------------------|-----------------------|-----------------------|
| AMD Ryzen 7 7840U         | Radeon 780M      | RDNA3          | gfx1103            | Dynamic + carveout |              12 | 32 or 64         |         128 | N/A                    |                2 |                       256 |                      32 |                      16 |                           32 |               512 |                32 |                    11 |                     0 |
| AMD Ryzen 9 270           | Radeon 780M      | RDNA3          | gfx1103            | Dynamic + carveout |              12 | 32 or 64         |         128 | N/A                    |                2 |                       256 |                      32 |                      16 |                           32 |               512 |                32 |                    11 |                     0 |
| AMD Ryzen AI 9 HX 375     | Radeon 890M      | RDNA3.5        | gfx1150            | Dynamic + carveout |              16 | 32 or 64         |         128 | N/A                    |                2 |                       256 |                      32 |                      16 |                           32 |               512 |                32 |                    11 |                     5 |
| AMD Ryzen AI Max+ PRO 395 | Radeon 8060S     | RDNA3.5        | gfx1151            | Dynamic + carveout |              40 | 32 or 64         |         128 | 32                     |                2 |                       256 |                      32 |                      16 |                           32 |               768 |                32 |                    11 |                     5 |
| AMD Ryzen AI 7 350        | Radeon 860M      | RDNA3.5        | gfx1152            | Dynamic + carveout |               8 | 32 or 64         |         128 | N/A                    |                1 |                       256 |                      32 |                      16 |                           32 |               512 |                32 |                    11 |                     5 |

---

Source: https://rocm.docs.amd.com/en/latest/reference/gpu-arch/index.html

# AMD GPU architecture documentation

This page collects AMD GPU architecture documentation in one place,
including instruction set architecture (ISA) references, white papers,
microarchitecture overviews, and performance counter documentation for
AMD Instinct™ and Radeon™ GPU families.

These documents can be useful when studying hardware capabilities, tuning kernel
performance, interpreting profiling data, or understanding low-level
architectural behavior.

**CDNA™ 4** (AMD Instinct MI350 Series)

* [AMD CDNA 4 (Instinct MI350 Series) ISA reference](https://www.amd.com/content/dam/amd/en/documents/instinct-tech-docs/instruction-set-architectures/amd-instinct-cdna4-instruction-set-architecture.pdf)
* [White paper](https://www.amd.com/content/dam/amd/en/documents/instinct-tech-docs/white-papers/amd-cdna-4-architecture-whitepaper.pdf)
* [Performance counters](https://rocm.docs.amd.com/en/latest/reference/gpu-arch/mi350-performance-counters.html)

**CDNA 3** (AMD Instinct MI300 Series)

* [AMD CDNA 3 (Instinct MI300 Series) ISA reference](https://www.amd.com/content/dam/amd/en/documents/instinct-tech-docs/instruction-set-architectures/amd-instinct-mi300-cdna3-instruction-set-architecture.pdf)
* [White paper](https://www.amd.com/content/dam/amd/en/documents/instinct-tech-docs/white-papers/amd-cdna-3-white-paper.pdf)
* [AMD Instinct MI300 microarchitecture](https://rocm.docs.amd.com/en/latest/reference/gpu-arch/mi300.html)
* [Performance counters](https://rocm.docs.amd.com/en/latest/reference/gpu-arch/mi300-mi200-performance-counters.html)

**CDNA 2** (AMD Instinct MI200 Series)

* [AMD CDNA 2 (Instinct MI200 Series) ISA reference](https://www.amd.com/system/files/TechDocs/instinct-mi200-cdna2-instruction-set-architecture.pdf)
* [White paper](https://www.amd.com/content/dam/amd/en/documents/instinct-business-docs/white-papers/amd-cdna2-white-paper.pdf)
* [AMD Instinct MI250 microarchitecture](https://rocm.docs.amd.com/en/latest/reference/gpu-arch/mi250.html)
* [Performance counters](https://rocm.docs.amd.com/en/latest/reference/gpu-arch/mi300-mi200-performance-counters.html)

**CDNA** (Instinct MI100)

* [AMD CDNA (AMD Instinct MI100) ISA reference](https://www.amd.com/system/files/TechDocs/instinct-mi100-cdna1-shader-instruction-set-architecture%C2%A0.pdf)
* [AMD Instinct MI100 microarchitecture](https://rocm.docs.amd.com/en/latest/reference/gpu-arch/mi100.html)
* [White paper](https://www.amd.com/content/dam/amd/en/documents/instinct-business-docs/white-papers/amd-cdna-white-paper.pdf)

**RDNA™** (Radeon and Ryzen)

* [AMD RDNA 4 ISA reference](https://www.amd.com/content/dam/amd/en/documents/radeon-tech-docs/instruction-set-architectures/rdna4-instruction-set-architecture.pdf)
* [AMD RDNA 3 ISA reference](https://www.amd.com/system/files/TechDocs/rdna3-shader-instruction-set-architecture-feb-2023_0.pdf)
* [AMD RDNA 2 ISA reference](https://www.amd.com/system/files/TechDocs/rdna2-shader-instruction-set-architecture.pdf)
* [AMD RDNA ISA reference](https://www.amd.com/system/files/TechDocs/rdna-shader-instruction-set-architecture.pdf)

**Older architectures**

* [AMD Vega 7nm (Instinct MI50) ISA reference](https://www.amd.com/system/files/TechDocs/vega-7nm-shader-instruction-set-architecture.pdf)
* [AMD Vega (Instinct MI25) ISA reference](https://www.amd.com/system/files/TechDocs/vega-shader-instruction-set-architecture.pdf)
* [AMD GCN 3 ISA reference](https://www.amd.com/system/files/TechDocs/gcn3-instruction-set-architecture.pdf)

---

Source: https://rocm.docs.amd.com/en/latest/reference/gpu-arch/mi350.html

# AMD Instinct MI350 Series microarchitecture

This section covers the AMD Instinct MI350 Series GPU architecture based on AMD
CDNA4. It includes the following topics:

* [AMD CDNA 4 (Instinct MI350 Series) ISA reference](https://www.amd.com/content/dam/amd/en/documents/instinct-tech-docs/instruction-set-architectures/amd-instinct-cdna4-instruction-set-architecture.pdf) –
  Instruction set architecture reference for the AMD Instinct MI350 Series.
* [White paper](https://www.amd.com/content/dam/amd/en/documents/instinct-tech-docs/white-papers/amd-cdna-4-architecture-whitepaper.pdf) –
  Architectural overview of the AMD CDNA 4 design.
* [MI350 Series performance counters](https://rocm.docs.amd.com/en/latest/reference/gpu-arch/mi350-performance-counters.html) –
  Hardware performance counters and derived metrics available for profiling
  MI350 Series GPUs using ROCprofiler-SDK and ROCm Compute Profiler.

---

Source: https://rocm.docs.amd.com/en/latest/reference/gpu-arch/mi350-performance-counters.html

# MI350 Series performance counters

This topic lists and describes the hardware performance counters and derived metrics available on the AMD Instinct MI350 and MI355 GPUs. These counters are available for profiling using [ROCprofiler-SDK](https://rocm.docs.amd.com/projects/rocprofiler-sdk/en/latest/index.html) and [ROCm Compute Profiler](https://rocm.docs.amd.com/projects/rocprofiler-compute/en/latest/).

The following sections list the performance counters based on the IP blocks.

## Command processor packet processor counters (CPC)

| Hardware counter              | Definition                                                                                   |
|-------------------------------|----------------------------------------------------------------------------------------------|
| CPC_ALWAYS_COUNT              | Always count.                                                                                |
| CPC_ADC_VALID_CHUNK_NOT_AVAIL | ADC valid chunk is not available when dispatch walking is in progress in the multi-xcc mode. |
| CPC_ADC_DISPATCH_ALLOC_DONE   | ADC dispatch allocation is done.                                                             |
| CPC_ADC_VALID_CHUNK_END       | ADC crawler’s valid chunk end in the multi-xcc mode.                                         |
| CPC_SYNC_FIFO_FULL_LEVEL      | SYNC FIFO full last cycles.                                                                  |
| CPC_SYNC_FIFO_FULL            | SYNC FIFO full times.                                                                        |
| CPC_GD_BUSY                   | ADC busy.                                                                                    |
| CPC_TG_SEND                   | ADC thread group send.                                                                       |
| CPC_WALK_NEXT_CHUNK           | ADC walking next valid chunk in the multi-xcc mode.                                          |
| CPC_STALLED_BY_SE0_SPI        | ADC CSDATA stalled by SE0SPI.                                                                |
| CPC_STALLED_BY_SE1_SPI        | ADC CSDATA stalled by SE1SPI.                                                                |
| CPC_STALLED_BY_SE2_SPI        | ADC CSDATA stalled by SE2SPI.                                                                |
| CPC_STALLED_BY_SE3_SPI        | ADC CSDATA stalled by SE3SPI.                                                                |
| CPC_LTE_ALL                   | CPC sync counter LteAll. Only Master XCD manages LteAll.                                     |
| CPC_SYNC_WRREQ_FIFO_BUSY      | CPC sync counter request FIFO is not empty.                                                  |
| CPC_CANE_BUSY                 | CPC CANE bus is busy, which indicates the presence of inflight sync counter requests.        |
| CPC_CANE_STALL                | CPC sync counter sending is stalled by CANE.                                                 |

## Shader pipe interpolators (SPI) counters

| Hardware counter         | Definition                                                     |
|--------------------------|----------------------------------------------------------------|
| SPI_CS0_WINDOW_VALID     | Clock count enabled by PIPE0 perfcounter_start event.          |
| SPI_CS0_BUSY             | Number of clocks with outstanding waves for PIPE0 (SPI or SH). |
| SPI_CS0_NUM_THREADGROUPS | Number of thread groups launched for PIPE0.                    |
| SPI_CS0_CRAWLER_STALL    | Number of clocks when PIPE0 event or wave order FIFO is full.  |
| SPI_CS0_EVENT_WAVE       | Number of PIPE0 events and waves.                              |
| SPI_CS0_WAVE             | Number of PIPE0 waves.                                         |
| SPI_CS1_WINDOW_VALID     | Clock count enabled by PIPE1 perfcounter_start event.          |
| SPI_CS1_BUSY             | Number of clocks with outstanding waves for PIPE1 (SPI or SH). |
| SPI_CS1_NUM_THREADGROUPS | Number of thread groups launched for PIPE1.                    |
| SPI_CS1_CRAWLER_STALL    | Number of clocks when PIPE1 event or wave order FIFO is full.  |
| SPI_CS1_EVENT_WAVE       | Number of PIPE1 events and waves.                              |
| SPI_CS1_WAVE             | Number of PIPE1 waves.                                         |
| SPI_CS2_WINDOW_VALID     | Clock count enabled by PIPE2 perfcounter_start event.          |
| SPI_CS2_BUSY             | Number of clocks with outstanding waves for PIPE2 (SPI or SH). |
| SPI_CS2_NUM_THREADGROUPS | Number of thread groups launched for PIPE2.                    |
| SPI_CS2_CRAWLER_STALL    | Number of clocks when PIPE2 event or wave order FIFO is full.  |
| SPI_CS2_EVENT_WAVE       | Number of PIPE2 events and waves.                              |
| SPI_CS2_WAVE             | Number of PIPE2 waves.                                         |
| SPI_CS3_WINDOW_VALID     | Clock count enabled by PIPE3 perfcounter_start event.          |
| SPI_CS3_BUSY             | Number of clocks with outstanding waves for PIPE3 (SPI or SH). |
| SPI_CS3_NUM_THREADGROUPS | Number of thread groups launched for PIPE3.                    |
| SPI_CS3_CRAWLER_STALL    | Number of clocks when PIPE3 event or wave order FIFO is full.  |
| SPI_CS3_EVENT_WAVE       | Number of PIPE3 events and waves.                              |
| SPI_CS3_WAVE             | Number of PIPE3 waves.                                         |
| SPI_CSQ_P0_Q0_OCCUPANCY  | Sum of occupancy info for PIPE0 Queue0.                        |
| SPI_CSQ_P0_Q1_OCCUPANCY  | Sum of occupancy info for PIPE0 Queue1.                        |
| SPI_CSQ_P0_Q2_OCCUPANCY  | Sum of occupancy info for PIPE0 Queue2.                        |
| SPI_CSQ_P0_Q3_OCCUPANCY  | Sum of occupancy info for PIPE0 Queue3.                        |
| SPI_CSQ_P0_Q4_OCCUPANCY  | Sum of occupancy info for PIPE0 Queue4.                        |
| SPI_CSQ_P0_Q5_OCCUPANCY  | Sum of occupancy info for PIPE0 Queue5.                        |
| SPI_CSQ_P0_Q6_OCCUPANCY  | Sum of occupancy info for PIPE0 Queue6.                        |
| SPI_CSQ_P0_Q7_OCCUPANCY  | Sum of occupancy info for PIPE0 Queue7.                        |
| SPI_CSQ_P1_Q0_OCCUPANCY  | Sum of occupancy info for PIPE1 Queue0.                        |
| SPI_CSQ_P1_Q1_OCCUPANCY  | Sum of occupancy info for PIPE1 Queue1.                        |
| SPI_CSQ_P1_Q2_OCCUPANCY  | Sum of occupancy info for PIPE1 Queue2.                        |
| SPI_CSQ_P1_Q3_OCCUPANCY  | Sum of occupancy info for PIPE1 Queue3.                        |
| SPI_CSQ_P1_Q4_OCCUPANCY  | Sum of occupancy info for PIPE1 Queue4.                        |
| SPI_CSQ_P1_Q5_OCCUPANCY  | Sum of occupancy info for PIPE1 Queue5.                        |
| SPI_CSQ_P1_Q6_OCCUPANCY  | Sum of occupancy info for PIPE1 Queue6.                        |
| SPI_CSQ_P1_Q7_OCCUPANCY  | Sum of occupancy info for PIPE1 Queue7.                        |
| SPI_CSQ_P2_Q0_OCCUPANCY  | Sum of occupancy info for PIPE2 Queue0.                        |
| SPI_CSQ_P2_Q1_OCCUPANCY  | Sum of occupancy info for PIPE2 Queue1.                        |
| SPI_CSQ_P2_Q2_OCCUPANCY  | Sum of occupancy info for PIPE2 Queue2.                        |
| SPI_CSQ_P2_Q3_OCCUPANCY  | Sum of occupancy info for PIPE2 Queue3.                        |
| SPI_CSQ_P2_Q4_OCCUPANCY  | Sum of occupancy info for PIPE2 Queue4.                        |
| SPI_CSQ_P2_Q5_OCCUPANCY  | Sum of occupancy info for PIPE2 Queue5.                        |
| SPI_CSQ_P2_Q6_OCCUPANCY  | Sum of occupancy info for PIPE2 Queue6.                        |
| SPI_CSQ_P2_Q7_OCCUPANCY  | Sum of occupancy info for PIPE2 Queue7.                        |
| SPI_CSQ_P3_Q0_OCCUPANCY  | Sum of occupancy info for PIPE3 Queue0.                        |
| SPI_CSQ_P3_Q1_OCCUPANCY  | Sum of occupancy info for PIPE3 Queue1.                        |
| SPI_CSQ_P3_Q2_OCCUPANCY  | Sum of occupancy info for PIPE3 Queue2.                        |
| SPI_CSQ_P3_Q3_OCCUPANCY  | Sum of occupancy info for PIPE3 Queue3.                        |
| SPI_CSQ_P3_Q4_OCCUPANCY  | Sum of occupancy info for PIPE3 Queue4.                        |
| SPI_CSQ_P3_Q5_OCCUPANCY  | Sum of occupancy info for PIPE3 Queue5.                        |
| SPI_CSQ_P3_Q6_OCCUPANCY  | Sum of occupancy info for PIPE3 Queue6.                        |
| SPI_CSQ_P3_Q7_OCCUPANCY  | Sum of occupancy info for PIPE3 Queue7.                        |
| SPI_CSQ_P0_OCCUPANCY     | Sum of occupancy info for all PIPE0 queues.                    |
| SPI_CSQ_P1_OCCUPANCY     | Sum of occupancy info for all PIPE1 queues.                    |
| SPI_CSQ_P2_OCCUPANCY     | Sum of occupancy info for all PIPE2 queues.                    |
| SPI_CSQ_P3_OCCUPANCY     | Sum of occupancy info for all PIPE3 queues.                    |
| SPI_VWC0_VDATA_VALID_WR  | Number of clocks VGPR bus_0 writes VGPRs.                      |
| SPI_VWC1_VDATA_VALID_WR  | Number of clocks VGPR bus_1 writes VGPRs.                      |
| SPI_CSC_WAVE_CNT_BUSY    | Number of cycles when there is any wave in the pipe.           |

## Compute unit (SQ) counters

| Hardware counter               | Definition                                                                                                                             |
|--------------------------------|----------------------------------------------------------------------------------------------------------------------------------------|
| SQ_INSTS_VALU_MFMA_F6F4        | Number of VALU V_MFMA_\*_F6F4 instructions.                                                                                            |
| SQ_INSTS_VALU_MFMA_MOPS_F6F4   | Number of VALU matrix with the performed math operations (add or mul) divided by 512, assuming a full EXEC mask of F6 or F4 data type. |
| SQ_ACTIVE_INST_VALU2           | Number of quad-cycles when two VALU instructions are issued (per-simd, nondeterministic).                                              |
| SQ_INSTS_LDS_LOAD              | Number of LDS load instructions issued (per-simd, emulated).                                                                           |
| SQ_INSTS_LDS_STORE             | Number of LDS store instructions issued (per-simd, emulated).                                                                          |
| SQ_INSTS_LDS_ATOMIC            | Number of LDS atomic instructions issued (per-simd, emulated).                                                                         |
| SQ_INSTS_LDS_LOAD_BANDWIDTH    | Total number of 64-bytes loaded (instrSize \* CountOnes(EXEC))/64 (per-simd, emulated).                                                |
| SQ_INSTS_LDS_STORE_BANDWIDTH   | Total number of 64-bytes written (instrSize \* CountOnes(EXEC))/64 (per-simd, emulated).                                               |
| SQ_INSTS_LDS_ATOMIC_BANDWIDTH  | Total number of 64-bytes atomic (instrSize \* CountOnes(EXEC))/64 (per-simd, emulated).                                                |
| SQ_INSTS_VALU_FLOPS_FP16       | Counts FLOPS per instruction on float 16 excluding MFMA/SMFMA.                                                                         |
| SQ_INSTS_VALU_FLOPS_FP32       | Counts FLOPS per instruction on float 32 excluding MFMA/SMFMA.                                                                         |
| SQ_INSTS_VALU_FLOPS_FP64       | Counts FLOPS per instruction on float 64 excluding MFMA/SMFMA.                                                                         |
| SQ_INSTS_VALU_FLOPS_FP16_TRANS | Counts FLOPS per instruction on float 16 trans excluding MFMA/SMFMA.                                                                   |
| SQ_INSTS_VALU_FLOPS_FP32_TRANS | Counts FLOPS per instruction on float 32 trans excluding MFMA/SMFMA.                                                                   |
| SQ_INSTS_VALU_FLOPS_FP64_TRANS | Counts FLOPS per instruction on float 64 trans excluding MFMA/SMFMA.                                                                   |
| SQ_INSTS_VALU_IOPS             | Counts OPS per instruction on integer or unsigned or bit data (per-simd, emulated).                                                    |
| SQ_LDS_DATA_FIFO_FULL          | Number of cycles LDS data FIFO is full (nondeterministic, unwindowed).                                                                 |
| SQ_LDS_CMD_FIFO_FULL           | Number of cycles LDS command FIFO is full (nondeterministic, unwindowed).                                                              |
| SQ_VMEM_TA_ADDR_FIFO_FULL      | Number of cycles texture requests are stalled due to full address FIFO in TA (nondeterministic, unwindowed).                           |
| SQ_VMEM_TA_CMD_FIFO_FULL       | Number of cycles texture requests are stalled due to full cmd FIFO in TA (nondeterministic, unwindowed).                               |
| SQ_VMEM_WR_TA_DATA_FIFO_FULL   | Number of cycles texture writes are stalled due to full data FIFO in TA (nondeterministic, unwindowed).                                |
| SQC_ICACHE_MISSES_DUPLICATE    | Number of duplicate misses (access to a non-resident, miss pending CL) (per-SQ, per-Bank, nondeterministic).                           |
| SQC_DCACHE_MISSES_DUPLICATE    | Number of duplicate misses (access to a non-resident, miss pending CL) (per-SQ, per-Bank, nondeterministic).                           |

## Texture addressing (TA) unit counters

| Hardware counter              | Definition                                                           |
|-------------------------------|----------------------------------------------------------------------|
| TA_BUFFER_READ_LDS_WAVEFRONTS | Number of buffer read wavefronts for LDS return processed by the TA. |
| TA_FLAT_READ_LDS_WAVEFRONTS   | Number of flat opcode reads for LDS return processed by the TA.      |

## Texture data (TD) unit counters

| Hardware counter        | Definition                                                 |
|-------------------------|------------------------------------------------------------|
| TD_WRITE_ACKT_WAVEFRONT | Number of write acknowledgments, sent to SQ and not to SP. |
| TD_TD_SP_TRAFFIC        | Number of times this TD sends data to the SP.              |

## Texture cache per pipe (TCP) counters

| Hardware counter                         | Definition                                                                                                                                                                                                                                                                                 |
|------------------------------------------|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| TCP_TCP_TA_ADDR_STALL_CYCLES             | TCP stalls TA addr interface.                                                                                                                                                                                                                                                              |
| TCP_TCP_TA_DATA_STALL_CYCLES             | TCP stalls TA data interface. Now windowed.                                                                                                                                                                                                                                                |
| TCP_LFIFO_STALL_CYCLES                   | Memory latency FIFOs full stall.                                                                                                                                                                                                                                                           |
| TCP_RFIFO_STALL_CYCLES                   | Memory Request FIFOs full stall.                                                                                                                                                                                                                                                           |
| TCP_TCR_RDRET_STALL                      | Write into cache stalled by read return from TCR.                                                                                                                                                                                                                                          |
| TCP_PENDING_STALL_CYCLES                 | Stall due to data pending from L2.                                                                                                                                                                                                                                                         |
| TCP_UTCL1_SERIALIZATION_STALL            | Total number of stalls caused due to serializing translation requests through the UTCL1.                                                                                                                                                                                                   |
| TCP_UTCL1_THRASHING_STALL                | Stall caused by thrashing feature in any probe. Lacks accuracy when the stall signal overlaps between probe0 and probe1, which is worse with MECO of thrashing deadlock. Some probe0 events could miss being counted in with MECO on. This perf count provides a rough thrashing estimate. |
| TCP_UTCL1_TRANSLATION_MISS_UNDER_MISS    | Translation miss_under_miss.                                                                                                                                                                                                                                                               |
| TCP_UTCL1_STALL_INFLIGHT_MAX             | Total UTCL1 stalls due to inflight counter saturation.                                                                                                                                                                                                                                     |
| TCP_UTCL1_STALL_LRU_INFLIGHT             | Total UTCL1 stalls due to LRU cache line with inflight traffic.                                                                                                                                                                                                                            |
| TCP_UTCL1_STALL_MULTI_MISS               | Total UTCL1 stalls due to arbitrated multiple misses.                                                                                                                                                                                                                                      |
| TCP_UTCL1_LFIFO_FULL                     | Total UTCL1 and UTCL2 latency, which hides FIFO full cycles.                                                                                                                                                                                                                               |
| TCP_UTCL1_STALL_LFIFO_NOT_RES            | Total UTCL1 stalls due to UTCL2 latency, which hides FIFO output (not resident).                                                                                                                                                                                                           |
| TCP_UTCL1_STALL_UTCL2_REQ_OUT_OF_CREDITS | Total UTCL1 stalls due to UTCL2_req being out of credits.                                                                                                                                                                                                                                  |
| TCP_CLIENT_UTCL1_INFLIGHT                | The sum of inflight client to UTCL1 requests per cycle.                                                                                                                                                                                                                                    |
| TCP_TAGRAM0_REQ                          | Total L2 requests mapping to TagRAM 0 from this TCP to all TCCs.                                                                                                                                                                                                                           |
| TCP_TAGRAM1_REQ                          | Total L2 requests mapping to TagRAM 1 from this TCP to all TCCs.                                                                                                                                                                                                                           |
| TCP_TAGRAM2_REQ                          | Total L2 requests mapping to TagRAM 2 from this TCP to all TCCs.                                                                                                                                                                                                                           |
| TCP_TAGRAM3_REQ                          | Total L2 requests mapping to TagRAM 3 from this TCP to all TCCs.                                                                                                                                                                                                                           |
| TCP_TCP_LATENCY                          | Total TCP wave latency (from the first clock of wave entering to the first clock of wave leaving). Divide by TA_TCP_STATE_READ to find average wave latency.                                                                                                                               |
| TCP_TCC_READ_REQ_LATENCY                 | Total TCP to TCC request latency for reads and atomics with return. Not Windowed.                                                                                                                                                                                                          |
| TCP_TCC_WRITE_REQ_LATENCY                | Total TCP to TCC request latency for writes and atomics without return. Not Windowed.                                                                                                                                                                                                      |
| TCP_TCC_WRITE_REQ_HOLE_LATENCY           | Total TCP req to TCC hole latency for writes and atomics. Not Windowed.                                                                                                                                                                                                                    |

## Texture cache per channel (TCC) counters

| Hardware counter              | Definition                                                                                                                  |
|-------------------------------|-----------------------------------------------------------------------------------------------------------------------------|
| TCC_READ_SECTORS              | Total number of 32B data sectors in read requests.                                                                          |
| TCC_WRITE_SECTORS             | Total number of 32B data sectors in write requests.                                                                         |
| TCC_ATOMIC_SECTORS            | Total number of 32B data sectors in atomic requests.                                                                        |
| TCC_BYPASS_REQ                | Number of bypass requests. This is measured at the tag block.                                                               |
| TCC_LATENCY_FIFO_FULL         | Number of cycles when the latency FIFO is full.                                                                             |
| TCC_SRC_FIFO_FULL             | Number of cycles when the SRC FIFO is assumed to be full as measured at the IB block.                                       |
| TCC_EA0_RDREQ_64B             | Number of 64-byte TCC/EA read requests.                                                                                     |
| TCC_EA0_RDREQ_128B            | Number of 128-byte TCC/EA read requests.                                                                                    |
| TCC_IB_REQ                    | Number of requests through the IB. This measures the number of raw requests from graphics clients to this TCC.              |
| TCC_IB_STALL                  | Number of cycles when the IB output is stalled.                                                                             |
| TCC_EA0_WRREQ_WRITE_DRAM      | Number of TCC/EA write requests (32-byte or 64-byte) destined for DRAM (MC).                                                |
| TCC_EA0_WRREQ_ATOMIC_DRAM     | Number of TCC/EA atomic requests (32-byte or 64-byte) destined for DRAM (MC).                                               |
| TCC_EA0_RDREQ_DRAM_32B        | Number of 32-byte TCC/EA read requests due to DRAM traffic. One 64-byte request is counted as two and one 128-byte as four. |
| TCC_EA0_RDREQ_GMI_32B         | Number of 32-byte TCC/EA read requests due to GMI traffic. One 64-byte request is counted as two and one 128-byte as four.  |
| TCC_EA0_RDREQ_IO_32B          | Number of 32-byte TCC/EA read requests due to IO traffic. One 64-byte request is counted as two and one 128-byte as four.   |
| TCC_EA0_WRREQ_WRITE_DRAM_32B  | Number of 32-byte TCC/EA write requests due to DRAM traffic. One 64-byte request is counted as two.                         |
| TCC_EA0_WRREQ_ATOMIC_DRAM_32B | Number of 32-byte TCC/EA atomic requests due to DRAM traffic. One 64-byte request is counted as two.                        |
| TCC_EA0_WRREQ_WRITE_GMI_32B   | Number of 32-byte TCC/EA write requests due to GMI traffic. One 64-byte request is counted as two.                          |
| TCC_EA0_WRREQ_ATOMIC_GMI_32B  | Number of 32-byte TCC/EA atomic requests due to GMI traffic. One 64-byte request is counted as two.                         |
| TCC_EA0_WRREQ_WRITE_IO_32B    | Number of 32-byte TCC/EA write requests due to IO traffic. One 64-byte request is counted as two.                           |
| TCC_EA0_WRREQ_ATOMIC_IO_32B   | Number of 32-byte TCC/EA atomic requests due to IO traffic. One 64-byte request is counted as two.                          |

---

Source: https://rocm.docs.amd.com/en/latest/reference/gpu-arch/mi300.html

# AMD Instinct™ MI300 Series microarchitecture

The AMD Instinct MI300 Series GPUs are based on the AMD CDNA 3
architecture which was designed to deliver leadership performance for HPC, artificial intelligence (AI), and machine
learning (ML) workloads. The AMD Instinct MI300 Series GPUs are well-suited for extreme scalability and compute performance, running
on everything from individual servers to the world’s largest exascale supercomputers.

With the MI300 Series, AMD is introducing the Accelerator Complex Die (XCD), which contains the
GPU computational elements of the processor along with the lower levels of the cache hierarchy.

The following image depicts the structure of a single XCD in the AMD Instinct MI300 GPU Series.

![image](images/shared/xcd-sys-arch.png)

On the XCD, four Asynchronous Compute Engines (ACEs) send compute shader workgroups to the
Compute Units (CUs). The XCD has 40 CUs: 38 active CUs at the aggregate level and 2 disabled CUs for
yield management. The CUs all share a 4 MB L2 cache that serves to coalesce all memory traffic for the
die. With less than half of the CUs of the AMD Instinct MI200 Series compute die, the AMD CDNA™ 3
XCD die is a smaller building block. However, it uses more advanced packaging and the processor
can include 6 or 8 XCDs for up to 304 CUs, roughly 40% more than MI250X.

The MI300 Series integrate up to 8 vertically stacked XCDs, 8 stacks of
High-Bandwidth Memory 3 (HBM3) and 4 I/O dies (containing system
infrastructure) using the AMD Infinity Fabric™ technology as interconnect.

The Matrix Cores inside the CDNA 3 CUs have significant improvements, emphasizing AI and machine
learning, enhancing throughput of existing data types while adding support for new data types.
CDNA 2 Matrix Cores support FP16 and BF16, while offering INT8 for inference. Compared to MI250X
GPUs, CDNA 3 Matrix Cores triple the performance for FP16 and BF16, while providing a
performance gain of 6.8 times for INT8. FP8 has a performance gain of 16 times compared to FP32,
while TF32 has a gain of 4 times compared to FP32.

#### Peak-performance capabilities of the MI300X for different data types.

| Computation and Data Type   |   FLOPS/CLOCK/CU |   Peak TFLOPS |
|-----------------------------|------------------|---------------|
| Matrix FP64                 |              256 |         163.4 |
| Vector FP64                 |              128 |          81.7 |
| Matrix FP32                 |              256 |         163.4 |
| Vector FP32                 |              256 |         163.4 |
| Vector TF32                 |             1024 |         653.7 |
| Matrix FP16                 |             2048 |        1307.4 |
| Matrix BF16                 |             2048 |        1307.4 |
| Matrix FP8                  |             4096 |        2614.9 |
| Matrix INT8                 |             4096 |        2614.9 |

The above table summarizes the aggregated peak performance of the AMD Instinct MI300X Open
Compute Platform (OCP) Open Accelerator Modules (OAMs) for different data types and command
processors. The middle column lists the peak performance (number of data elements processed in a
single instruction) of a single compute unit if a SIMD (or matrix) instruction is submitted in each clock
cycle. The third column lists the theoretical peak performance of the OAM. The theoretical aggregated
peak memory bandwidth of the GPU is 5.3 TB per second.

The following image shows the block diagram of the APU (left) and the OAM package (right) both
connected via AMD Infinity Fabric™ network on-chip.

![](reference/gpu-arch/images/image008.png)

## Node-level architecture

![image](images/shared/mi300-node-level-arch.png)

The image above shows the node-level architecture of a system with AMD EPYC processors in a
dual-socket configuration and eight AMD Instinct MI300X GPUs. The MI300X OAMs attach to the
host system via PCIe Gen 5 x16 links (yellow lines). The GPUs are using seven high-bandwidth,
low-latency AMD Infinity Fabric™ links (red lines) to form a fully connected 8-GPU system.

---

Source: https://rocm.docs.amd.com/en/latest/reference/gpu-arch/mi300-mi200-performance-counters.html

# MI300 and MI200 Series performance counters and metrics

This document lists and describes the hardware performance counters and derived metrics available
for the AMD Instinct™ MI300 and MI200 GPU. You can also access this information using the
[ROCprofiler-SDK](https://rocm.docs.amd.com/projects/rocprofiler-sdk/en/latest/how-to/using-rocprofv3.html).

## MI300 and MI200 Series performance counters

Series performance counters include the following categories:

* [Command processor counters](#command-processor-counters)
* [Graphics register bus manager counters](#graphics-register-bus-manager-counters)
* [Shader processor input counters](#spi-counters)
* [Compute unit counters](#compute-unit-counters)
* [L1 instruction cache (L1i) and scalar L1 data cache (L1d) counters](#l1i-and-sl1d-cache-counters)
* [Vector L1 cache subsystem counters](#vector-l1-cache-subsystem-counters)
* [L2 cache access counters](#l2-cache-access-counters)

The following sections provide additional details for each category.

#### NOTE
Preliminary validation of all MI300 and MI200 Series performance counters is in progress. Those with
an asterisk (\*) require further evaluation.

### Command processor counters

Command processor counters are further classified into command processor-fetcher and command
processor-compute.

#### Command processor-fetcher counters

| Hardware counter                     | Unit   | Definition                                                                                                        |
|--------------------------------------|--------|-------------------------------------------------------------------------------------------------------------------|
| `CPF_CMP_UTCL1_STALL_ON_TRANSLATION` | Cycles | Number of cycles one of the compute unified translation caches (L1) is stalled waiting on translation             |
| `CPF_CPF_STAT_BUSY`                  | Cycles | Number of cycles command processor-fetcher is busy                                                                |
| `CPF_CPF_STAT_IDLE`                  | Cycles | Number of cycles command processor-fetcher is idle                                                                |
| `CPF_CPF_STAT_STALL`                 | Cycles | Number of cycles command processor-fetcher is stalled                                                             |
| `CPF_CPF_TCIU_BUSY`                  | Cycles | Number of cycles command processor-fetcher texture cache interface unit interface is busy                         |
| `CPF_CPF_TCIU_IDLE`                  | Cycles | Number of cycles command processor-fetcher texture cache interface unit interface is idle                         |
| `CPF_CPF_TCIU_STALL`                 | Cycles | Number of cycles command processor-fetcher texture cache interface unit interface is stalled waiting on free tags |

The texture cache interface unit is the interface between the command processor and the memory
system.

#### Command processor-compute counters

| Hardware counter                 | Unit   | Definition                                                                                     |
|----------------------------------|--------|------------------------------------------------------------------------------------------------|
| `CPC_ME1_BUSY_FOR_PACKET_DECODE` | Cycles | Number of cycles command processor-compute micro engine is busy decoding packets               |
| `CPC_UTCL1_STALL_ON_TRANSLATION` | Cycles | Number of cycles one of the unified translation caches (L1) is stalled waiting on translation  |
| `CPC_CPC_STAT_BUSY`              | Cycles | Number of cycles command processor-compute is busy                                             |
| `CPC_CPC_STAT_IDLE`              | Cycles | Number of cycles command processor-compute is idle                                             |
| `CPC_CPC_STAT_STALL`             | Cycles | Number of cycles command processor-compute is stalled                                          |
| `CPC_CPC_TCIU_BUSY`              | Cycles | Number of cycles command processor-compute texture cache interface unit interface is busy      |
| `CPC_CPC_TCIU_IDLE`              | Cycles | Number of cycles command processor-compute texture cache interface unit interface is idle      |
| `CPC_CPC_UTCL2IU_BUSY`           | Cycles | Number of cycles command processor-compute unified translation cache (L2) interface is busy    |
| `CPC_CPC_UTCL2IU_IDLE`           | Cycles | Number of cycles command processor-compute unified translation cache (L2) interface is idle    |
| `CPC_CPC_UTCL2IU_STALL`          | Cycles | Number of cycles command processor-compute unified translation cache (L2) interface is stalled |
| `CPC_ME1_DC0_SPI_BUSY`           | Cycles | Number of cycles command processor-compute micro engine processor is busy                      |

The micro engine runs packet-processing firmware on the command processor-compute counter.

### Graphics register bus manager counters

| Hardware counter   | Unit   | Definition                                                                        |
|--------------------|--------|-----------------------------------------------------------------------------------|
| `GRBM_COUNT`       | Cycles | Number of free-running GPU cycles                                                 |
| `GRBM_GUI_ACTIVE`  | Cycles | Number of GPU active cycles                                                       |
| `GRBM_CP_BUSY`     | Cycles | Number of cycles any of the command processor blocks are busy                     |
| `GRBM_SPI_BUSY`    | Cycles | Number of cycles any of the shader processor input is busy in the shader engines  |
| `GRBM_TA_BUSY`     | Cycles | Number of cycles any of the texture addressing unit is busy in the shader engines |
| `GRBM_TC_BUSY`     | Cycles | Number of cycles any of the texture cache blocks are busy                         |
| `GRBM_CPC_BUSY`    | Cycles | Number of cycles the command processor-compute is busy                            |
| `GRBM_CPF_BUSY`    | Cycles | Number of cycles the command processor-fetcher is busy                            |
| `GRBM_UTCL2_BUSY`  | Cycles | Number of cycles the unified translation cache (Level 2 [L2]) block is busy       |
| `GRBM_EA_BUSY`     | Cycles | Number of cycles the efficiency arbiter block is busy                             |

Texture cache blocks include:

* Texture cache arbiter
* Texture cache per pipe, also known as vector Level 1 (L1) cache
* Texture cache per channel, also known as known as L2 cache
* Texture cache interface

### Shader processor input counters

| Hardware counter | Unit | Definition |
|-----------------------------|-------------|--------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| `SPI_CSN_BUSY` | Cycles | Number of cycles with outstanding waves |
| `SPI_CSN_WINDOW_VALID` | Cycles | Number of cycles enabled by `perfcounter_start` event |
| `SPI_CSN_NUM_THREADGROUPS` | Workgroups | Number of dispatched workgroups |
| `SPI_CSN_WAVE` | Wavefronts | Number of dispatched wavefronts |
| `SPI_RA_REQ_NO_ALLOC` | Cycles | Number of arbiter cycles with requests but no allocation |
| `SPI_RA_REQ_NO_ALLOC_CSN` | Cycles | Number of arbiter cycles with compute shader (nth pipe) requests but no compute shader (nth pipe) allocation |
| `SPI_RA_RES_STALL_CSN` | Cycles | Number of arbiter stall cycles due to shortage of compute shader (nth pipe) pipeline slots |
| `SPI_RA_TMP_STALL_CSN` | Cycles | Number of stall cycles due to shortage of temp space |
| `SPI_RA_WAVE_SIMD_FULL_CSN` | SIMD-cycles | Accumulated number of single instruction, multiple data (SIMD) per cycle affected by shortage of wave slots for compute shader (nth pipe) wave dispatch |
| `SPI_RA_VGPR_SIMD_FULL_CSN` | SIMD-cycles | Accumulated number of SIMDs per cycle affected by shortage of vector general-purpose register (VGPR) slots for compute shader (nth pipe) wave dispatch |
| `SPI_RA_SGPR_SIMD_FULL_CSN` | SIMD-cycles | Accumulated number of SIMDs per cycle affected by shortage of scalar general-purpose register (SGPR) slots for compute shader (nth pipe) wave dispatch |
| `SPI_RA_LDS_CU_FULL_CSN` | CU | Number of compute units affected by shortage of local data share (LDS) space for compute shader (nth pipe) wave dispatch |
| `SPI_RA_BAR_CU_FULL_CSN` | CU | Number of compute units with compute shader (nth pipe) waves waiting at a BARRIER |
| `SPI_RA_BULKY_CU_FULL_CSN` | CU | Number of compute units with compute shader (nth pipe) waves waiting for BULKY resource |
| `SPI_RA_TGLIM_CU_FULL_CSN` | Cycles | Number of compute shader (nth pipe) wave stall cycles due to restriction of `tg_limit` for thread group size |
| `SPI_RA_WVLIM_STALL_CSN` | Cycles | Number of cycles compute shader (nth pipe) is stalled due to `WAVE_LIMIT` |
| `SPI_VWC_CSC_WR` | Qcycles | Number of quad-cycles taken to initialize VGPRs when launching waves |
| `SPI_SWC_CSC_WR` | Qcycles | Number of quad-cycles taken to initialize SGPRs when launching waves |

### Compute unit counters

The compute unit counters are further classified into instruction mix, matrix fused multiply-add (FMA)
operation counters, level counters, wavefront counters, wavefront cycle counters, and LDS counters.

#### Instruction mix

| Hardware counter          | Unit   | Definition                                                                                                                           |
|---------------------------|--------|--------------------------------------------------------------------------------------------------------------------------------------|
| `SQ_INSTS`                | Instr  | Number of instructions issued                                                                                                        |
| `SQ_INSTS_VALU`           | Instr  | Number of vector arithmetic logic unit (VALU) instructions including matrix FMA issued                                               |
| `SQ_INSTS_VALU_ADD_F16`   | Instr  | Number of VALU half-precision floating-point (F16) `ADD` or `SUB` instructions issued                                                |
| `SQ_INSTS_VALU_MUL_F16`   | Instr  | Number of VALU F16 Multiply instructions issued                                                                                      |
| `SQ_INSTS_VALU_FMA_F16`   | Instr  | Number of VALU F16 FMA or multiply-add instructions issued                                                                           |
| `SQ_INSTS_VALU_TRANS_F16` | Instr  | Number of VALU F16 Transcendental instructions issued                                                                                |
| `SQ_INSTS_VALU_ADD_F32`   | Instr  | Number of VALU full-precision floating-point (F32) `ADD` or `SUB` instructions issued                                                |
| `SQ_INSTS_VALU_MUL_F32`   | Instr  | Number of VALU F32 Multiply instructions issued                                                                                      |
| `SQ_INSTS_VALU_FMA_F32`   | Instr  | Number of VALU F32 FMAor multiply-add instructions issued                                                                            |
| `SQ_INSTS_VALU_TRANS_F32` | Instr  | Number of VALU F32 Transcendental instructions issued                                                                                |
| `SQ_INSTS_VALU_ADD_F64`   | Instr  | Number of VALU F64 `ADD` or `SUB` instructions issued                                                                                |
| `SQ_INSTS_VALU_MUL_F64`   | Instr  | Number of VALU F64 Multiply instructions issued                                                                                      |
| `SQ_INSTS_VALU_FMA_F64`   | Instr  | Number of VALU F64 FMA or multiply-add instructions issued                                                                           |
| `SQ_INSTS_VALU_TRANS_F64` | Instr  | Number of VALU F64 Transcendental instructions issued                                                                                |
| `SQ_INSTS_VALU_INT32`     | Instr  | Number of VALU 32-bit integer instructions (signed or unsigned) issued                                                               |
| `SQ_INSTS_VALU_INT64`     | Instr  | Number of VALU 64-bit integer instructions (signed or unsigned) issued                                                               |
| `SQ_INSTS_VALU_CVT`       | Instr  | Number of VALU Conversion instructions issued                                                                                        |
| `SQ_INSTS_VALU_MFMA_I8`   | Instr  | Number of 8-bit Integer matrix FMA instructions issued                                                                               |
| `SQ_INSTS_VALU_MFMA_F16`  | Instr  | Number of F16 matrix FMA instructions issued                                                                                         |
| `SQ_INSTS_VALU_MFMA_F32`  | Instr  | Number of F32 matrix FMA instructions issued                                                                                         |
| `SQ_INSTS_VALU_MFMA_F64`  | Instr  | Number of F64 matrix FMA instructions issued                                                                                         |
| `SQ_INSTS_MFMA`           | Instr  | Number of matrix FMA instructions issued                                                                                             |
| `SQ_INSTS_VMEM_WR`        | Instr  | Number of vector memory write instructions (including flat) issued                                                                   |
| `SQ_INSTS_VMEM_RD`        | Instr  | Number of vector memory read instructions (including flat) issued                                                                    |
| `SQ_INSTS_VMEM`           | Instr  | Number of vector memory instructions issued, including both flat and buffer instructions                                             |
| `SQ_INSTS_SALU`           | Instr  | Number of scalar arithmetic logic unit (SALU) instructions issued                                                                    |
| `SQ_INSTS_SMEM`           | Instr  | Number of scalar memory instructions issued                                                                                          |
| `SQ_INSTS_SMEM_NORM`      | Instr  | Number of scalar memory instructions normalized to match `smem_level` issued                                                         |
| `SQ_INSTS_FLAT`           | Instr  | Number of flat instructions issued                                                                                                   |
| `SQ_INSTS_FLAT_LDS_ONLY`  | Instr  | **MI200 Series only** Number of FLAT instructions that read/write only from/to LDS issued. Works only if `EARLY_TA_DONE` is enabled. |
| `SQ_INSTS_LDS`            | Instr  | Number of LDS instructions issued **(MI200: includes flat; MI300: does not include flat)**                                           |
| `SQ_INSTS_GDS`            | Instr  | Number of global data share instructions issued                                                                                      |
| `SQ_INSTS_EXP_GDS`        | Instr  | Number of EXP and global data share instructions excluding skipped export instructions issued                                        |
| `SQ_INSTS_BRANCH`         | Instr  | Number of branch instructions issued                                                                                                 |
| `SQ_INSTS_SENDMSG`        | Instr  | Number of `SENDMSG` instructions including `s_endpgm` issued                                                                         |
| `SQ_INSTS_VSKIPPED`       | Instr  | Number of vector instructions skipped                                                                                                |

Flat instructions allow read, write, and atomic access to a generic memory address pointer that can
resolve to any of the following physical memories:

* Global Memory
* Scratch (“private”)
* LDS (“shared”)
* Invalid - `MEM_VIOL` TrapStatus

#### Matrix fused multiply-add operation counters

| Hardware counter               | Unit   | Definition                                                |
|--------------------------------|--------|-----------------------------------------------------------|
| `SQ_INSTS_VALU_MFMA_MOPS_I8`   | IOP    | Number of 8-bit integer matrix FMA ops in the unit of 512 |
| `SQ_INSTS_VALU_MFMA_MOPS_F16`  | FLOP   | Number of F16 floating matrix FMA ops in the unit of 512  |
| `SQ_INSTS_VALU_MFMA_MOPS_BF16` | FLOP   | Number of BF16 floating matrix FMA ops in the unit of 512 |
| `SQ_INSTS_VALU_MFMA_MOPS_F32`  | FLOP   | Number of F32 floating matrix FMA ops in the unit of 512  |
| `SQ_INSTS_VALU_MFMA_MOPS_F64`  | FLOP   | Number of F64 floating matrix FMA ops in the unit of 512  |

#### Level counters

#### NOTE
All level counters must be followed by `SQ_ACCUM_PREV_HIRES` counter to measure average latency.

| Hardware counter      | Unit   | Definition                                                                             |
|-----------------------|--------|----------------------------------------------------------------------------------------|
| `SQ_ACCUM_PREV`       | Count  | Accumulated counter sample value where accumulation takes place once every four cycles |
| `SQ_ACCUM_PREV_HIRES` | Count  | Accumulated counter sample value where accumulation takes place once every cycle       |
| `SQ_LEVEL_WAVES`      | Waves  | Number of inflight waves                                                               |
| `SQ_INST_LEVEL_VMEM`  | Instr  | Number of inflight vector memory (including flat) instructions                         |
| `SQ_INST_LEVEL_SMEM`  | Instr  | Number of inflight scalar memory instructions                                          |
| `SQ_INST_LEVEL_LDS`   | Instr  | Number of inflight LDS (including flat) instructions                                   |
| `SQ_IFETCH_LEVEL`     | Instr  | Number of inflight instruction fetch requests from the cache                           |

Use the following formulae to calculate latencies:

* Vector memory latency = `SQ_ACCUM_PREV_HIRES` divided by `SQ_INSTS_VMEM`
* Wave latency = `SQ_ACCUM_PREV_HIRES` divided by `SQ_WAVE`
* LDS latency = `SQ_ACCUM_PREV_HIRES` divided by `SQ_INSTS_LDS`
* Scalar memory latency = `SQ_ACCUM_PREV_HIRES` divided by `SQ_INSTS_SMEM_NORM`
* Instruction fetch latency = `SQ_ACCUM_PREV_HIRES` divided by `SQ_IFETCH`

#### Wavefront counters

| Hardware counter    | Unit   | Definition                                                                                |
|---------------------|--------|-------------------------------------------------------------------------------------------|
| `SQ_WAVES`          | Waves  | Number of wavefronts dispatched to sequencers, including both new and restored wavefronts |
| `SQ_WAVES_SAVED`    | Waves  | Number of context-saved waves                                                             |
| `SQ_WAVES_RESTORED` | Waves  | Number of context-restored waves sent to sequencers                                       |
| `SQ_WAVES_EQ_64`    | Waves  | Number of wavefronts with exactly 64 active threads sent to sequencers                    |
| `SQ_WAVES_LT_64`    | Waves  | Number of wavefronts with less than 64 active threads sent to sequencers                  |
| `SQ_WAVES_LT_48`    | Waves  | Number of wavefronts with less than 48 active threads sent to sequencers                  |
| `SQ_WAVES_LT_32`    | Waves  | Number of wavefronts with less than 32 active threads sent to sequencers                  |
| `SQ_WAVES_LT_16`    | Waves  | Number of wavefronts with less than 16 active threads sent to sequencers                  |

#### Wavefront cycle counters

| Hardware counter           | Unit    | Definition                                                                                                      |
|----------------------------|---------|-----------------------------------------------------------------------------------------------------------------|
| `SQ_CYCLES`                | Cycles  | Clock cycles                                                                                                    |
| `SQ_BUSY_CYCLES`           | Cycles  | Number of cycles while sequencers reports it to be busy                                                         |
| `SQ_BUSY_CU_CYCLES`        | Qcycles | Number of quad-cycles each compute unit is busy                                                                 |
| `SQ_VALU_MFMA_BUSY_CYCLES` | Cycles  | Number of cycles the matrix FMA arithmetic logic unit (ALU) is busy                                             |
| `SQ_WAVE_CYCLES`           | Qcycles | Number of quad-cycles spent by waves in the compute units                                                       |
| `SQ_WAIT_ANY`              | Qcycles | Number of quad-cycles spent waiting for anything                                                                |
| `SQ_WAIT_INST_ANY`         | Qcycles | Number of quad-cycles spent waiting for any instruction to be issued                                            |
| `SQ_ACTIVE_INST_ANY`       | Qcycles | Number of quad-cycles spent by each wave to work on an instruction                                              |
| `SQ_ACTIVE_INST_VMEM`      | Qcycles | Number of quad-cycles spent by the sequencer instruction arbiter to work on a vector memory instruction         |
| `SQ_ACTIVE_INST_LDS`       | Qcycles | Number of quad-cycles spent by the sequencer instruction arbiter to work on an LDS instruction                  |
| `SQ_ACTIVE_INST_VALU`      | Qcycles | Number of quad-cycles spent by the sequencer instruction arbiter to work on a VALU instruction                  |
| `SQ_ACTIVE_INST_SCA`       | Qcycles | Number of quad-cycles spent by the sequencer instruction arbiter to work on a SALU or scalar memory instruction |
| `SQ_ACTIVE_INST_EXP_GDS`   | Qcycles | Number of quad-cycles spent by the sequencer instruction arbiter to work on an `EXPORT` or `GDS` instruction    |
| `SQ_ACTIVE_INST_MISC`      | Qcycles | Number of quad-cycles spent by the sequencer instruction arbiter to work on a `BRANCH` or `SENDMSG` instruction |
| `SQ_ACTIVE_INST_FLAT`      | Qcycles | Number of quad-cycles spent by the sequencer instruction arbiter to work on a flat instruction                  |
| `SQ_INST_CYCLES_VMEM_WR`   | Qcycles | Number of quad-cycles spent to send addr and cmd data for vector memory write instructions                      |
| `SQ_INST_CYCLES_VMEM_RD`   | Qcycles | Number of quad-cycles spent to send addr and cmd data for vector memory read instructions                       |
| `SQ_INST_CYCLES_SMEM`      | Qcycles | Number of quad-cycles spent to execute scalar memory reads                                                      |
| `SQ_INST_CYCLES_SALU`      | Qcycles | Number of quad-cycles spent to execute non-memory read scalar operations                                        |
| `SQ_THREAD_CYCLES_VALU`    | Qcycles | Number of quad-cycles spent to execute VALU operations on active threads                                        |
| `SQ_WAIT_INST_LDS`         | Qcycles | Number of quad-cycles spent waiting for LDS instruction to be issued                                            |

`SQ_THREAD_CYCLES_VALU` is similar to `INST_CYCLES_VALU`, but it’s multiplied by the number of
active threads.

#### LDS counters

| Hardware counter         | Unit   | Definition                                                                         |
|--------------------------|--------|------------------------------------------------------------------------------------|
| `SQ_LDS_ATOMIC_RETURN`   | Cycles | Number of atomic return cycles in LDS                                              |
| `SQ_LDS_BANK_CONFLICT`   | Cycles | Number of cycles LDS is stalled by bank conflicts                                  |
| `SQ_LDS_ADDR_CONFLICT`   | Cycles | Number of cycles LDS is stalled by address conflicts                               |
| `SQ_LDS_UNALIGNED_STALL` | Cycles | Number of cycles LDS is stalled processing flat unaligned load or store operations |
| `SQ_LDS_MEM_VIOLATIONS`  | Count  | Number of threads that have a memory violation in the LDS                          |
| `SQ_LDS_IDX_ACTIVE`      | Cycles | Number of cycles LDS is used for indexed operations                                |

#### Miscellaneous counters

| Hardware counter   | Unit    | Definition                                                      |
|--------------------|---------|-----------------------------------------------------------------|
| `SQ_IFETCH`        | Count   | Number of instruction fetch requests from L1i, in 32-byte width |
| `SQ_ITEMS`         | Threads | Number of valid items per wave                                  |

### L1 instruction cache (L1i) and scalar L1 data cache (L1d) counters

| Hardware counter                | Unit   | Definition                                                                                                  |
|---------------------------------|--------|-------------------------------------------------------------------------------------------------------------|
| `SQC_ICACHE_REQ`                | Req    | Number of L1 instruction (L1i) cache requests                                                               |
| `SQC_ICACHE_HITS`               | Count  | Number of L1i cache hits                                                                                    |
| `SQC_ICACHE_MISSES`             | Count  | Number of non-duplicate L1i cache misses including uncached requests                                        |
| `SQC_ICACHE_MISSES_DUPLICATE`   | Count  | Number of duplicate L1i cache misses whose previous lookup miss on the same cache line is not fulfilled yet |
| `SQC_DCACHE_REQ`                | Req    | Number of scalar L1d requests                                                                               |
| `SQC_DCACHE_INPUT_VALID_READYB` | Cycles | Number of cycles while sequencer input is valid but scalar L1d is not ready                                 |
| `SQC_DCACHE_HITS`               | Count  | Number of scalar L1d hits                                                                                   |
| `SQC_DCACHE_MISSES`             | Count  | Number of non-duplicate scalar L1d misses including uncached requests                                       |
| `SQC_DCACHE_MISSES_DUPLICATE`   | Count  | Number of duplicate scalar L1d misses                                                                       |
| `SQC_DCACHE_REQ_READ_1`         | Req    | Number of constant cache read requests in a single 32-bit data word                                         |
| `SQC_DCACHE_REQ_READ_2`         | Req    | Number of constant cache read requests in two 32-bit data words                                             |
| `SQC_DCACHE_REQ_READ_4`         | Req    | Number of constant cache read requests in four 32-bit data words                                            |
| `SQC_DCACHE_REQ_READ_8`         | Req    | Number of constant cache read requests in eight 32-bit data words                                           |
| `SQC_DCACHE_REQ_READ_16`        | Req    | Number of constant cache read requests in 16 32-bit data words                                              |
| `SQC_DCACHE_ATOMIC`             | Req    | Number of atomic requests                                                                                   |
| `SQC_TC_REQ`                    | Req    | Number of texture cache requests that were issued by instruction and constant caches                        |
| `SQC_TC_INST_REQ`               | Req    | Number of instruction requests to the L2 cache                                                              |
| `SQC_TC_DATA_READ_REQ`          | Req    | Number of data Read requests to the L2 cache                                                                |
| `SQC_TC_DATA_WRITE_REQ`         | Req    | Number of data write requests to the L2 cache                                                               |
| `SQC_TC_DATA_ATOMIC_REQ`        | Req    | Number of data atomic requests to the L2 cache                                                              |
| `SQC_TC_STALL`                  | Cycles | Number of cycles while the valid requests to the L2 cache are stalled                                       |

### Vector L1 cache subsystem counters

The vector L1 cache subsystem counters are further classified into texture addressing unit, texture data
unit, vector L1d or texture cache per pipe, and texture cache arbiter counters.

#### Texture addressing unit counters

| Hardware counter                      | Unit   | Definition                                                                            | Value range for `n`   |
|---------------------------------------|--------|---------------------------------------------------------------------------------------|-----------------------|
| `TA_TA_BUSY[n]`                       | Cycles | Texture addressing unit busy cycles                                                   | 0-15                  |
| `TA_TOTAL_WAVEFRONTS[n]`              | Instr  | Number of wavefronts processed by texture addressing unit                             | 0-15                  |
| `TA_BUFFER_WAVEFRONTS[n]`             | Instr  | Number of buffer wavefronts processed by texture addressing unit                      | 0-15                  |
| `TA_BUFFER_READ_WAVEFRONTS[n]`        | Instr  | Number of buffer read wavefronts processed by texture addressing unit                 | 0-15                  |
| `TA_BUFFER_WRITE_WAVEFRONTS[n]`       | Instr  | Number of buffer write wavefronts processed by texture addressing unit                | 0-15                  |
| `TA_BUFFER_ATOMIC_WAVEFRONTS[n]`      | Instr  | Number of buffer atomic wavefronts processed by texture addressing unit               | 0-15                  |
| `TA_BUFFER_TOTAL_CYCLES[n]`           | Cycles | Number of buffer cycles (including read and write) issued to texture cache            | 0-15                  |
| `TA_BUFFER_COALESCED_READ_CYCLES[n]`  | Cycles | Number of coalesced buffer read cycles issued to texture cache                        | 0-15                  |
| `TA_BUFFER_COALESCED_WRITE_CYCLES[n]` | Cycles | Number of coalesced buffer write cycles issued to texture cache                       | 0-15                  |
| `TA_ADDR_STALLED_BY_TC_CYCLES[n]`     | Cycles | Number of cycles texture addressing unit address path is stalled by texture cache     | 0-15                  |
| `TA_DATA_STALLED_BY_TC_CYCLES[n]`     | Cycles | Number of cycles texture addressing unit data path is stalled by texture cache        | 0-15                  |
| `TA_ADDR_STALLED_BY_TD_CYCLES[n]`     | Cycles | Number of cycles texture addressing unit address path is stalled by texture data unit | 0-15                  |
| `TA_FLAT_WAVEFRONTS[n]`               | Instr  | Number of flat opcode wavefronts processed by texture addressing unit                 | 0-15                  |
| `TA_FLAT_READ_WAVEFRONTS[n]`          | Instr  | Number of flat opcode read wavefronts processed by texture addressing unit            | 0-15                  |
| `TA_FLAT_WRITE_WAVEFRONTS[n]`         | Instr  | Number of flat opcode write wavefronts processed by texture addressing unit           | 0-15                  |
| `TA_FLAT_ATOMIC_WAVEFRONTS[n]`        | Instr  | Number of flat opcode atomic wavefronts processed by texture addressing unit          | 0-15                  |

#### Texture data unit counters

| Hardware counter              | Unit   | Definition                                                                   | Value range for `n`   |
|-------------------------------|--------|------------------------------------------------------------------------------|-----------------------|
| `TD_TD_BUSY[n]`               | Cycle  | Texture data unit busy cycles while it is processing or waiting for data     | 0-15                  |
| `TD_TC_STALL[n]`              | Cycle  | Number of cycles texture data unit is stalled waiting for texture cache data | 0-15                  |
| `TD_SPI_STALL[n]`             | Cycle  | Number of cycles texture data unit is stalled by shader processor input      | 0-15                  |
| `TD_LOAD_WAVEFRONT[n]`        | Instr  | Number of wavefront instructions (read, write, atomic)                       | 0-15                  |
| `TD_STORE_WAVEFRONT[n]`       | Instr  | Number of write wavefront instructions                                       | 0-15                  |
| `TD_ATOMIC_WAVEFRONT[n]`      | Instr  | Number of atomic wavefront instructions                                      | 0-15                  |
| `TD_COALESCABLE_WAVEFRONT[n]` | Instr  | Number of coalescable wavefronts according to texture addressing unit        | 0-15                  |

#### Texture cache per pipe counters

| Hardware counter                         | Unit   | Definition                                                                                                       | Value range for `n`   |
|------------------------------------------|--------|------------------------------------------------------------------------------------------------------------------|-----------------------|
| `TCP_GATE_EN1[n]`                        | Cycles | Number of cycles vector L1d interface clocks are turned on                                                       | 0-15                  |
| `TCP_GATE_EN2[n]`                        | Cycles | Number of cycles vector L1d core clocks are turned on                                                            | 0-15                  |
| `TCP_TD_TCP_STALL_CYCLES[n]`             | Cycles | Number of cycles texture data unit stalls vector L1d                                                             | 0-15                  |
| `TCP_TCR_TCP_STALL_CYCLES[n]`            | Cycles | Number of cycles texture cache router stalls vector L1d                                                          | 0-15                  |
| `TCP_READ_TAGCONFLICT_STALL_CYCLES[n]`   | Cycles | Number of cycles tag RAM conflict stalls on a read                                                               | 0-15                  |
| `TCP_WRITE_TAGCONFLICT_STALL_CYCLES[n]`  | Cycles | Number of cycles tag RAM conflict stalls on a write                                                              | 0-15                  |
| `TCP_ATOMIC_TAGCONFLICT_STALL_CYCLES[n]` | Cycles | Number of cycles tag RAM conflict stalls on an atomic                                                            | 0-15                  |
| `TCP_PENDING_STALL_CYCLES[n]`            | Cycles | Number of cycles vector L1d is stalled due to data pending from L2 Cache                                         | 0-15                  |
| `TCP_TCP_TA_DATA_STALL_CYCLES`           | Cycles | Number of cycles texture cache per pipe stalls texture addressing unit data interface                            | NA                    |
| `TCP_TA_TCP_STATE_READ[n]`               | Req    | Number of state reads                                                                                            | 0-15                  |
| `TCP_VOLATILE[n]`                        | Req    | Number of L1 volatile pixels or buffers from texture addressing unit                                             | 0-15                  |
| `TCP_TOTAL_ACCESSES[n]`                  | Req    | Number of vector L1d accesses. Equals `TCP_PERF_SEL_TOTAL_READ`+`TCP_PERF_SEL_TOTAL_NONREAD`                     | 0-15                  |
| `TCP_TOTAL_READ[n]`                      | Req    | Number of vector L1d read accesses                                                                               | 0-15                  |
| `TCP_TOTAL_WRITE[n]`                     | Req    | Number of vector L1d write accesses                                                                              | 0-15                  |
| `TCP_TOTAL_ATOMIC_WITH_RET[n]`           | Req    | Number of vector L1d atomic requests with return                                                                 | 0-15                  |
| `TCP_TOTAL_ATOMIC_WITHOUT_RET[n]`        | Req    | Number of vector L1d atomic without return                                                                       | 0-15                  |
| `TCP_TOTAL_WRITEBACK_INVALIDATES[n]`     | Count  | Total number of vector L1d writebacks and invalidates                                                            | 0-15                  |
| `TCP_UTCL1_REQUEST[n]`                   | Req    | Number of address translation requests to unified translation cache (L1)                                         | 0-15                  |
| `TCP_UTCL1_TRANSLATION_HIT[n]`           | Req    | Number of unified translation cache (L1) translation hits                                                        | 0-15                  |
| `TCP_UTCL1_TRANSLATION_MISS[n]`          | Req    | Number of unified translation cache (L1) translation misses                                                      | 0-15                  |
| `TCP_UTCL1_PERMISSION_MISS[n]`           | Req    | Number of unified translation cache (L1) permission misses                                                       | 0-15                  |
| `TCP_TOTAL_CACHE_ACCESSES[n]`            | Req    | Number of vector L1d cache accesses including hits and misses                                                    | 0-15                  |
| `TCP_TCP_LATENCY[n]`                     | Cycles | **MI200 Series only** Accumulated wave access latency to vL1D over all wavefronts                                | 0-15                  |
| `TCP_TCC_READ_REQ_LATENCY[n]`            | Cycles | **MI200 Series only** Total vL1D to L2 request latency over all wavefronts for reads and atomics with return     | 0-15                  |
| `TCP_TCC_WRITE_REQ_LATENCY[n]`           | Cycles | **MI200 Series only** Total vL1D to L2 request latency over all wavefronts for writes and atomics without return | 0-15                  |
| `TCP_TCC_READ_REQ[n]`                    | Req    | Number of read requests to L2 cache                                                                              | 0-15                  |
| `TCP_TCC_WRITE_REQ[n]`                   | Req    | Number of write requests to L2 cache                                                                             | 0-15                  |
| `TCP_TCC_ATOMIC_WITH_RET_REQ[n]`         | Req    | Number of atomic requests to L2 cache with return                                                                | 0-15                  |
| `TCP_TCC_ATOMIC_WITHOUT_RET_REQ[n]`      | Req    | Number of atomic requests to L2 cache without return                                                             | 0-15                  |
| `TCP_TCC_NC_READ_REQ[n]`                 | Req    | Number of non-coherently cached read requests to L2 cache                                                        | 0-15                  |
| `TCP_TCC_UC_READ_REQ[n]`                 | Req    | Number of uncached read requests to L2 cache                                                                     | 0-15                  |
| `TCP_TCC_CC_READ_REQ[n]`                 | Req    | Number of coherently cached read requests to L2 cache                                                            | 0-15                  |
| `TCP_TCC_RW_READ_REQ[n]`                 | Req    | Number of coherently cached with write read requests to L2 cache                                                 | 0-15                  |
| `TCP_TCC_NC_WRITE_REQ[n]`                | Req    | Number of non-coherently cached write requests to L2 cache                                                       | 0-15                  |
| `TCP_TCC_UC_WRITE_REQ[n]`                | Req    | Number of uncached write requests to L2 cache                                                                    | 0-15                  |
| `TCP_TCC_CC_WRITE_REQ[n]`                | Req    | Number of coherently cached write requests to L2 cache                                                           | 0-15                  |
| `TCP_TCC_RW_WRITE_REQ[n]`                | Req    | Number of coherently cached with write write requests to L2 cache                                                | 0-15                  |
| `TCP_TCC_NC_ATOMIC_REQ[n]`               | Req    | Number of non-coherently cached atomic requests to L2 cache                                                      | 0-15                  |
| `TCP_TCC_UC_ATOMIC_REQ[n]`               | Req    | Number of uncached atomic requests to L2 cache                                                                   | 0-15                  |
| `TCP_TCC_CC_ATOMIC_REQ[n]`               | Req    | Number of coherently cached atomic requests to L2 cache                                                          | 0-15                  |
| `TCP_TCC_RW_ATOMIC_REQ[n]`               | Req    | Number of coherently cached with write atomic requests to L2 cache                                               | 0-15                  |

Note that:

* `TCP_TOTAL_READ[n]` = `TCP_PERF_SEL_TOTAL_HIT_LRU_READ` + `TCP_PERF_SEL_TOTAL_MISS_LRU_READ` + `TCP_PERF_SEL_TOTAL_MISS_EVICT_READ`
* `TCP_TOTAL_WRITE[n]` = `TCP_PERF_SEL_TOTAL_MISS_LRU_WRITE``+ ``TCP_PERF_SEL_TOTAL_MISS_EVICT_WRITE`
* `TCP_TOTAL_WRITEBACK_INVALIDATES[n]` = `TCP_PERF_SEL_TOTAL_WBINVL1``+ ``TCP_PERF_SEL_TOTAL_WBINVL1_VOL``+ ``TCP_PERF_SEL_CP_TCP_INVALIDATE``+ ``TCP_PERF_SEL_SQ_TCP_INVALIDATE_VOL`

#### Texture cache arbiter counters

| Hardware counter   | Unit   | Definition                                                   | Value range for `n`   |
|--------------------|--------|--------------------------------------------------------------|-----------------------|
| `TCA_CYCLE[n]`     | Cycles | Number of texture cache arbiter cycles                       | 0-31                  |
| `TCA_BUSY[n]`      | Cycles | Number of cycles texture cache arbiter has a pending request | 0-31                  |

### L2 cache access counters

L2 cache is also known as texture cache per channel.

**MI300 hardware counter**

| Hardware counter                     | Unit   | Definition                                                                                                                                                               | Value range for `n`   |
|--------------------------------------|--------|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------|-----------------------|
| `TCC_CYCLE[n]`                       | Cycles | Number of L2 cache free-running clocks                                                                                                                                   | 0-31                  |
| `TCC_BUSY[n]`                        | Cycles | Number of L2 cache busy cycles                                                                                                                                           | 0-31                  |
| `TCC_REQ[n]`                         | Req    | Number of L2 cache requests of all types (measured at the tag block)                                                                                                     | 0-31                  |
| `TCC_STREAMING_REQ[n]`               | Req    | Number of L2 cache streaming requests (measured at the tag block)                                                                                                        | 0-31                  |
| `TCC_NC_REQ[n]`                      | Req    | Number of non-coherently cached requests (measured at the tag block)                                                                                                     | 0-31                  |
| `TCC_UC_REQ[n]`                      | Req    | Number of uncached requests. This is measured at the tag block                                                                                                           | 0-31                  |
| `TCC_CC_REQ[n]`                      | Req    | Number of coherently cached requests. This is measured at the tag block                                                                                                  | 0-31                  |
| `TCC_RW_REQ[n]`                      | Req    | Number of coherently cached with write requests. This is measured at the tag block                                                                                       | 0-31                  |
| `TCC_PROBE[n]`                       | Req    | Number of probe requests                                                                                                                                                 | 0-31                  |
| `TCC_PROBE_ALL[n]`                   | Req    | Number of external probe requests with `EA_TCC_preq_all == 1`                                                                                                            | 0-31                  |
| `TCC_READ[n]`                        | Req    | Number of L2 cache read requests (includes compressed reads but not metadata reads)                                                                                      | 0-31                  |
| `TCC_WRITE[n]`                       | Req    | Number of L2 cache write requests                                                                                                                                        | 0-31                  |
| `TCC_ATOMIC[n]`                      | Req    | Number of L2 cache atomic requests of all types                                                                                                                          | 0-31                  |
| `TCC_HIT[n]`                         | Req    | Number of L2 cache hits                                                                                                                                                  | 0-31                  |
| `TCC_MISS[n]`                        | Req    | Number of L2 cache misses                                                                                                                                                | 0-31                  |
| `TCC_WRITEBACK[n]`                   | Req    | Number of lines written back to the main memory, including writebacks of dirty lines and uncached write or atomic requests                                               | 0-31                  |
| `TCC_EA0_WRREQ[n]`                   | Req    | Number of 32-byte and 64-byte transactions going over the `TC_EA_wrreq` interface (doesn’t include probe commands)                                                       | 0-31                  |
| `TCC_EA0_WRREQ_64B[n]`               | Req    | Total number of 64-byte transactions (write or `CMPSWAP`) going over the `TC_EA_wrreq` interface                                                                         | 0-31                  |
| `TCC_EA0_WR_UNCACHED_32B[n]`         | Req    | Number of 32 or 64-byte write or atomic going over the `TC_EA_wrreq` interface due to uncached traffic                                                                   | 0-31                  |
| `TCC_EA0_WRREQ_STALL[n]`             | Cycles | Number of cycles a write request is stalled                                                                                                                              | 0-31                  |
| `TCC_EA0_WRREQ_IO_CREDIT_STALL[n]`   | Cycles | Number of cycles an efficiency arbiter write request is stalled due to the interface running out of input-output (IO) credits                                            | 0-31                  |
| `TCC_EA0_WRREQ_GMI_CREDIT_STALL[n]`  | Cycles | Number of cycles an efficiency arbiter write request is stalled due to the interface running out of GMI credits                                                          | 0-31                  |
| `TCC_EA0_WRREQ_DRAM_CREDIT_STALL[n]` | Cycles | Number of cycles an efficiency arbiter write request is stalled due to the interface running out of DRAM credits                                                         | 0-31                  |
| `TCC_TOO_MANY_EA_WRREQS_STALL[n]`    | Cycles | Number of cycles the L2 cache is unable to send an efficiency arbiter write request due to it reaching its maximum capacity of pending efficiency arbiter write requests | 0-31                  |
| `TCC_EA0_WRREQ_LEVEL[n]`             | Req    | The accumulated number of efficiency arbiter write requests in flight                                                                                                    | 0-31                  |
| `TCC_EA0_ATOMIC[n]`                  | Req    | Number of 32-byte or 64-byte atomic requests going over the `TC_EA_wrreq` interface                                                                                      | 0-31                  |
| `TCC_EA0_ATOMIC_LEVEL[n]`            | Req    | The accumulated number of efficiency arbiter atomic requests in flight                                                                                                   | 0-31                  |
| `TCC_EA0_RDREQ[n]`                   | Req    | Number of 32-byte or 64-byte read requests to efficiency arbiter                                                                                                         | 0-31                  |
| `TCC_EA0_RDREQ_32B[n]`               | Req    | Number of 32-byte read requests to efficiency arbiter                                                                                                                    | 0-31                  |
| `TCC_EA0_RD_UNCACHED_32B[n]`         | Req    | Number of 32-byte efficiency arbiter reads due to uncached traffic. A 64-byte request is counted as 2                                                                    | 0-31                  |
| `TCC_EA0_RDREQ_IO_CREDIT_STALL[n]`   | Cycles | Number of cycles there is a stall due to the read request interface running out of IO credits                                                                            | 0-31                  |
| `TCC_EA0_RDREQ_GMI_CREDIT_STALL[n]`  | Cycles | Number of cycles there is a stall due to the read request interface running out of GMI credits                                                                           | 0-31                  |
| `TCC_EA0_RDREQ_DRAM_CREDIT_STALL[n]` | Cycles | Number of cycles there is a stall due to the read request interface running out of DRAM credits                                                                          | 0-31                  |
| `TCC_EA0_RDREQ_LEVEL[n]`             | Req    | The accumulated number of efficiency arbiter read requests in flight                                                                                                     | 0-31                  |
| `TCC_EA0_RDREQ_DRAM[n]`              | Req    | Number of 32-byte or 64-byte efficiency arbiter read requests to High Bandwidth Memory (HBM)                                                                             | 0-31                  |
| `TCC_EA0_WRREQ_DRAM[n]`              | Req    | Number of 32-byte or 64-byte efficiency arbiter write requests to HBM                                                                                                    | 0-31                  |
| `TCC_TAG_STALL[n]`                   | Cycles | Number of cycles the normal request pipeline in the tag is stalled for any reason                                                                                        | 0-31                  |
| `TCC_NORMAL_WRITEBACK[n]`            | Req    | Number of writebacks due to requests that are not writeback requests                                                                                                     | 0-31                  |
| `TCC_ALL_TC_OP_WB_WRITEBACK[n]`      | Req    | Number of writebacks due to all `TC_OP` writeback requests                                                                                                               | 0-31                  |
| `TCC_NORMAL_EVICT[n]`                | Req    | Number of evictions due to requests that are not invalidate or probe requests                                                                                            | 0-31                  |
| `TCC_ALL_TC_OP_INV_EVICT[n]`         | Req    | Number of evictions due to all `TC_OP` invalidate requests                                                                                                               | 0-31                  |

**MI200 hardware counter**

| Hardware counter                    | Unit   | Definition                                                                                                                                                               | Value range for `n`   |
|-------------------------------------|--------|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------|-----------------------|
| `TCC_CYCLE[n]`                      | Cycles | Number of L2 cache free-running clocks                                                                                                                                   | 0-31                  |
| `TCC_BUSY[n]`                       | Cycles | Number of L2 cache busy cycles                                                                                                                                           | 0-31                  |
| `TCC_REQ[n]`                        | Req    | Number of L2 cache requests of all types (measured at the tag block)                                                                                                     | 0-31                  |
| `TCC_STREAMING_REQ[n]`              | Req    | Number of L2 cache streaming requests (measured at the tag block)                                                                                                        | 0-31                  |
| `TCC_NC_REQ[n]`                     | Req    | Number of non-coherently cached requests (measured at the tag block)                                                                                                     | 0-31                  |
| `TCC_UC_REQ[n]`                     | Req    | Number of uncached requests. This is measured at the tag block                                                                                                           | 0-31                  |
| `TCC_CC_REQ[n]`                     | Req    | Number of coherently cached requests. This is measured at the tag block                                                                                                  | 0-31                  |
| `TCC_RW_REQ[n]`                     | Req    | Number of coherently cached with write requests. This is measured at the tag block                                                                                       | 0-31                  |
| `TCC_PROBE[n]`                      | Req    | Number of probe requests                                                                                                                                                 | 0-31                  |
| `TCC_PROBE_ALL[n]`                  | Req    | Number of external probe requests with `EA_TCC_preq_all == 1`                                                                                                            | 0-31                  |
| `TCC_READ[n]`                       | Req    | Number of L2 cache read requests (includes compressed reads but not metadata reads)                                                                                      | 0-31                  |
| `TCC_WRITE[n]`                      | Req    | Number of L2 cache write requests                                                                                                                                        | 0-31                  |
| `TCC_ATOMIC[n]`                     | Req    | Number of L2 cache atomic requests of all types                                                                                                                          | 0-31                  |
| `TCC_HIT[n]`                        | Req    | Number of L2 cache hits                                                                                                                                                  | 0-31                  |
| `TCC_MISS[n]`                       | Req    | Number of L2 cache misses                                                                                                                                                | 0-31                  |
| `TCC_WRITEBACK[n]`                  | Req    | Number of lines written back to the main memory, including writebacks of dirty lines and uncached write or atomic requests                                               | 0-31                  |
| `TCC_EA_WRREQ[n]`                   | Req    | Number of 32-byte and 64-byte transactions going over the `TC_EA_wrreq` interface (doesn’t include probe commands)                                                       | 0-31                  |
| `TCC_EA_WRREQ_64B[n]`               | Req    | Total number of 64-byte transactions (write or `CMPSWAP`) going over the `TC_EA_wrreq` interface                                                                         | 0-31                  |
| `TCC_EA_WR_UNCACHED_32B[n]`         | Req    | Number of 32 write or atomic going over the `TC_EA_wrreq` interface due to uncached traffic. A 64-byte request will be counted as 2                                      | 0-31                  |
| `TCC_EA_WRREQ_STALL[n]`             | Cycles | Number of cycles a write request is stalled                                                                                                                              | 0-31                  |
| `TCC_EA_WRREQ_IO_CREDIT_STALL[n]`   | Cycles | Number of cycles an efficiency arbiter write request is stalled due to the interface running out of input-output (IO) credits                                            | 0-31                  |
| `TCC_EA_WRREQ_GMI_CREDIT_STALL[n]`  | Cycles | Number of cycles an efficiency arbiter write request is stalled due to the interface running out of GMI credits                                                          | 0-31                  |
| `TCC_EA_WRREQ_DRAM_CREDIT_STALL[n]` | Cycles | Number of cycles an efficiency arbiter write request is stalled due to the interface running out of DRAM credits                                                         | 0-31                  |
| `TCC_TOO_MANY_EA_WRREQS_STALL[n]`   | Cycles | Number of cycles the L2 cache is unable to send an efficiency arbiter write request due to it reaching its maximum capacity of pending efficiency arbiter write requests | 0-31                  |
| `TCC_EA_WRREQ_LEVEL[n]`             | Req    | The accumulated number of efficiency arbiter write requests in flight                                                                                                    | 0-31                  |
| `TCC_EA_ATOMIC[n]`                  | Req    | Number of 32-byte or 64-byte atomic requests going over the `TC_EA_wrreq` interface                                                                                      | 0-31                  |
| `TCC_EA_ATOMIC_LEVEL[n]`            | Req    | The accumulated number of efficiency arbiter atomic requests in flight                                                                                                   | 0-31                  |
| `TCC_EA_RDREQ[n]`                   | Req    | Number of 32-byte or 64-byte read requests to efficiency arbiter                                                                                                         | 0-31                  |
| `TCC_EA_RDREQ_32B[n]`               | Req    | Number of 32-byte read requests to efficiency arbiter                                                                                                                    | 0-31                  |
| `TCC_EA_RD_UNCACHED_32B[n]`         | Req    | Number of 32-byte efficiency arbiter reads due to uncached traffic. A 64-byte request is counted as 2                                                                    | 0-31                  |
| `TCC_EA_RDREQ_IO_CREDIT_STALL[n]`   | Cycles | Number of cycles there is a stall due to the read request interface running out of IO credits                                                                            | 0-31                  |
| `TCC_EA_RDREQ_GMI_CREDIT_STALL[n]`  | Cycles | Number of cycles there is a stall due to the read request interface running out of GMI credits                                                                           | 0-31                  |
| `TCC_EA_RDREQ_DRAM_CREDIT_STALL[n]` | Cycles | Number of cycles there is a stall due to the read request interface running out of DRAM credits                                                                          | 0-31                  |
| `TCC_EA_RDREQ_LEVEL[n]`             | Req    | The accumulated number of efficiency arbiter read requests in flight                                                                                                     | 0-31                  |
| `TCC_EA_RDREQ_DRAM[n]`              | Req    | Number of 32-byte or 64-byte efficiency arbiter read requests to High Bandwidth Memory (HBM)                                                                             | 0-31                  |
| `TCC_EA_WRREQ_DRAM[n]`              | Req    | Number of 32-byte or 64-byte efficiency arbiter write requests to HBM                                                                                                    | 0-31                  |
| `TCC_TAG_STALL[n]`                  | Cycles | Number of cycles the normal request pipeline in the tag is stalled for any reason                                                                                        | 0-31                  |
| `TCC_NORMAL_WRITEBACK[n]`           | Req    | Number of writebacks due to requests that are not writeback requests                                                                                                     | 0-31                  |
| `TCC_ALL_TC_OP_WB_WRITEBACK[n]`     | Req    | Number of writebacks due to all `TC_OP` writeback requests                                                                                                               | 0-31                  |
| `TCC_NORMAL_EVICT[n]`               | Req    | Number of evictions due to requests that are not invalidate or probe requests                                                                                            | 0-31                  |
| `TCC_ALL_TC_OP_INV_EVICT[n]`        | Req    | Number of evictions due to all `TC_OP` invalidate requests                                                                                                               | 0-31                  |

Note the following:

* `TCC_REQ[n]` may be more than the number of requests arriving at the texture cache per channel,
  but it’s a good indication of the total amount of work that needs to be performed.
* For `TCC_EA0_WRREQ[n]`, atomics may travel over the same interface and are generally classified as
  write requests.
* CC mtypes can produce uncached requests, and those are included in
  `TCC_EA0_WR_UNCACHED_32B[n]`
* `TCC_EA0_WRREQ_LEVEL[n]` is primarily intended to measure average efficiency arbiter write latency.
  * Average write latency = `TCC_PERF_SEL_EA0_WRREQ_LEVEL` divided by `TCC_PERF_SEL_EA0_WRREQ`
* `TCC_EA0_ATOMIC_LEVEL[n]` is primarily intended to measure average efficiency arbiter atomic
  latency
  * Average atomic latency = `TCC_PERF_SEL_EA0_WRREQ_ATOMIC_LEVEL` divided by `TCC_PERF_SEL_EA0_WRREQ_ATOMIC`
* `TCC_EA0_RDREQ_LEVEL[n]` is primarily intended to measure average efficiency arbiter read latency.
  * Average read latency = `TCC_PERF_SEL_EA0_RDREQ_LEVEL` divided by `TCC_PERF_SEL_EA0_RDREQ`
* Stalls can occur regardless of the need for a read to be performed
* Normally, stalls are measured exactly at one point in the pipeline however in the case of
  `TCC_TAG_STALL[n]`, probes can stall the pipeline at a variety of places. There is no single point that
  can accurately measure the total stalls

## MI300 and MI200 Series derived metrics list

| Hardware counter   | Definition                                                                                                                                                                                                                            |
|--------------------|---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| `ALUStalledByLDS`  | Percentage of GPU time ALU units are stalled due to the LDS input queue being full or the output queue not being ready (value range: 0% (optimal) to 100%)                                                                            |
| `FetchSize`        | Total kilobytes fetched from the video memory; measured with all extra fetches and any cache or memory effects taken into account                                                                                                     |
| `FlatLDSInsts`     | Average number of flat instructions that read from or write to LDS, run per work item (affected by flow control)                                                                                                                      |
| `FlatVMemInsts`    | Average number of flat instructions that read from or write to the video memory, run per work item (affected by flow control). Includes flat instructions that read from or write to scratch                                          |
| `GDSInsts`         | Average number of global data share read or write instructions run per work item (affected by flow control)                                                                                                                           |
| `GPUBusy`          | Percentage of time GPU is busy                                                                                                                                                                                                        |
| `L2CacheHit`       | Percentage of fetch, write, atomic, and other instructions that hit the data in L2 cache (value range: 0% (no hit) to 100% (optimal))                                                                                                 |
| `LDSBankConflict`  | Percentage of GPU time LDS is stalled by bank conflicts (value range: 0% (optimal) to 100%)                                                                                                                                           |
| `LDSInsts`         | Average number of LDS read or write instructions run per work item (affected by flow control). Excludes flat instructions that read from or write to LDS.                                                                             |
| `MemUnitBusy`      | Percentage of GPU time the memory unit is active, which is measured with all extra fetches and writes and any cache or memory effects taken into account (value range: 0% to 100% (fetch-bound))                                      |
| `MemUnitStalled`   | Percentage of GPU time the memory unit is stalled (value range: 0% (optimal) to 100%)                                                                                                                                                 |
| `MemWrites32B`     | Total number of effective 32B write transactions to the memory                                                                                                                                                                        |
| `TCA_BUSY_sum`     | Total number of cycles texture cache arbiter has a pending request, over all texture cache arbiter instances                                                                                                                          |
| `TCA_CYCLE_sum`    | Total number of cycles over all texture cache arbiter instances                                                                                                                                                                       |
| `SALUBusy`         | Percentage of GPU time scalar ALU instructions are processed (value range: 0% to 100% (optimal))                                                                                                                                      |
| `SALUInsts`        | Average number of scalar ALU instructions run per work item (affected by flow control)                                                                                                                                                |
| `SFetchInsts`      | Average number of scalar fetch instructions from the video memory run per work item (affected by flow control)                                                                                                                        |
| `VALUBusy`         | Percentage of GPU time vector ALU instructions are processed (value range: 0% to 100% (optimal))                                                                                                                                      |
| `VALUInsts`        | Average number of vector ALU instructions run per work item (affected by flow control)                                                                                                                                                |
| `VALUUtilization`  | Percentage of active vector ALU threads in a wave, where a lower number can mean either more thread divergence in a wave or that the work-group size is not a multiple of 64 (value range: 0%, 100% (optimal - no thread divergence)) |
| `VFetchInsts`      | Average number of vector fetch instructions from the video memory run per work-item (affected by flow control); excludes flat instructions that fetch from video memory                                                               |
| `VWriteInsts`      | Average number of vector write instructions to the video memory run per work-item (affected by flow control); excludes flat instructions that write to video memory                                                                   |
| `Wavefronts`       | Total wavefronts                                                                                                                                                                                                                      |
| `WRITE_REQ_32B`    | Total number of 32-byte effective memory writes                                                                                                                                                                                       |
| `WriteSize`        | Total kilobytes written to the video memory; measured with all extra fetches and any cache or memory effects taken into account                                                                                                       |
| `WriteUnitStalled` | Percentage of GPU time the write unit is stalled (value range: 0% (optimal) to 100%)                                                                                                                                                  |

You can lower `ALUStalledByLDS` by reducing LDS bank conflicts or number of LDS accesses.
You can lower `MemUnitStalled` by reducing the number or size of fetches and writes.
`MemUnitBusy` includes the stall time (`MemUnitStalled`).

### Hardware counters by and over all texture addressing unit instances

The following table shows the hardware counters *by* all texture addressing unit instances.

| Hardware counter                       | Definition                                                                       |
|----------------------------------------|----------------------------------------------------------------------------------|
| `TA_BUFFER_WAVEFRONTS_sum`             | Total number of buffer wavefronts processed                                      |
| `TA_BUFFER_READ_WAVEFRONTS_sum`        | Total number of buffer read wavefronts processed                                 |
| `TA_BUFFER_WRITE_WAVEFRONTS_sum`       | Total number of buffer write wavefronts processed                                |
| `TA_BUFFER_ATOMIC_WAVEFRONTS_sum`      | Total number of buffer atomic wavefronts processed                               |
| `TA_BUFFER_TOTAL_CYCLES_sum`           | Total number of buffer cycles (including read and write) issued to texture cache |
| `TA_BUFFER_COALESCED_READ_CYCLES_sum`  | Total number of coalesced buffer read cycles issued to texture cache             |
| `TA_BUFFER_COALESCED_WRITE_CYCLES_sum` | Total number of coalesced buffer write cycles issued to texture cache            |
| `TA_FLAT_READ_WAVEFRONTS_sum`          | Sum of flat opcode reads processed                                               |
| `TA_FLAT_WRITE_WAVEFRONTS_sum`         | Sum of flat opcode writes processed                                              |
| `TA_FLAT_WAVEFRONTS_sum`               | Total number of flat opcode wavefronts processed                                 |
| `TA_FLAT_ATOMIC_WAVEFRONTS_sum`        | Total number of flat opcode atomic wavefronts processed                          |
| `TA_TOTAL_WAVEFRONTS_sum`              | Total number of wavefronts processed                                             |

The following table shows the hardware counters *over* all texture addressing unit instances.

| Hardware counter                   | Definition                                                                                  |
|------------------------------------|---------------------------------------------------------------------------------------------|
| `TA_ADDR_STALLED_BY_TC_CYCLES_sum` | Total number of cycles texture addressing unit address path is stalled by texture cache     |
| `TA_ADDR_STALLED_BY_TD_CYCLES_sum` | Total number of cycles texture addressing unit address path is stalled by texture data unit |
| `TA_BUSY_avr`                      | Average number of busy cycles                                                               |
| `TA_BUSY_max`                      | Maximum number of texture addressing unit busy cycles                                       |
| `TA_BUSY_min`                      | Minimum number of texture addressing unit busy cycles                                       |
| `TA_DATA_STALLED_BY_TC_CYCLES_sum` | Total number of cycles texture addressing unit data path is stalled by texture cache        |
| `TA_TA_BUSY_sum`                   | Total number of texture addressing unit busy cycles                                         |

### Hardware counters over all texture cache per channel instances

| Hardware counter                      | Definition                                                                                                                                                                                                                                    |
|---------------------------------------|-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| `TCC_ALL_TC_OP_WB_WRITEBACK_sum`      | Total number of writebacks due to all `TC_OP` writeback requests.                                                                                                                                                                             |
| `TCC_ALL_TC_OP_INV_EVICT_sum`         | Total number of evictions due to all `TC_OP` invalidate requests.                                                                                                                                                                             |
| `TCC_ATOMIC_sum`                      | Total number of L2 cache atomic requests of all types.                                                                                                                                                                                        |
| `TCC_BUSY_avr`                        | Average number of L2 cache busy cycles.                                                                                                                                                                                                       |
| `TCC_BUSY_sum`                        | Total number of L2 cache busy cycles.                                                                                                                                                                                                         |
| `TCC_CC_REQ_sum`                      | Total number of coherently cached requests.                                                                                                                                                                                                   |
| `TCC_CYCLE_sum`                       | Total number of L2 cache free running clocks.                                                                                                                                                                                                 |
| `TCC_EA0_WRREQ_sum`                   | Total number of 32-byte and 64-byte transactions going over the `TC_EA0_wrreq` interface. Atomics may travel over the same interface and are generally classified as write requests. This does not include probe commands.                    |
| `TCC_EA0_WRREQ_64B_sum`               | Total number of 64-byte transactions (write or CMPSWAP) going over the `TC_EA0_wrreq` interface.                                                                                                                                              |
| `TCC_EA0_WR_UNCACHED_32B_sum`         | Total Number of 32-byte write or atomic going over the `TC_EA0_wrreq` interface due to uncached traffic. Note that coherently cached mtypes can produce uncached requests, and those are included in this. A 64-byte request is counted as 2. |
| `TCC_EA0_WRREQ_STALL_sum`             | Total Number of cycles a write request is stalled, over all instances.                                                                                                                                                                        |
| `TCC_EA0_WRREQ_IO_CREDIT_STALL_sum`   | Total number of cycles an efficiency arbiter write request is stalled due to the interface running out of IO credits, over all instances.                                                                                                     |
| `TCC_EA0_WRREQ_GMI_CREDIT_STALL_sum`  | Total number of cycles an efficiency arbiter write request is stalled due to the interface running out of GMI credits, over all instances.                                                                                                    |
| `TCC_EA0_WRREQ_DRAM_CREDIT_STALL_sum` | Total number of cycles an efficiency arbiter write request is stalled due to the interface running out of DRAM credits, over all instances.                                                                                                   |
| `TCC_EA0_WRREQ_LEVEL_sum`             | Total number of efficiency arbiter write requests in flight.                                                                                                                                                                                  |
| `TCC_EA0_RDREQ_LEVEL_sum`             | Total number of efficiency arbiter read requests in flight.                                                                                                                                                                                   |
| `TCC_EA0_ATOMIC_sum`                  | Total Number of 32-byte or 64-byte atomic requests going over the `TC_EA0_wrreq` interface.                                                                                                                                                   |
| `TCC_EA0_ATOMIC_LEVEL_sum`            | Total number of efficiency arbiter atomic requests in flight.                                                                                                                                                                                 |
| `TCC_EA0_RDREQ_sum`                   | Total number of 32-byte or 64-byte read requests to efficiency arbiter.                                                                                                                                                                       |
| `TCC_EA0_RDREQ_32B_sum`               | Total number of 32-byte read requests to efficiency arbiter.                                                                                                                                                                                  |
| `TCC_EA0_RD_UNCACHED_32B_sum`         | Total number of 32-byte efficiency arbiter reads due to uncached traffic.                                                                                                                                                                     |
| `TCC_EA0_RDREQ_IO_CREDIT_STALL_sum`   | Total number of cycles there is a stall due to the read request interface running out of IO credits.                                                                                                                                          |
| `TCC_EA0_RDREQ_GMI_CREDIT_STALL_sum`  | Total number of cycles there is a stall due to the read request interface running out of GMI credits.                                                                                                                                         |
| `TCC_EA0_RDREQ_DRAM_CREDIT_STALL_sum` | Total number of cycles there is a stall due to the read request interface running out of DRAM credits.                                                                                                                                        |
| `TCC_EA0_RDREQ_DRAM_sum`              | Total number of 32-byte or 64-byte efficiency arbiter read requests to HBM.                                                                                                                                                                   |
| `TCC_EA0_WRREQ_DRAM_sum`              | Total number of 32-byte or 64-byte efficiency arbiter write requests to HBM.                                                                                                                                                                  |
| `TCC_HIT_sum`                         | Total number of L2 cache hits.                                                                                                                                                                                                                |
| `TCC_MISS_sum`                        | Total number of L2 cache misses.                                                                                                                                                                                                              |
| `TCC_NC_REQ_sum`                      | Total number of non-coherently cached requests.                                                                                                                                                                                               |
| `TCC_NORMAL_WRITEBACK_sum`            | Total number of writebacks due to requests that are not writeback requests.                                                                                                                                                                   |
| `TCC_NORMAL_EVICT_sum`                | Total number of evictions due to requests that are not invalidate or probe requests.                                                                                                                                                          |
| `TCC_PROBE_sum`                       | Total number of probe requests.                                                                                                                                                                                                               |
| `TCC_PROBE_ALL_sum`                   | Total number of external probe requests with `EA0_TCC_preq_all == 1`.                                                                                                                                                                         |
| `TCC_READ_sum`                        | Total number of L2 cache read requests (including compressed reads but not metadata reads).                                                                                                                                                   |
| `TCC_REQ_sum`                         | Total number of all types of L2 cache requests.                                                                                                                                                                                               |
| `TCC_RW_REQ_sum`                      | Total number of coherently cached with write requests.                                                                                                                                                                                        |
| `TCC_STREAMING_REQ_sum`               | Total number of L2 cache streaming requests.                                                                                                                                                                                                  |
| `TCC_TAG_STALL_sum`                   | Total number of cycles the normal request pipeline in the tag is stalled for any reason.                                                                                                                                                      |
| `TCC_TOO_MANY_EA0_WRREQS_STALL_sum`   | Total number of cycles L2 cache is unable to send an efficiency arbiter write request due to it reaching its maximum capacity of pending efficiency arbiter write requests.                                                                   |
| `TCC_UC_REQ_sum`                      | Total number of uncached requests.                                                                                                                                                                                                            |
| `TCC_WRITE_sum`                       | Total number of L2 cache write requests.                                                                                                                                                                                                      |
| `TCC_WRITEBACK_sum`                   | Total number of lines written back to the main memory including writebacks of dirty lines and uncached write or atomic requests.                                                                                                              |
| `TCC_WRREQ_STALL_max`                 | Maximum number of cycles a write request is stalled.                                                                                                                                                                                          |

### Hardware counters by, for, or over all texture cache per pipe instances

The following table shows the hardware counters *by* all texture cache per pipe instances.

| Hardware counter                 | Definition                                                                     |
|----------------------------------|--------------------------------------------------------------------------------|
| `TCP_TA_TCP_STATE_READ_sum`      | Total number of state reads by ATCPPI                                          |
| `TCP_TOTAL_CACHE_ACCESSES_sum`   | Total number of vector L1d accesses (including hits and misses)                |
| `TCP_UTCL1_PERMISSION_MISS_sum`  | Total number of unified translation cache (L1) permission misses               |
| `TCP_UTCL1_REQUEST_sum`          | Total number of address translation requests to unified translation cache (L1) |
| `TCP_UTCL1_TRANSLATION_MISS_sum` | Total number of unified translation cache (L1) translation misses              |
| `TCP_UTCL1_TRANSLATION_HIT_sum`  | Total number of unified translation cache (L1) translation hits                |

The following table shows the hardware counters *for* all texture cache per pipe instances.

| Hardware counter                | Definition                                                                                       |
|---------------------------------|--------------------------------------------------------------------------------------------------|
| `TCP_TCC_READ_REQ_LATENCY_sum`  | Total vector L1d to L2 request latency over all wavefronts for reads and atomics with return     |
| `TCP_TCC_WRITE_REQ_LATENCY_sum` | Total vector L1d to L2 request latency over all wavefronts for writes and atomics without return |
| `TCP_TCP_LATENCY_sum`           | Total wave access latency to vector L1d over all wavefronts                                      |

The following table shows the hardware counters *over* all texture cache per pipe instances.

| Hardware counter                          | Definition                                                                           |
|-------------------------------------------|--------------------------------------------------------------------------------------|
| `TCP_ATOMIC_TAGCONFLICT_STALL_CYCLES_sum` | Total number of cycles tag RAM conflict stalls on an atomic                          |
| `TCP_GATE_EN1_sum`                        | Total number of cycles vector L1d interface clocks are turned on                     |
| `TCP_GATE_EN2_sum`                        | Total number of cycles vector L1d core clocks are turned on                          |
| `TCP_PENDING_STALL_CYCLES_sum`            | Total number of cycles vector L1d cache is stalled due to data pending from L2 Cache |
| `TCP_READ_TAGCONFLICT_STALL_CYCLES_sum`   | Total number of cycles tag RAM conflict stalls on a read                             |
| `TCP_TCC_ATOMIC_WITH_RET_REQ_sum`         | Total number of atomic requests to L2 cache with return                              |
| `TCP_TCC_ATOMIC_WITHOUT_RET_REQ_sum`      | Total number of atomic requests to L2 cache without return                           |
| `TCP_TCC_CC_READ_REQ_sum`                 | Total number of coherently cached read requests to L2 cache                          |
| `TCP_TCC_CC_WRITE_REQ_sum`                | Total number of coherently cached write requests to L2 cache                         |
| `TCP_TCC_CC_ATOMIC_REQ_sum`               | Total number of coherently cached atomic requests to L2 cache                        |
| `TCP_TCC_NC_READ_REQ_sum`                 | Total number of non-coherently cached read requests to L2 cache                      |
| `TCP_TCC_NC_WRITE_REQ_sum`                | Total number of non-coherently cached write requests to L2 cache                     |
| `TCP_TCC_NC_ATOMIC_REQ_sum`               | Total number of non-coherently cached atomic requests to L2 cache                    |
| `TCP_TCC_READ_REQ_sum`                    | Total number of read requests to L2 cache                                            |
| `TCP_TCC_RW_READ_REQ_sum`                 | Total number of coherently cached with write read requests to L2 cache               |
| `TCP_TCC_RW_WRITE_REQ_sum`                | Total number of coherently cached with write write requests to L2 cache              |
| `TCP_TCC_RW_ATOMIC_REQ_sum`               | Total number of coherently cached with write atomic requests to L2 cache             |
| `TCP_TCC_UC_READ_REQ_sum`                 | Total number of uncached read requests to L2 cache                                   |
| `TCP_TCC_UC_WRITE_REQ_sum`                | Total number of uncached write requests to L2 cache                                  |
| `TCP_TCC_UC_ATOMIC_REQ_sum`               | Total number of uncached atomic requests to L2 cache                                 |
| `TCP_TCC_WRITE_REQ_sum`                   | Total number of write requests to L2 cache                                           |
| `TCP_TCR_TCP_STALL_CYCLES_sum`            | Total number of cycles texture cache router stalls vector L1d                        |
| `TCP_TD_TCP_STALL_CYCLES_sum`             | Total number of cycles texture data unit stalls vector L1d                           |
| `TCP_TOTAL_ACCESSES_sum`                  | Total number of vector L1d accesses                                                  |
| `TCP_TOTAL_READ_sum`                      | Total number of vector L1d read accesses                                             |
| `TCP_TOTAL_WRITE_sum`                     | Total number of vector L1d write accesses                                            |
| `TCP_TOTAL_ATOMIC_WITH_RET_sum`           | Total number of vector L1d atomic requests with return                               |
| `TCP_TOTAL_ATOMIC_WITHOUT_RET_sum`        | Total number of vector L1d atomic requests without return                            |
| `TCP_TOTAL_WRITEBACK_INVALIDATES_sum`     | Total number of vector L1d writebacks and invalidates                                |
| `TCP_VOLATILE_sum`                        | Total number of L1 volatile pixels or buffers from texture addressing unit           |
| `TCP_WRITE_TAGCONFLICT_STALL_CYCLES_sum`  | Total number of cycles tag RAM conflict stalls on a write                            |

### Hardware counter over all texture data unit instances

| Hardware counter               | Definition                                                                               |
|--------------------------------|------------------------------------------------------------------------------------------|
| `TD_ATOMIC_WAVEFRONT_sum`      | Total number of atomic wavefront instructions                                            |
| `TD_COALESCABLE_WAVEFRONT_sum` | Total number of coalescable wavefronts according to texture addressing unit              |
| `TD_LOAD_WAVEFRONT_sum`        | Total number of wavefront instructions (read, write, atomic)                             |
| `TD_SPI_STALL_sum`             | Total number of cycles texture data unit is stalled by shader processor input            |
| `TD_STORE_WAVEFRONT_sum`       | Total number of write wavefront instructions                                             |
| `TD_TC_STALL_sum`              | Total number of cycles texture data unit is stalled waiting for texture cache data       |
| `TD_TD_BUSY_sum`               | Total number of texture data unit busy cycles while it is processing or waiting for data |

---

Source: https://rocm.docs.amd.com/en/latest/reference/gpu-arch/mi250.html

# AMD Instinct™ MI250 microarchitecture

The microarchitecture of the AMD Instinct MI250 GPU is based on the
AMD CDNA 2 architecture that targets compute applications such as HPC,
artificial intelligence (AI), and machine learning (ML) and that run on
everything from individual servers to the world’s largest exascale
supercomputers. The overall system architecture is designed for extreme
scalability and compute performance.

The following image shows the components of a single Graphics Compute Die (GCD) of the CDNA 2 architecture. On the top and the bottom are AMD Infinity Fabric™
interfaces and their physical links that are used to connect the GPU die to the
other system-level components of the node (see also Section 2.2). Both
interfaces can drive four AMD Infinity Fabric links. One of the AMD Infinity
Fabric links of the controller at the bottom can be configured as a PCIe link.
Each of the AMD Infinity Fabric links between GPUs can run at up to 25 GT/sec,
which correlates to a peak transfer bandwidth of 50 GB/sec for a 16-wide link (
two bytes per transaction). Section 2.2 has more details on the number of AMD
Infinity Fabric links and the resulting transfer rates between the system-level
components.

To the left and the right are memory controllers that attach the High Bandwidth
Memory (HBM) modules to the GCD. AMD Instinct MI250 GPUs use HBM2e, which offers
a peak memory bandwidth of 1.6 TB/sec per GCD.

The execution units of the GPU are depicted in the following image as Compute
Units (CU). The MI250 GCD has 104 active CUs. Each compute unit is further
subdivided into four SIMD units that process SIMD instructions of 16 data
elements per instruction (for the FP64 data type). This enables the CU to
process 64 work items (a so-called “wavefront”) at a peak clock frequency of 1.7
GHz. Therefore, the theoretical maximum FP64 peak performance per GCD is 22.6
TFLOPS for vector instructions. This equates to 45.3 TFLOPS for vector instructions for both GCDs together. The MI250 compute units also provide specialized
execution units (also called matrix cores), which are geared toward executing
matrix operations like matrix-matrix multiplications. For FP64, the peak
performance of these units amounts to 90.5 TFLOPS.

![Structure of a single GCD in the AMD Instinct MI250 GPU.](reference/gpu-arch/images/image001.png)

#### Peak-performance capabilities of the MI250 OAM for different data types.

| Computation and Data Type   |   FLOPS/CLOCK/CU |   Peak TFLOPS |
|-----------------------------|------------------|---------------|
| Matrix FP64                 |              256 |          90.5 |
| Vector FP64                 |              128 |          45.3 |
| Matrix FP32                 |              256 |          90.5 |
| Packed FP32                 |              256 |          90.5 |
| Vector FP32                 |              128 |          45.3 |
| Matrix FP16                 |             1024 |         362.1 |
| Matrix BF16                 |             1024 |         362.1 |
| Matrix INT8                 |             1024 |         362.1 |

The above table summarizes the aggregated peak performance of the AMD Instinct MI250 Open Compute Platform (OCP) Open Accelerator Modules (OAMs) and its two GCDs for different data types and execution units. The middle column lists the peak performance (number of data elements processed in a single instruction) of a single compute unit if a SIMD (or matrix) instruction is being retired in each clock cycle. The third column lists the theoretical peak performance of the OAM module. The theoretical aggregated peak memory bandwidth of the GPU is 3.2 TB/sec (1.6 TB/sec per GCD).

![Dual-GCD architecture of the AMD Instinct MI250 GPUs](reference/gpu-arch/images/image002.png)

The following image shows the block diagram of an OAM package that consists
of two GCDs, each of which constitutes one GPU device in the system. The two
GCDs in the package are connected via four AMD Infinity Fabric links running at
a theoretical peak rate of 25 GT/sec, giving 200 GB/sec peak transfer bandwidth
between the two GCDs of an OAM, or a bidirectional peak transfer bandwidth of
400 GB/sec for the same.

## Node-level architecture

The following image shows the node-level architecture of a system that is
based on the AMD Instinct MI250 GPU. The MI250 OAMs attach to the host
system via PCIe Gen 4 x16 links (yellow lines). Each GCD maintains its own PCIe
x16 link to the host part of the system. Depending on the server platform, the
GCD can attach to the AMD EPYC processor directly or via an optional PCIe switch
. Note that some platforms may offer an x8 interface to the GCDs, which reduces
the available host-to-GPU bandwidth.

![Block diagram of AMD Instinct MI250 GPUs with 3rd Generation AMD EPYC processor](reference/gpu-arch/images/image003.png)

The preceding image shows the node-level architecture of a system with AMD
EPYC processors in a dual-socket configuration and four AMD Instinct MI250
GPUs. The MI250 OAMs attach to the host processors system via PCIe Gen 4
x16 links (yellow lines). Depending on the system design, a PCIe switch may
exist to make more PCIe lanes available for additional components like network
interfaces and/or storage devices. Each GCD maintains its own PCIe x16 link to
the host part of the system or to the PCIe switch. Please note, some platforms
may offer an x8 interface to the GCDs, which will reduce the available
host-to-GPU bandwidth.

Between the OAMs and their respective GCDs, a peer-to-peer (P2P) network allows
for direct data exchange between the GPU dies via AMD Infinity Fabric links (
black, green, and red lines). Each of these 16-wide links connects to one of the
two GPU dies in the MI250 OAM and operates at 25 GT/sec, which corresponds to a
theoretical peak transfer rate of 50 GB/sec per link (or 100 GB/sec
bidirectional peak transfer bandwidth). The GCD pairs 2 and 6 as well as GCDs 0
and 4 connect via two XGMI links, which is indicated by the thicker red line in
the preceding image.

---

Source: https://rocm.docs.amd.com/en/latest/reference/gpu-arch/mi100.html

# AMD Instinct™ MI100 microarchitecture

The following image shows the node-level architecture of a system that
comprises two AMD EPYC™ processors and (up to) eight AMD Instinct™ GPUs.
The two EPYC processors are connected to each other with the AMD Infinity™
fabric which provides a high-bandwidth (up to 18 GT/sec) and coherent links such
that each processor can access the available node memory as a single
shared-memory domain in a non-uniform memory architecture (NUMA) fashion. In a
2P, or dual-socket, configuration, three AMD Infinity™ fabric links are
available to connect the processors plus one PCIe Gen 4 x16 link per processor
can attach additional I/O devices such as the host adapters for the network
fabric.

![Structure of a single GCD in the AMD Instinct MI100 GPU](reference/gpu-arch/images/image004.png)

In a typical node configuration, each processor can host up to four AMD
Instinct™ GPUs that are attached using PCIe Gen 4 links at 16 GT/sec,
which corresponds to a peak bidirectional link bandwidth of 32 GB/sec. Each hive
of four GPUs can participate in a fully connected, coherent AMD
Instinct™ fabric that connects the four GPUs using 23 GT/sec AMD
Infinity fabric links that run at a higher frequency than the inter-processor
links. This inter-GPU link can be established in certified server systems if the
GPUs are mounted in neighboring PCIe slots by installing the AMD Infinity
Fabric™ bridge for the AMD Instinct™ GPUs.

## Microarchitecture

The microarchitecture of the AMD Instinct GPUs is based on the AMD CDNA
architecture, which targets compute applications such as high-performance
computing (HPC) and AI & machine learning (ML) that run on everything from
individual servers to the world’s largest exascale supercomputers. The overall
system architecture is designed for extreme scalability and compute performance.

![Structure of the AMD Instinct GPU (MI100 generation)](reference/gpu-arch/images/image005.png)

The above image shows the AMD Instinct GPU with its PCIe Gen 4 x16
link (16 GT/sec, at the bottom) that connects the GPU to (one of) the host
processor(s). It also shows the three AMD Infinity Fabric ports that provide
high-speed links (23 GT/sec, also at the bottom) to the other GPUs of the local
hive.

On the left and right of the floor plan, the High Bandwidth Memory (HBM)
attaches via the GPU memory controller.  The MI100 generation of the AMD
Instinct GPU offers four stacks of HBM generation 2 (HBM2) for a total
of 32GB with a 4,096bit-wide memory interface. The peak memory bandwidth of the
attached HBM2 is 1.228 TB/sec at a memory clock frequency of 1.2 GHz.

The execution units of the GPU are depicted in the above image as Compute
Units (CU). There are a total 120 compute units that are physically organized
into eight Shader Engines (SE) with fifteen compute units per shader engine.
Each compute unit is further sub-divided into four SIMD units that process SIMD
instructions of 16 data elements per instruction. This enables the CU to process
64 data elements (a so-called ‘wavefront’) at a peak clock frequency of 1.5 GHz.
Therefore, the theoretical maximum FP64 peak performance is 11.5 TFLOPS
(`4 [SIMD units] x 16 [elements per instruction] x 120 [CU] x 1.5 [GHz]`).

![Block diagram of an MI100 compute unit with detailed SIMD view of the AMD CDNA architecture](reference/gpu-arch/images/image006.png)

The preceding image shows the block diagram of a single CU of an AMD Instinct™
MI100 GPU and summarizes how instructions flow through the execution
engines. The CU fetches the instructions via a 32KB instruction cache and moves
them forward to execution via a dispatcher. The CU can handle up to ten
wavefronts at a time and feed their instructions into the execution unit. The
execution unit contains 256 vector general-purpose registers (VGPR) and 800
scalar general-purpose registers (SGPR). The VGPR and SGPR are dynamically
allocated to the executing wavefronts. A wavefront can access a maximum of 102
scalar registers. Excess scalar-register usage will cause register spilling and
thus may affect execution performance.

A wavefront can occupy any number of VGPRs from 0 to 256, directly affecting
occupancy; that is, the number of concurrently active wavefronts in the CU. For
instance, with 119 VGPRs used, only two wavefronts can be active in the CU at
the same time. With the instruction latency of four cycles per SIMD instruction,
the occupancy should be as high as possible such that the compute unit can
improve execution efficiency by scheduling instructions from multiple
wavefronts.

#### Peak-performance capabilities of MI100 for different data types.

| Computation and Data Type   |   FLOPS/CLOCK/CU |   Peak TFLOPS |
|-----------------------------|------------------|---------------|
| Vector FP64                 |               64 |          11.5 |
| Matrix FP32                 |              256 |          46.1 |
| Vector FP32                 |              128 |          23.1 |
| Matrix FP16                 |             1024 |         184.6 |
| Matrix BF16                 |              512 |          92.3 |

---

Source: https://rocm.docs.amd.com/en/latest/reference/system-optimization/index.html

# AMD GPU system optimization

This guide outlines system setup and tuning suggestions for AMD hardware to
optimize performance for specific types of workloads or use cases. The contents
are structured according to the hardware:

* [AMD Instinct MI355X](https://instinct.docs.amd.com/projects/system-acceptance/en/latest/gpus/mi355x.html)
* [AMD Instinct MI350X](https://instinct.docs.amd.com/projects/system-acceptance/en/latest/gpus/mi350x.html)
* [AMD Instinct MI325X](https://instinct.docs.amd.com/projects/system-acceptance/en/latest/gpus/mi325x.html)
* [AMD Instinct MI300X](https://instinct.docs.amd.com/projects/system-acceptance/en/latest/gpus/mi300x.html)
* [AMD Instinct MI300A](https://instinct.docs.amd.com/projects/system-acceptance/en/latest/gpus/mi300a.html)
* [AMD Instinct MI250 / MI250X](https://instinct.docs.amd.com/projects/system-acceptance/en/latest/gpus/mi250.html)
* [AMD Instinct MI210](https://instinct.docs.amd.com/projects/system-acceptance/en/latest/gpus/mi210.html)
* [AMD Instinct MI100](https://instinct.docs.amd.com/projects/system-acceptance/en/latest/gpus/mi100.html)

* [AMD RDNA3.5 system optimization](https://rocm.docs.amd.com/en/latest/reference/system-optimization/rdna3-5.html)
* [AMD RDNA2 system optimization](https://rocm.docs.amd.com/en/latest/reference/system-optimization/rdna2.html)

* [GPU isolation techniques](https://rocm.docs.amd.com/en/latest/reference/system-optimization/gpu-isolation.html)
* [BAR access limits](https://rocm.docs.amd.com/en/latest/reference/system-optimization/bar-access-limits.html)

---

Source: https://rocm.docs.amd.com/en/latest/reference/system-optimization/cdna.html

# AMD Instinct system optimization

---

Source: https://rocm.docs.amd.com/en/latest/reference/system-optimization/rdna.html

# AMD Radeon and Ryzen system optimization

This topic outlines system setup and tuning suggestions for AMD Radeon and
Ryzen hardware to optimize performance for specific workloads or use cases.

---

Source: https://rocm.docs.amd.com/en/latest/reference/system-optimization/rdna3-5.html

# AMD RDNA3.5 system optimization

This topic describes how to optimize systems powered by AMD Ryzen APUs with
RDNA3.5 architecture. These APUs combine high-performance CPU cores with
integrated RDNA3.5 graphics, and support LPDDR5X-8000 or DDR5 memory, making
them particularly well-suited for:

* LLM development and inference systems
* High-performance workstations
* Virtualization hosts running multiple VMs
* GPU compute and parallel processing
* Gaming systems
* Home servers and AI development platforms

## Memory settings

AMD Ryzen APUs with RDNA3.5 architecture (gfx1150, gfx1151, and gfx1152 LLVM
targets) handle memory access through GPU Virtual Memory (GPUVM), which
provides per-process GPU virtual address spaces (VMIDs) rather than a separate,
discrete VRAM pool.

As a result, memory on RDNA3.5 APUs is mapped rather than physically
partitioned. The terms Graphics Address Remapping Table (GART) and Graphics
Translation Table (GTT) describe limits on how much system memory can be mapped
into GPU address spaces and who can use it, rather than distinct types of
physical memory.

* **GART**

Defines the amount of platform address space (system RAM or Memory-Mapped I/O)
  that can be mapped into the GPU virtual address space used by the kernel driver.
  On systems with physically shared CPU and GPU memory, such as RDNA3.5-based
  systems, this mapped system memory effectively serves as VRAM for the GPU.
  GART is typically kept relatively small to limit GPU page-table size and is
  primarily used for driver-internal operations.
* **GTT**

Defines the amount of system RAM that can be mapped into GPU virtual address
  spaces for user processes. This is the memory pool used by applications such
  as PyTorch and other AI/compute workloads. GTT allocations are dynamic and
  not permanently reserved, allowing the operating system to reclaim memory when
  the GPU isn’t actively using it. By default, the GTT limit is set to
  approximately 50 percent of total system RAM.

#### NOTE
On systems with physically shared CPU and GPU memory, such as RDNA3.5-based
systems, several terms are often used interchangeably in firmware menus,
documentation, and community discussions:

* VRAM
* Carve-out
* GART
* Dedicated GPU memory
* Firmware-reserved GPU memory

In this topic, VRAM will be used going forward.

You can adjust the amount of memory available to the GPU by:

* Increasing the VRAM in BIOS, or
* Reducing the configured GTT size to be smaller than the reserved amount.

If the GTT size is larger than the VRAM, the AMD GPU driver performs VRAM
allocations using GTT (GTT-backed allocations), as described in the
[torvalds/linux@759e764](https://github.com/torvalds/linux/commit/759e764f7d587283b4e0b01ff930faca64370e59)
GitHub commit.

Because memory is physically shared, there’s no performance distinction
like that of discrete GPUs, where dedicated VRAM is significantly faster than
system memory. Firmware may optionally reserve some memory exclusively for GPU
use, but this provides little benefit for most workloads while permanently
reducing available system memory.

For this reason, AI frameworks work more efficiently with GTT-backed allocations. GTT
allows large, flexible mappings without permanently reserving memory, resulting
in better overall system utilization on unified memory systems.

### Configuring shared memory limits on Linux

The maximum amount of shared GPU-accessible memory can be increased by changing
the kernel **Translation Table Manager (TTM)** page limit. This setting controls
how many system memory pages can be mapped for GPU use and is exposed at:

```default
/sys/module/ttm/parameters/pages_limit
```

The value is expressed in pages, and not bytes or gigabytes (GB).

#### NOTE
It’s recommended to keep the dedicated VRAM reservation in BIOS small
(for example, 0.5 GB) and increase the shared (TTM/GTT) limit instead.

A helper utility is available to simplify configuration.

1. Install `pipx`:
 ```default
 sudo apt install pipx
 pipx ensurepath
 ```
2. Install the AMD debug tools:
 ```default
 pipx install amd-debug-tools
 ```
3. Query the current shared memory configuration:
 ```default
 amd-ttm
 ```
4. Set the usable shared memory (in GB):
 ```default
 amd-ttm --set <NUM>
 ```
5. Reboot for changes to take effect.

#### NOTE
The amd-ttm command converts the pages to GB to help the users.

#### Example with output

Check the current settings:

```default
amd-ttm
💻 Current TTM pages limit: 16469033 pages (62.82 GB)
💻 Total system memory: 125.65 GB
```

Change the usable shared memory:

```default
❯ amd-ttm --set 100
🐧 Successfully set TTM pages limit to 26214400 pages (100.00 GB)
🐧 Configuration written to /etc/modprobe.d/ttm.conf
○ NOTE: You need to reboot for changes to take effect.
Would you like to reboot the system now? (y/n): y
```

Revert to kernel defaults:

```default
 ❯ amd-ttm --clear
🐧 Configuration /etc/modprobe.d/ttm.conf removed
 Would you like to reboot the system now? (y/n): y
```

## Operating system support

The ROCm operating system requirements are available in
[Operating system support](https://rocm.docs.amd.com/en/latest/about/release-notes.html#release-supported-os).

AMD Ryzen AI Max series APUs (gfx1151) have additional kernel version
requirements, as described in the  following section.

### Required Linux kernel version

Support for AMD Ryzen AI Max series APUs requires specific Linux kernel fixes
that update internal limits in the AMD KFD driver to ensure correct queue
creation and memory availability checks. Without these updates, GPU compute
workloads might fail to initialize or exhibit unpredictable behavior.

The following commits are required for AMD Ryzen AI Max series support:

- [gregkh/linux@7f26af7](https://github.com/gregkh/linux/commit/7f26af7bf9b76c2c2a1a761aab5803e52be21eea)
- [gregkh/linux@7445db6](https://github.com/gregkh/linux/commit/7445db6a7d5a0242d8214582b480600b266cba9e)

These patches are available in the following minimum kernel versions:

- Ubuntu 24.04 Hardware Enablement (HWE): `6.17.0-19.19~24.04.2` or later
- Ubuntu 24.04 Original Equipment Manufacturer (OEM): `6.14.0-1018` or later
- All other distributions: Linux kernel `6.18.4` or later

The table below reflects compatibility for AMD-released pre-built ROCm
binaries only. Distributions that ship native ROCm packaging might
provide different support levels.

| ❌   | Unsupported combination           |
|------|-----------------------------------|
| ⚠️   | Unstable/experimental combination |
| ✅   | Stable and supported combination  |

| ROCm Release | Ubuntu 24.04 HWE (>= 6.17.0-19.19~24.04.2), Ubuntu 24.04 OEM (>= 6.14.0-1018) or Ubuntu 26.04 Generic | Other distributions >= 6.18.4 | Other distributions < 6.18.4 |
|-----------------------|-----------------------------------------------------------------------------------------------------------------------------------------------|---------------------------------|--------------------------------|
| 7.11.0 or 7.12.0 | ✅ | ✅ | ⚠️ |
| 7.9.0 or 7.10.0 | ❌ | ❌ | ⚠️ |
| 7.2.1, 7.2.2 or 7.2.3 | ✅ | ✅ | ⚠️ |
| 7.2.0 | ✅ | ✅ | ❌ |
| 7.1.x | ❌ | ❌ | ⚠️ |
| 6.4.x | ❌ | ❌ | ⚠️ |

#### NOTE
Ubuntu 24.04 HWE kernels earlier than `6.17.0-19.19~24.04.2` and Ubuntu
24.04 OEM kernels earlier than `6.14.0-1018` are not supported for
RDNA3.5 APUs.

The following distributions include the required fixes in their native
packaging, independent of AMD pre-built binaries:

- Fedora 43
- Ubuntu 26.04
- Arch Linux 2026.02.01

---

Source: https://rocm.docs.amd.com/en/latest/reference/system-optimization/rdna2.html

# AMD RDNA2 system optimization

The topic covers specific software requirements and processes needed to use
these GPUs for Single Root I/O Virtualization (SR-IOV) and machine learning
tasks.

The main purpose of this document is to help users utilize the RDNA™ 2 GPUs to
their full potential.

## Workstation workloads

Workstation workloads, much like those for HPC, have a unique set of
requirements: a blend of both graphics and compute, certification, stability,
and more.

For more information about RDNA2, see the [RDNA 2 Instruction Set Architecture: Reference Guide](https://www.amd.com/system/files/TechDocs/rdna2-shader-instruction-set-architecture.pdf).

## System settings

This section discusses system settings used to configure the system
for ROCm virtualization on RDNA2-based AMD Radeon™ PRO GPUs. Installing ROCm on
Bare Metal follows the routine [ROCm installation procedure](https://rocm.docs.amd.com/en/latest/install/rocm.html).

To enable ROCm virtualization on V620, set up Single Root I/O
Virtualization (SR-IOV) in the BIOS via settings found in
[System BIOS settings](#bios-settings). A tested configuration can be followed in
[Operating system settings](#os-settings).

#### ATTENTION
SR-IOV is supported on V620 and unsupported on W6800.

### System BIOS settings

| Advanced / North Bridge Configuration   | IOMMU           | Enabled   | Input-output Memory Management Unit   |
|-----------------------------------------|-----------------|-----------|---------------------------------------|
| Advanced / North Bridge Configuration   | ACS Enable      | Enabled   | Access Control Service                |
| Advanced / PCIe/PCI/PnP Configuration   | SR-IOV Support  | Enabled   | Single Root I/O Virtualization        |
| Advanced / ACPI settings                | PCI AER Support | Enabled   | Advanced Error Reporting              |

To set up the host, update SBIOS to version 1.2a.

### Operating system settings

| Server       | [SMC 4124](https://www.supermicro.com/en/Aplus/system/4U/4124/AS-4124GS-TNR.cfm) [AS -4124GS-TNR]   |
|--------------|-----------------------------------------------------------------------------------------------------|
| Host OS      | Ubuntu 20.04.3 LTS                                                                                  |
| Host Kernel  | 5.4.0-97-generic                                                                                    |
| CPU          | AMD EPYC 7552 48-Core Processor                                                                     |
| GPU          | RDNA2 V620 (D603GLXE)                                                                               |
| SBIOS        | Version SMC_r_1.2a                                                                                  |
| VBIOS        | 113-D603GLXE-077                                                                                    |
| Guest OS 1   | Ubuntu 20.04.5 LTS                                                                                  |
| Guest OS 2   | RHEL 9.0                                                                                            |
| GIM Driver   | gim-dkms_1.0.0.1234577_all                                                                          |
| VM CPU Cores | 32                                                                                                  |
| VM RAM       | 64 GB                                                                                               |

Install the following Kernel-based Virtual Machine (KVM) Hypervisor packages:

```shell
sudo apt-get -y install qemu-kvm qemu-utils  bridge-utils virt-manager  gir1.2-spiceclientgtk*  gir1.2-spice-client-gtk* libvirt-daemon-system dnsmasq-base
sudo virsh net-start default /*to enable Virtual network by default
```

Enable input-output memory management unit (IOMMU) in GRUB settings by adding the following line to `/etc/default/grub`:

```none
GRUB_CMDLINE_LINUX_DEFAULT="quiet splash" for AMD CPU
```

Update grub and reboot

```shell
sudo update=grub
sudo reboot
```

Install the GPU-IOV Module (GIM, where IOV is I/O Virtualization) driver and
follow the steps below.z

```shell
sudo dpkg -i <gim_driver>
sudo reboot
# Load Host Driver to Create 1VF
sudo modprobe gim vf_num=1
# Note: If GIM driver loaded successfully, we could see "gim info:(gim_init:213) *****Running GIM*****" in dmesg
lspci -d 1002:
```

Which should output something like:

```none
01:00.0 PCI bridge: Advanced Micro Devices, Inc. [AMD/ATI] Device 1478
02:00.0 PCI bridge: Advanced Micro Devices, Inc. [AMD/ATI] Device 1479
03:00.0 Display controller: Advanced Micro Devices, Inc. [AMD/ATI] Device 73a1
03:02.0 Display controller: Advanced Micro Devices, Inc. [AMD/ATI] Device 73ae → VF
```

### Guest OS installation

First, assign GPU virtual function (VF) to VM using the following steps.

1. Stop the VM.
2. Run `virt-manager`
3. In the **Virtual Machine Manager** GUI, select the **VM** and click **Open**.

![Virtual Machine Manager](reference/system-optimization/images/tuning014.png)
4. In the VM GUI, go to **Show Virtual Hardware Details > Add Hardware** to
   configure hardware.

![Show virtual hardware details](reference/system-optimization/images/tuning015.png)
5. Go to **Add Hardware > PCI Host Device > VF** and click **Finish**.

![VF Selection](reference/system-optimization/images/tuning016.png)

Then start the VM.

Lastly, install ROCm on the virtual machine (VM). For detailed instructions,
refer to the [Linux install guide](https://rocm.docs.amd.com/en/latest/install/rocm.html).

---

Source: https://rocm.docs.amd.com/en/latest/reference/system-optimization/common.html

# Common system settings

These topics discuss system-level GPU configuration that applies across AMD
hardware and workload types.

* [GPU isolation techniques](https://rocm.docs.amd.com/en/latest/reference/system-optimization/gpu-isolation.html) – Restrict application
  access to a subset of GPUs using environment variables, cgroups, or
  virtual machines.
* [BAR access limits](https://rocm.docs.amd.com/en/latest/reference/system-optimization/bar-access-limits.html) – Understand Base Address
  Register (BAR) physical addressing limits and how to handle peer-to-peer
  DMA access restrictions.

---

Source: https://rocm.docs.amd.com/en/latest/reference/system-optimization/gpu-isolation.html

# GPU isolation techniques

Restricting the access of applications to a subset of GPUs, aka isolating
GPUs allows users to hide GPU resources from programs. The programs by default
will only use the “exposed” GPUs ignoring other (hidden) GPUs in the system.

There are multiple ways to achieve isolation of GPUs in the ROCm software stack,
differing in which applications they apply to and the security they provide.
This page serves as an overview of the techniques.

## Environment variables

The runtimes in the ROCm software stack read these environment variables to
select the exposed or default device to present to applications using them.
See [ROCm environment variables](https://rocm.docs.amd.com/en/latest/reference/environment-variables/index.html#env-variables-reference) for more information.

Environment variables shouldn’t be used for isolating untrusted applications,
as an application can reset them before initializing the runtime.

### `ROCR_VISIBLE_DEVICES`

A list of device indices or s
that will be exposed to applications.

Runtime
: ROCm Software Runtime. Applies to all applications using the user mode ROCm
software stack.

```shell
export ROCR_VISIBLE_DEVICES="0,GPU-4b2c1a9f-8d3e-6f7a-b5c9-2e4d8a1f6c3b"
```

### `GPU_DEVICE_ORDINAL`

Devices indices exposed to OpenCL and HIP applications.

Runtime
: ROCm Compute Language Runtime (`ROCclr`). Applies to applications and runtimes
using the `ROCclr` abstraction layer including HIP and OpenCL applications.

```shell
export GPU_DEVICE_ORDINAL="0,2"
```

### `HIP_VISIBLE_DEVICES`

Device indices exposed to HIP applications.

Runtime: HIP runtime. Applies only to applications using HIP on the AMD platform.

```shell
export HIP_VISIBLE_DEVICES="0,2"
```

### `CUDA_VISIBLE_DEVICES`

Provided for CUDA compatibility, has the same effect as `HIP_VISIBLE_DEVICES`
on the AMD platform.

Runtime
: HIP or CUDA Runtime. Applies to HIP applications on the AMD or NVIDIA platform
and CUDA applications.

### `OMP_DEFAULT_DEVICE`

Default device used for OpenMP target offloading.

Runtime
: OpenMP Runtime. Applies only to applications using OpenMP offloading.

```shell
export OMP_DEFAULT_DEVICE="2"
```

## Docker

Docker uses Linux kernel namespaces to provide isolated environments for
applications. This isolation applies to most devices by default, including
GPUs. To access them in containers explicit access must be granted, please see
docker-access-gpus-in-container for details.
Specifically refer to [Restricting GPU access](https://rocm.docs.amd.com/projects/install-on-linux/en/latest/how-to/docker.html#docker-restrict-gpus) on exposing just a subset
of all GPUs.

Docker isolation is more secure than environment variables, and applies
to all programs that use the `amdgpu` kernel module interfaces.
Even programs that don’t use the ROCm runtime, like graphics applications
using OpenGL or Vulkan, can only access the GPUs exposed to the container.

## GPU passthrough to virtual machines

Virtual machines achieve the highest level of isolation, because even the kernel
of the virtual machine is isolated from the host. Devices physically installed
in the host system can be passed to the virtual machine using PCIe passthrough.
This allows for using the GPU with a different operating systems like a Windows
guest from a Linux host.

Setting up PCIe passthrough is specific to the hypervisor used. ROCm
supports [VMware ESXi](https://www.vmware.com/products/esxi-and-esx.html)
for select GPUs.

---

Source: https://rocm.docs.amd.com/en/latest/reference/system-optimization/bar-access-limits.html

# BAR access limits and physical addressing for AMD GPUs

Direct Memory Access (DMA) to PCIe devices using Base Address Registers (BARs) can be restricted due to physical addressing limits. These restrictions can result in data access failures between the system components. Peer-to-peer (P2P) DMA is used to access resources such as registers and memory between devices. PCIe devices need memory-mapped input/output (MMIO) space for DMA, and these MMIO spaces are defined in the PCIe BARs.

These BARs are a set of 32-bit or 64-bit registers that are used to define the resources that PCIe devices provide. The CPU and other system devices also use these to access the resources of the PCIe devices. P2P DMA only works when one device can directly access the local BAR memory of another. If the memory address of a BAR memory exceeds the physical addressing limit of a device, the device will not be able to access that BAR. This could be the device’s own BAR or the BAR of another device in the system.

If the BAR memory exceeds than the physical addressing limit of the device, the device will not be able to access the remote BAR.

To handle any BAR access issues that might occur, you need to be aware of the physical address limitations of the devices and understand the [BAR configuration of AMD GPUs](#bar-configuration). This information is important when setting up additional MMIO apertures for PCIe devices in the system’s physical address space.

## Handling physical address limitation

When a system boots, the system BIOS allocates the physical address space for the components in the system, including system memory and MMIO apertures. On modern 64-bit platforms, there are generally two or more MMIO apertures: one located below 4 GB of physical address space for 32-bit compatibility, and one or more above 4 GB for devices needing more space.

You can control the memory address of the high MMIO aperture from the system BIOS configuration options. This lets you configure the additional MMIO space to align with the physical addressing limit and allows P2P DMA between the devices. For example, if a PCIe device is limited to 44-bit of physical addressing, you should ensure that the MMIO aperture is set below 44-bit in the system physical address space.

There are two ways to handle this:

* Ensure that the high MMIO aperture is within the physical addressing limits of the devices in the system. For example, if the devices have a 44-bit physical addressing limit, set the `MMIO High Base` and `MMIO High size` options in the BIOS such that the aperture is within the 44-bit address range, and ensure that the `Above 4G Decoding` option is Enabled.
* Enable the Input-Output Memory Management Unit (IOMMU). When the IOMMU is enabled in non-passthrough mode, it will create a virtual I/O address space for each device on the system. It also ensures that all virtual addresses created in that space are within the physical addressing limits of the device. For more information on IOMMU, see [Input-Output Memory Management Unit (IOMMU)](https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/conceptual/iommu.html).

## BAR configuration for AMD GPUs

The following table shows how the BARs are configured for AMD GPUs.

| BAR Type         | Value                            | Description                                                                                                                                                                                                                                                                                                                                      |
|------------------|----------------------------------|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| BAR0-1 registers | 64-bit, Prefetchable, GPU memory | 8 GB or 16 GB depending on GPU. Set to less than 2^44 to support P2P access from other GPUs with a 44-bit physical address limit. Prefetchable memory enables faster read operation for high-performance computing (HPC) by fetching the contiguous data from the same data source even before requested as an anticipation of a future request. |
| BAR2-3 registers | 64-bit, Prefetchable, Doorbell   | Set to less than 2^44 to support P2P access from other GPUs with a 44-bit physical address limit. As a Doorbell BAR, it indicates to the GPU that a new operation is in its queue to be processed.                                                                                                                                               |
| BAR4 register    | Optional                         | Not a boot device                                                                                                                                                                                                                                                                                                                                |
| BAR5 register    | 32-bit, Non-prefetchable, MMIO   | Is set to less than 4 GB.                                                                                                                                                                                                                                                                                                                        |

### Example of BAR usage on AMD GPUs

Following is an example configuration of BARs set by the system BIOS on GFX8 GPUs with the 40-bit physical addressing limit:

```shell
11:00.0 Display controller: Advanced Micro Devices, Inc. [AMD/ATI] Fiji [Radeon R9 FURY / NANO
Series] (rev c1)

Subsystem: Advanced Micro Devices, Inc. [AMD/ATI] Device 0b35

Flags: bus master, fast devsel, latency 0, IRQ 119

Memory at bf40000000 (64-bit, prefetchable) [size=256M]

Memory at bf50000000 (64-bit, prefetchable) [size=2M]

I/O ports at 3000 [size=256]

Memory at c7400000 (32-bit, non-prefetchable) [size=256K]

Expansion ROM at c7440000 [disabled] [size=128K]
```

Details of the BARs configured in the example are:

**GPU Frame Buffer BAR:** `Memory at bf40000000 (64-bit, prefetchable) [size=256M]`

The size of the BAR in the example is 256 MB. Generally, it will be the size of the GPU memory (typically 4 GB+). Depending upon the physical address limit and generation of AMD GPUs, the BAR can be set below 2^40, 2^44, or 2^48.

**Doorbell BAR:** `Memory at bf50000000 (64-bit, prefetchable) [size=2M]`

The size of the BAR should typically be less than 10 MB for this generation of GPUs and has been set to 2 MB in the example. This BAR is placed less than 2^40 to allow peer-to-peer access from other generations of AMD GPUs.

**I/O BAR:** `I/O ports at 3000 [size=256]`

This is for legacy VGA and boot device support. Because the GPUs used are not connected to a display (VGA devices), this is not a concern, even if it isn’t set up in the system BIOS.

**MMIO BAR:** `Memory at c7400000 (32-bit, non-prefetchable) [size=256K]`

The AMD Driver requires this to access the configuration registers. Since the reminder of the BAR available is only 1 DWORD (32-bit), this is set less than 4 GB. In the example, it is fixed at 256 KB.

**Expansion ROM:** `Expansion ROM at c7440000 [disabled] [size=128K]`

This is required by the AMD Driver to access the GPU video-BIOS. In the example, it is fixed at 128 KB.

---

Source: https://rocm.docs.amd.com/en/latest/reference/precision-support.html

# Data types and precision support

This topic summarizes the data types supported on AMD GPUs and
ROCm libraries, along with corresponding [HIP](https://rocm.docs.amd.com/projects/HIP/en/latest/index.html) data types.

## Integral types

The signed and unsigned integral types supported by ROCm are listed in
the following table.

| Type name   | HIP type              | Description                         |
|-------------|-----------------------|-------------------------------------|
| int8        | `int8_t`, `uint8_t`   | A signed or unsigned 8-bit integer  |
| int16       | `int16_t`, `uint16_t` | A signed or unsigned 16-bit integer |
| int32       | `int32_t`, `uint32_t` | A signed or unsigned 32-bit integer |
| int64       | `int64_t`, `uint64_t` | A signed or unsigned 64-bit integer |

## Floating-point types

The floating-point types supported by ROCm are listed in the following table.

![Supported floating-point types](reference/images/floating-point-data-types.png)

| Type name | HIP type | Description |
|---------------|-----------------------------------------------------------------|-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| float4 (E2M1) | `__hip_fp4_e2m1` | A 4-bit floating-point number with **E2M1** bit layout, as described in [low precision floating point types page](https://rocm.docs.amd.com/projects/HIP/en/latest/reference/low_fp_types.html). |
| float6 (E3M2) | `__hip_fp6_e3m2` | A 6-bit floating-point number with **E3M2** bit layout, as described in [low precision floating point types page](https://rocm.docs.amd.com/projects/HIP/en/latest/reference/low_fp_types.html). |
| float6 (E2M3) | `__hip_fp6_e2m3` | A 6-bit floating-point number with **E2M3** bit layout, as described in [low precision floating point types page](https://rocm.docs.amd.com/projects/HIP/en/latest/reference/low_fp_types.html). |
| float8 (E4M3) | `__hip_fp8_e4m3_fnuz`, `__hip_fp8_e4m3` | An 8-bit floating-point number with **E4M3** bit layout, as described in [low precision floating point types page](https://rocm.docs.amd.com/projects/HIP/en/latest/reference/low_fp_types.html). The FNUZ variant has expanded range with no infinity or signed zero (NaN represented as negative zero), while the OCP variant follows the Open Compute Project specification. |
| float8 (E5M2) | `__hip_fp8_e5m2_fnuz`, `__hip_fp8_e5m2` | An 8-bit floating-point number with **E5M2** bit layout, as described in [low precision floating point types page](https://rocm.docs.amd.com/projects/HIP/en/latest/reference/low_fp_types.html). The FNUZ variant has expanded range with no infinity or signed zero (NaN represented as negative zero), while the OCP variant follows the Open Compute Project specification. |
| float16 | `half` | A 16-bit floating-point number that conforms to the IEEE 754-2008 half-precision storage format. |
| bfloat16 | `bfloat16` | A shortened 16-bit version of the IEEE 754 single-precision storage format. |
| tensorfloat32 | Not available | A floating-point number that occupies 32 bits or less of storage, providing improved range compared to half (16-bit) format, at (potentially) greater throughput than single-precision (32-bit) formats. |
| float32 | `float` | A 32-bit floating-point number that conforms to the IEEE 754 single-precision storage format. |
| float64 | `double` | A 64-bit floating-point number that conforms to the IEEE 754 double-precision storage format. |

#### NOTE
* The float8 and tensorfloat32 types are internal types used in calculations
  in Matrix Cores and can be stored in any type of the same size.
* CDNA3 natively supports FP8 FNUZ (E4M3 and E5M2), which differs from the customized
  FP8 format used with NVIDIA H100
  ([FP8 Formats for Deep Learning](https://arxiv.org/abs/2209.05433)).
* In some AMD documents and articles, float8 (E5M2) is referred to as bfloat8.
* The [low precision floating point types page](https://rocm.docs.amd.com/projects/HIP/en/latest/reference/low_fp_types.html)
  describes how to use these types in HIP with examples.

## Level of support definitions

In the following sections, icons represent the level of support. These icons,
described in the following table, are also used in the library data type support
pages.

| Icon   | Definition      |
|--------|-----------------|
| NA     | Not applicable  |
| ❌     | Not supported   |
| ⚠️     | Partial support |
| ✅     | Full support    |

#### NOTE
* Full support means that the type is supported natively or with hardware
  emulation.
* Native support means that the operations for that type are implemented in
  hardware. Types that are not natively supported are emulated with the
  available hardware. The performance of non-natively supported types can
  differ from the full instruction throughput rate. For example, 16-bit
  integer operations can be performed on the 32-bit integer ALUs at full rate;
  however, 64-bit integer operations might need several instructions on the
  32-bit integer ALUs.
* Any type can be emulated by software, but this page does not cover such
  cases.

## Data type support by hardware architecture

AMD’s GPU lineup spans multiple architecture generations:

* CDNA1 such as MI100
* CDNA2 such as MI210, MI250, and MI250X
* CDNA3 such as MI300A, MI300X, and MI325X
* CDNA4 such as MI350X and MI355X
* RDNA2 such as PRO W6800 and PRO V620
* RDNA3 such as RX 7900XT and RX 7900XTX
* RDNA4 such as RX 9070 and RX 9070XT

### HIP C++ type implementation support

The HIP C++ types available on different hardware platforms are listed in the
following table.

| HIP C++ Type          | CDNA1   | CDNA2   | CDNA3   | CDNA4   | RDNA2   | RDNA3   | RDNA4   |
|-----------------------|---------|---------|---------|---------|---------|---------|---------|
| `int8_t`, `uint8_t`   | ✅      | ✅      | ✅      | ✅      | ✅      | ✅      | ✅      |
| `int16_t`, `uint16_t` | ✅      | ✅      | ✅      | ✅      | ✅      | ✅      | ✅      |
| `int32_t`, `uint32_t` | ✅      | ✅      | ✅      | ✅      | ✅      | ✅      | ✅      |
| `int64_t`, `uint64_t` | ✅      | ✅      | ✅      | ✅      | ✅      | ✅      | ✅      |
| `__hip_fp4_e2m1`      | ❌      | ❌      | ❌      | ✅      | ❌      | ❌      | ❌      |
| `__hip_fp6_e2m3`      | ❌      | ❌      | ❌      | ✅      | ❌      | ❌      | ❌      |
| `__hip_fp6_e3m2`      | ❌      | ❌      | ❌      | ✅      | ❌      | ❌      | ❌      |
| `__hip_fp8_e4m3_fnuz` | ❌      | ❌      | ✅      | ❌      | ❌      | ❌      | ❌      |
| `__hip_fp8_e5m2_fnuz` | ❌      | ❌      | ✅      | ❌      | ❌      | ❌      | ❌      |
| `__hip_fp8_e4m3`      | ❌      | ❌      | ❌      | ✅      | ❌      | ❌      | ✅      |
| `__hip_fp8_e5m2`      | ❌      | ❌      | ❌      | ✅      | ❌      | ❌      | ✅      |
| `half`                | ✅      | ✅      | ✅      | ✅      | ✅      | ✅      | ✅      |
| `bfloat16`            | ✅      | ✅      | ✅      | ✅      | ✅      | ✅      | ✅      |
| `float`               | ✅      | ✅      | ✅      | ✅      | ✅      | ✅      | ✅      |
| `double`              | ✅      | ✅      | ✅      | ✅      | ✅      | ✅      | ✅      |

#### NOTE
Library support for specific data types is contingent upon hardware support.
Even if a ROCm library indicates support for a particular data type, that type
will only be fully functional if the underlying hardware architecture (as shown
in the table above) also supports it. For example, fp8 types are only available
on architectures shown with a checkmark in the relevant rows.

### Compute units support

The following table lists data type support for compute units.

**Integral types**

| Type name   | int8   | int16   | int32   | int64   |
|-------------|--------|---------|---------|---------|
| CDNA1       | ✅     | ✅      | ✅      | ✅      |
| CDNA2       | ✅     | ✅      | ✅      | ✅      |
| CDNA3       | ✅     | ✅      | ✅      | ✅      |
| CDNA4       | ✅     | ✅      | ✅      | ✅      |
| RDNA2       | ✅     | ✅      | ✅      | ✅      |
| RDNA3       | ✅     | ✅      | ✅      | ✅      |
| RDNA4       | ✅     | ✅      | ✅      | ✅      |

**Low precision floating-point types**

| Type name   | float4   | float6 (E2M3)   | float6 (E3M2)   | float8 (E4M3)   | float8 (E5M2)   |
|-------------|----------|-----------------|-----------------|-----------------|-----------------|
| CDNA1       | ❌       | ❌              | ❌              | ❌              | ❌              |
| CDNA2       | ❌       | ❌              | ❌              | ❌              | ❌              |
| CDNA3       | ❌       | ❌              | ❌              | ❌              | ❌              |
| CDNA4       | ❌       | ❌              | ❌              | ❌              | ❌              |
| RDNA2       | ❌       | ❌              | ❌              | ❌              | ❌              |
| RDNA3       | ❌       | ❌              | ❌              | ❌              | ❌              |
| RDNA4       | ❌       | ❌              | ❌              | ❌              | ❌              |

**High precision floating-point types**

| Type name   | float16   | bfloat16   | tensorfloat32   | float32   | float64   |
|-------------|-----------|------------|-----------------|-----------|-----------|
| CDNA1       | ✅        | ✅         | ❌              | ✅        | ✅        |
| CDNA2       | ✅        | ✅         | ❌              | ✅        | ✅        |
| CDNA3       | ✅        | ✅         | ❌              | ✅        | ✅        |
| CDNA4       | ✅        | ✅         | ❌              | ✅        | ✅        |
| RDNA2       | ✅        | ✅         | ❌              | ✅        | ✅        |
| RDNA3       | ✅        | ✅         | ❌              | ✅        | ✅        |
| RDNA4       | ✅        | ✅         | ❌              | ✅        | ✅        |

### Matrix core support

The following table lists data type support for AMD GPU matrix cores.

**Integral types**

| Type name   | int8   | int16   | int32   | int64   |
|-------------|--------|---------|---------|---------|
| CDNA1       | ✅     | ❌      | ❌      | ❌      |
| CDNA2       | ✅     | ❌      | ❌      | ❌      |
| CDNA3       | ✅     | ❌      | ❌      | ❌      |
| CDNA4       | ✅     | ❌      | ❌      | ❌      |
| RDNA2       | ✅     | ❌      | ❌      | ❌      |
| RDNA3       | ✅     | ❌      | ❌      | ❌      |
| RDNA4       | ✅     | ❌      | ❌      | ❌      |

**Low precision floating-point types**

| Type name   | float4   | float6 (E2M3)   | float6 (E3M2)   | float8 (E4M3)   | float8 (E5M2)   |
|-------------|----------|-----------------|-----------------|-----------------|-----------------|
| CDNA1       | ❌       | ❌              | ❌              | ❌              | ❌              |
| CDNA2       | ❌       | ❌              | ❌              | ❌              | ❌              |
| CDNA3       | ❌       | ❌              | ❌              | ✅              | ✅              |
| CDNA4       | ✅       | ✅              | ✅              | ✅              | ✅              |
| RDNA2       | ❌       | ❌              | ❌              | ❌              | ❌              |
| RDNA3       | ❌       | ❌              | ❌              | ❌              | ❌              |
| RDNA4       | ❌       | ❌              | ❌              | ✅              | ✅              |

**High precision floating-point types**

| Type name   | float16   | bfloat16   | tensorfloat32   | float32   | float64   |
|-------------|-----------|------------|-----------------|-----------|-----------|
| CDNA1       | ✅        | ✅         | ❌              | ✅        | ❌        |
| CDNA2       | ✅        | ✅         | ❌              | ✅        | ✅        |
| CDNA3       | ✅        | ✅         | ✅              | ✅        | ✅        |
| CDNA4       | ✅        | ✅         | ✅              | ✅        | ✅        |
| RDNA2       | ✅        | ✅         | ❌              | ❌        | ❌        |
| RDNA3       | ✅        | ✅         | ❌              | ❌        | ❌        |
| RDNA4       | ✅        | ✅         | ❌              | ❌        | ❌        |

### Atomic operations support

The following table lists which data types are supported for atomic
operations on AMD GPUs. The atomics operation type behavior is affected by the
memory locations, memory granularity, or scope of operations. For detailed
various support of atomic read-modify-write (atomicRMW) operations collected on
the [Hardware atomics operation support](https://rocm.docs.amd.com/en/latest/reference/gpu-atomics-operation.html#hw-atomics-operation-support)
page.

**Integral types**

| Type name   | int8   | int16   | int32   | int64   |
|-------------|--------|---------|---------|---------|
| CDNA 1      | ❌     | ❌      | ✅      | ✅      |
| CDNA 2      | ❌     | ❌      | ✅      | ✅      |
| CDNA 3      | ❌     | ❌      | ✅      | ✅      |
| RDNA 3      | ❌     | ❌      | ✅      | ✅      |
| RDNA 4      | ❌     | ❌      | ✅      | ✅      |

**Low precision floating-point types**

| Type name   | float4   | float6 (E2M3)   | float6 (E3M2)   | float8 (E4M3)   | float8 (E5M2)   |
|-------------|----------|-----------------|-----------------|-----------------|-----------------|
| CDNA 1      | ❌       | ❌              | ❌              | ❌              | ❌              |
| CDNA 2      | ❌       | ❌              | ❌              | ❌              | ❌              |
| CDNA 3      | ❌       | ❌              | ❌              | ❌              | ❌              |
| CDNA 4      | ❌       | ❌              | ❌              | ❌              | ❌              |
| RDNA 2      | ❌       | ❌              | ❌              | ❌              | ❌              |
| RDNA 3      | ❌       | ❌              | ❌              | ❌              | ❌              |
| RDNA 4      | ❌       | ❌              | ❌              | ❌              | ❌              |

**High precision floating-point types**

| Type name   | 2 x float16   | 2 x bfloat16   | tensorfloat32   | float32   | float64   |
|-------------|---------------|----------------|-----------------|-----------|-----------|
| CDNA 1      | ✅            | ✅             | ❌              | ✅        | ❌        |
| CDNA 2      | ✅            | ✅             | ❌              | ✅        | ✅        |
| CDNA 3      | ✅            | ✅             | ❌              | ✅        | ✅        |
| CDNA 4      | ✅            | ✅             | ❌              | ✅        | ✅        |
| RDNA 2      | ❌            | ❌             | ❌              | ✅        | ❌        |
| RDNA 3      | ❌            | ❌             | ❌              | ✅        | ❌        |
| RDNA 4      | ✅            | ✅             | ❌              | ✅        | ❌        |

#### NOTE
You can emulate atomic operations using software for cases that are not
natively supported. Software-emulated atomic operations have a high negative
performance impact when they frequently access the same memory address.

## Data type support in ROCm libraries

ROCm library support for int8, float8 (E4M3), float8 (E5M2), int16, float16,
bfloat16, int32, tensorfloat32, float32, int64, and float64 is listed in the
following tables.

### Libraries input/output type support

The following tables list ROCm library support for specific input and output
data types. Select a library from the below table to view the supported data
types.

For more information, please visit [Composable Kernel](https://rocm.docs.amd.com/projects/composable_kernel/en/latest/reference/Composable_Kernel_data_type_support.html).

| Data Type     | Support   |
|---------------|-----------|
| int8          | ✅        |
| int32         | ✅        |
| float4 (E2M1) | ✅        |
| float6 (E2M3) | ✅        |
| float6 (E3M2) | ✅        |
| float8 (E4M3) | ✅        |
| float8 (E5M2) | ✅        |
| float16       | ✅        |
| bfloat16      | ✅        |
| float32       | ✅        |
| float64       | ✅        |

For more information, please visit MIGraphX.

| Data Type     | Support   |
|---------------|-----------|
| int8          | ✅        |
| int16         | ✅        |
| int32         | ✅        |
| int64         | ✅        |
| float8 (E4M3) | ✅        |
| float8 (E5M2) | ✅        |
| float16       | ✅        |
| bfloat16      | ✅        |
| float32       | ✅        |
| float64       | ✅        |

For more information, please visit [MIOpen](https://rocm.docs.amd.com/projects/MIOpen/en/latest/reference/data_type_support.html).

| Data Type     | Support   |
|---------------|-----------|
| int8          | ⚠️        |
| int32         | ⚠️        |
| float8 (E4M3) | ⚠️        |
| float8 (E5M2) | ⚠️        |
| float16       | ✅        |
| bfloat16      | ⚠️        |
| float32       | ✅        |
| float64       | ⚠️        |

For more information, please visit [RCCL](https://rocm.docs.amd.com/projects/rccl/en/latest/api-reference/data-type-support.html).

| Data Type     | Support   |
|---------------|-----------|
| int8          | ✅        |
| int32         | ✅        |
| int64         | ✅        |
| float8 (E4M3) | ✅        |
| float8 (E5M2) | ✅        |
| float16       | ✅        |
| bfloat16      | ✅        |
| float32       | ✅        |
| float64       | ✅        |

For more information, please visit [hipBLAS](https://rocm.docs.amd.com/projects/hipBLAS/en/latest/reference/data-type-support.html).

| Data Type   | Support   |
|-------------|-----------|
| float16     | ⚠️        |
| bfloat16    | ⚠️        |
| float32     | ✅        |
| float64     | ✅        |
| cfloat32    | ✅        |
| cfloat64    | ✅        |

For more information, please visit [hipBLASLt](https://rocm.docs.amd.com/projects/hipBLASLt/en/latest/reference/data-type-support.html).

| Data Type     | Support    |
|---------------|------------|
| int8          | ✅         |
| float4 (E2M1) | Input only |
| float8 (E4M3) | ✅         |
| float8 (E5M2) | ✅         |
| float16       | ✅         |
| bfloat16      | ✅         |
| float32       | ✅         |

For more information, please visit [hipFFT](https://rocm.docs.amd.com/projects/hipFFT/en/latest/reference/data-type-support.html).

| Data Type   | Support   |
|-------------|-----------|
| float16     | ⚠️        |
| float32     | ✅        |
| float64     | ✅        |
| cfloat16    | ⚠️        |
| cfloat32    | ✅        |
| cfloat64    | ✅        |

For more information, please visit [hipRAND](https://rocm.docs.amd.com/projects/hipRAND/en/latest/api-reference/data-type-support.html).

| Data Type   | Support     |
|-------------|-------------|
| int8        | Output only |
| int16       | Output only |
| int32       | Output only |
| int64       | Output only |
| float16     | Output only |
| float32     | Output only |
| float64     | Output only |

For more information, please visit [hipSOLVER](https://rocm.docs.amd.com/projects/hipSOLVER/en/latest/reference/data-type-support.html).

| Data Type   | Support   |
|-------------|-----------|
| float32     | ✅        |
| float64     | ✅        |
| cfloat32    | ✅        |
| cfloat64    | ✅        |

For more information, please visit [hipSPARSE](https://rocm.docs.amd.com/projects/hipSPARSE/en/latest/reference/data-type-support.html).

| Data Type   | Support   |
|-------------|-----------|
| float32     | ✅        |
| float64     | ✅        |
| cfloat32    | ✅        |
| cfloat64    | ✅        |

For more information, please visit [hipSPARSELt](https://rocm.docs.amd.com/projects/hipSPARSELt/en/latest/reference/data-type-support.html).

| Data Type     | Support     |
|---------------|-------------|
| int8          | ✅          |
| int32         | Output only |
| float8 (E4M3) | Input only  |
| float8 (E5M2) | Input only  |
| float16       | ✅          |
| bfloat16      | ✅          |
| float32       | Output only |

For more information, please visit [rocBLAS](https://rocm.docs.amd.com/projects/rocBLAS/en/latest/reference/data-type-support.html).

| Data Type   | Support   |
|-------------|-----------|
| int8        | ⚠️        |
| int32       | ⚠️        |
| float16     | ⚠️        |
| bfloat16    | ⚠️        |
| float32     | ✅        |
| float64     | ✅        |
| cfloat32    | ✅        |
| cfloat64    | ✅        |

For more information, please visit [rocFFT](https://rocm.docs.amd.com/projects/rocFFT/en/latest/reference/data-type-support.html).

| Data Type   | Support   |
|-------------|-----------|
| float16     | ✅        |
| float32     | ✅        |
| float64     | ✅        |
| cfloat16    | ✅        |
| cfloat32    | ✅        |
| cfloat64    | ✅        |

For more information, please visit [rocRAND](https://rocm.docs.amd.com/projects/rocRAND/en/latest/api-reference/data-type-support.html).

For more information, please visit [rocSOLVER](https://rocm.docs.amd.com/projects/rocSOLVER/en/latest/reference/data_type_support.html).

| Data Type   | Support   |
|-------------|-----------|
| float32     | ✅        |
| float64     | ✅        |
| cfloat32    | ✅        |
| cfloat64    | ✅        |

For more information, please visit [rocSPARSE](https://rocm.docs.amd.com/projects/rocSPARSE/en/latest/reference/data_type_support.html).

| Data Type   | Support   |
|-------------|-----------|
| float32     | ✅        |
| float64     | ✅        |
| cfloat32    | ✅        |
| cfloat64    | ✅        |

For more information, please visit rocWMMA.

| Data Type     | Support     |
|---------------|-------------|
| int8          | ✅          |
| int32         | Output only |
| float8 (E4M3) | Input only  |
| float8 (E5M2) | Input only  |
| float16       | ✅          |
| bfloat16      | ✅          |
| tensorfloat32 | ✅          |
| float32       | ✅          |
| float64       | ✅          |

For more information, please visit [Tensile](https://rocm.docs.amd.com/projects/Tensile/en/latest/src/reference/precision-support.html).

| Data Type     | Support   |
|---------------|-----------|
| int8          | ✅        |
| int32         | ✅        |
| float8 (E4M3) | ✅        |
| float8 (E5M2) | ✅        |
| float16       | ✅        |
| bfloat16      | ✅        |
| tensorfloat32 | ✅        |
| float32       | ✅        |
| float64       | ✅        |
| cfloat32      | ✅        |
| cfloat64      | ✅        |

For more information, please visit [hipCUB](https://rocm.docs.amd.com/projects/hipCUB/en/latest/api-reference/data-type-support.html).

| Data Type   | Support   |
|-------------|-----------|
| int8        | ✅        |
| int16       | ✅        |
| int32       | ✅        |
| int64       | ✅        |
| float16     | ✅        |
| bfloat16    | ✅        |
| float32     | ✅        |
| float64     | ✅        |

For more information, please visit hipTensor.

| Data Type   | Support   |
|-------------|-----------|
| float16     | ✅        |
| bfloat16    | ✅        |
| float32     | ✅        |
| float64     | ⚠️        |
| cfloat32    | ⚠️        |
| cfloat64    | ⚠️        |

For more information, please visit rocPRIM.

For more information, please visit [rocThrust](https://rocm.docs.amd.com/projects/rocThrust/en/latest/reference/rocThrust-data-type-support.html).

| Data Type   | Support   |
|-------------|-----------|
| int8        | ✅        |
| int16       | ✅        |
| int32       | ✅        |
| int64       | ✅        |
| float16     | ⚠️        |
| bfloat16    | ⚠️        |
| float32     | ✅        |
| float64     | ✅        |

#### NOTE
The meaning of partial support depends on the library. Please refer to the individual
libraries’ documentation for more information.

#### NOTE
As random number generation libraries, rocRAND and hipRAND only specify output
data types for the random values they generate, with no need for input data
types.

#### NOTE
hipBLASLt supports additional data types as internal compute types, which may
differ from the supported input/output types shown in the tables above. While
TensorFloat32 is not supported as an input or output type in this library, it
is available as an internal compute type. For complete details on supported
compute types, refer to the [hipBLASLt](https://rocm.docs.amd.com/projects/hipBLASLt/en/latest/reference/data-type-support.html)
documentation.

### hipDataType enumeration

The `hipDataType` enumeration defines data precision types and is primarily
used when the data reference itself does not include type information, such as
in `void*` pointers. This enumeration is mainly utilized in BLAS libraries.
The HIP type equivalents of the `hipDataType` enumeration are listed in the
following table with descriptions and values.

| hipDataType          | HIP type              |   Value | Description                                                 |
|----------------------|-----------------------|---------|-------------------------------------------------------------|
| `HIP_R_8I`           | `int8_t`              |       3 | 8-bit real signed integer.                                  |
| `HIP_R_8U`           | `uint8_t`             |       8 | 8-bit real unsigned integer.                                |
| `HIP_R_16I`          | `int16_t`             |      20 | 16-bit real signed integer.                                 |
| `HIP_R_16U`          | `uint16_t`            |      22 | 16-bit real unsigned integer.                               |
| `HIP_R_32I`          | `int32_t`             |      10 | 32-bit real signed integer.                                 |
| `HIP_R_32U`          | `uint32_t`            |      12 | 32-bit real unsigned integer.                               |
| `HIP_R_32F`          | `float`               |       0 | 32-bit real single precision floating-point.                |
| `HIP_R_64F`          | `double`              |       1 | 64-bit real double precision floating-point.                |
| `HIP_R_16F`          | `half`                |       2 | 16-bit real half precision floating-point.                  |
| `HIP_R_16BF`         | `bfloat16`            |      14 | 16-bit real bfloat16 precision floating-point.              |
| `HIP_R_8F_E4M3`      | `__hip_fp8_e4m3`      |      28 | 8-bit real float8 precision floating-point (OCP version).   |
| `HIP_R_8F_E5M2`      | `__hip_fp8_e5m2`      |      29 | 8-bit real bfloat8 precision floating-point (OCP version).  |
| `HIP_R_6F_E2M3`      | `__hip_fp6_e2m3`      |      31 | 6-bit real float6 precision floating-point.                 |
| `HIP_R_6F_E3M2`      | `__hip_fp6_e3m2`      |      32 | 6-bit real bfloat6 precision floating-point.                |
| `HIP_R_4F_E2M1`      | `__hip_fp4_e2m1`      |      33 | 4-bit real float4 precision floating-point.                 |
| `HIP_R_8F_E4M3_FNUZ` | `__hip_fp8_e4m3_fnuz` |    1000 | 8-bit real float8 precision floating-point (FNUZ version).  |
| `HIP_R_8F_E5M2_FNUZ` | `__hip_fp8_e5m2_fnuz` |    1001 | 8-bit real bfloat8 precision floating-point (FNUZ version). |

The full list of the `hipDataType` enumeration listed in [library_types.h](https://github.com/ROCm/hip/blob/amd-staging/include/hip/library_types.h).

---

Source: https://rocm.docs.amd.com/en/latest/reference/graph-safe-support.html

# Graph-safe support for ROCm libraries

HIP graph-safe libraries operate safely in HIP execution graphs.
[HIP graphs](https://rocm.docs.amd.com/projects/HIP/en/latest/how-to/hip_runtime_api/hipgraph.html#how-to-hip-graph) are an alternative way of executing tasks on a GPU
that can provide performance benefits over launching kernels using the standard
method via streams.

Functions and routines from graph-safe libraries shouldn’t result in issues like
race conditions, deadlocks, or unintended dependencies.

The following table shows whether a ROCm library is graph-safe.

| ROCm library                                                   | Graph safe support                                                                                                                   |
|----------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------------------------------|
| [Composable Kernel](https://github.com/ROCm/composable_kernel) | ❌                                                                                                                                   |
| [hipBLAS](https://github.com/ROCm/hipBLAS)                     | ✅                                                                                                                                   |
| [hipBLASLt](https://github.com/ROCm/hipBLASLt)                 | ⚠️                                                                                                                                   |
| [hipCUB](https://github.com/ROCm/hipCUB)                       | ✅                                                                                                                                   |
| [hipFFT](https://github.com/ROCm/hipFFT)                       | ✅ (see [details](https://rocm.docs.amd.com/projects/hipFFT/en/latest/reference/hipfft-api-usage.html#hip-graph-support-for-hipfft)) |
| [hipRAND](https://github.com/ROCm/hipRAND)                     | ✅                                                                                                                                   |
| [hipSOLVER](https://github.com/ROCm/hipSOLVER)                 | ⚠️ (experimental)                                                                                                                    |
| [hipSPARSE](https://github.com/ROCm/hipSPARSE)                 | ✅                                                                                                                                   |
| [hipSPARSELt](https://github.com/ROCm/hipSPARSELt)             | ⚠️ (experimental)                                                                                                                    |
| [hipTensor](https://github.com/ROCm/hipTensor)                 | ❌                                                                                                                                   |
| [MIOpen](https://github.com/ROCm/MIOpen)                       | ❌                                                                                                                                   |
| [RCCL](https://github.com/ROCm/rccl)                           | ✅                                                                                                                                   |
| [rocAL](https://github.com/ROCm/rocAL)                         | ❌                                                                                                                                   |
| [rocALUTION](https://github.com/ROCm/rocALUTION)               | ❌                                                                                                                                   |
| [rocBLAS](https://github.com/ROCm/rocBLAS)                     | ✅ (see [details](https://rocm.docs.amd.com/projects/rocBLAS/en/latest/reference/beta-features.html))                                |
| [rocDecode](https://github.com/ROCm/rocDecode)                 | ❌                                                                                                                                   |
| [rocFFT](https://github.com/ROCm/rocFFT)                       | ✅ (see [details](https://rocm.docs.amd.com/projects/rocFFT/en/latest/reference/api.html#hip-graph-support-for-rocfft))              |
| [rocHPCG](https://github.com/ROCm/rocHPCG)                     | ❌                                                                                                                                   |
| [rocJPEG](https://github.com/ROCm/rocJPEG)                     | ❌                                                                                                                                   |
| [rocPRIM](https://github.com/ROCm/rocPRIM)                     | ✅                                                                                                                                   |
| [rocRAND](https://github.com/ROCm/rocRAND)                     | ✅                                                                                                                                   |
| [rocSOLVER](https://github.com/ROCm/rocSOLVER)                 | ⚠️ (experimental)                                                                                                                    |
| [rocSPARSE](https://github.com/ROCm/rocSPARSE)                 | ⚠️ (experimental)                                                                                                                    |
| [rocThrust](https://github.com/ROCm/rocThrust)                 | ❌                                                                                                                                   |
| [rocWMMA](https://github.com/ROCm/rocWMMA)                     | ❌                                                                                                                                   |
| [RPP](https://github.com/ROCm/rpp)                             | ⚠️                                                                                                                                   |
| [Tensile](https://github.com/ROCm/Tensile)                     | ✅                                                                                                                                   |

✅: full support

⚠️: partial support

❌: not supported

---

Source: https://rocm.docs.amd.com/en/latest/reference/environment-variables/index.html

# ROCm environment variables

ROCm provides a set of environment variables that allow users to configure and optimize their development
and runtime experience. These variables define key settings such as installation paths, platform selection,
and runtime behavior for applications running on AMD accelerators and GPUs.

This page outlines commonly used environment variables across different components of the ROCm software stack,
including HIP and ROCR-Runtime. Understanding these variables can help streamline software development and
execution in ROCm-based environments.

## HIP environment variables

The following tables list the HIP environment variables.

### GPU isolation variables

Restricting the access of applications to a subset of GPUs, also known as GPU
isolation, allows users to hide GPU resources from programs. The GPU isolation
environment variables in HIP are collected in the following table.

| **Environment variable** | **Links** | **Value** |
|------------------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------|-------------------------------------------------------|
| `ROCR_VISIBLE_DEVICES` A list of device indices or UUIDs that will be exposed to applications. | GPU isolation, Setting the number of compute units | Example: `0,GPU-4b2c1a9f-8d3e-6f7a-b5c9-2e4d8a1f6c3b` |
| `GPU_DEVICE_ORDINAL` Devices indices exposed to OpenCL and HIP applications. | GPU isolation | Example: `0,2` |
| `HIP_VISIBLE_DEVICES` or `CUDA_VISIBLE_DEVICES` Device indices exposed to HIP applications. | GPU isolation, [HIP debugging](https://rocm.docs.amd.com/projects/HIP/en/latest/how-to/debugging.html) | Example: `0,2` |

#### NOTE
### Recommendation

* On Linux, use `ROCR_VISIBLE_DEVICES`.
* On Windows, use `HIP_VISIBLE_DEVICES`.
* For portability across different vendors, use `CUDA_VISIBLE_DEVICES`.

### Profiling variables

The profiling environment variables in HIP are collected in the following table. For
more information, check setting the number of CUs page.

| **Environment variable** | **Value** |
|-----------------------------------------------------------------------------------------------------------------------------------------------------------------------|----------------------------------------------|
| `HSA_CU_MASK` Sets the mask on a lower level of queue creation in the driver, this mask will also be set for queues being profiled. | Example: `1:0-8` |
| `ROC_GLOBAL_CU_MASK` Sets the mask on queues created by the HIP or the OpenCL runtimes, this mask will also be set for queues being profiled. | Example: `0xf`, enables only 4 CUs |
| `HIP_FORCE_QUEUE_PROFILING` Used to run the app as if it were run in rocprof. Forces command queue profiling on by default. | 0: Disable 1: Enable |

### Debug variables

The debugging environment variables in HIP are collected in the following table. For
more information, check [Logging HIP activity](https://rocm.docs.amd.com/projects/HIP/en/latest/how-to/logging.html), [Debugging with HIP](https://rocm.docs.amd.com/projects/HIP/en/latest/how-to/debugging.html)
and GPU isolation.

| **Environment variable** | **Default value** | **Value** |
|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|---------------------|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| `AMD_LOG_LEVEL` Enables HIP log on various level. | `0` | 0: Disable log. 1: Enables error logs. 2: Enables warning logs next to lower-level logs. 3: Enables information logs next to lower-level logs. 4: Enables debug logs next to lower-level logs. 5: Enables debug extra logs next to lower-level logs. |
| `AMD_LOG_LEVEL_FILE` Sets output file for `AMD_LOG_LEVEL`. | stderr output | |
| `AMD_LOG_MASK` Specifies HIP log filters. Here is the [complete list of log masks](https://github.com/ROCm/rocm-systems/blob/develop/projects/clr/rocclr/utils/debug.hpp#L48). | `0x7FFFFFFF` | 0x1: Log API calls. 0x2: Kernel and copy commands and barriers. 0x4: Synchronization and waiting for commands to finish. 0x8: Decode and display AQL packets. 0x10: Queue commands and queue contents. 0x20: Signal creation, allocation, pool. 0x40: Locks and thread-safety code. 0x80: Kernel creations and arguments, etc. 0x100: Copy debug. 0x200: Detailed copy debug. 0x400: Resource allocation, performance-impacting events. 0x800: Initialization and shutdown. 0x1000: Misc debug, not yet classified. 0x2000: Show raw bytes of AQL packet. 0x4000: Show code creation debug. 0x8000: More detailed command info, including barrier commands. 0x10000: Log message location. 0x20000: Memory allocation. 0x40000: Memory pool allocation, including memory in graphs. 0x80000: Timestamp details. 0x100000: Comgr path information print. 0xFFFFFFFF: Log always even mask flag is zero. |
| `HIP_FORCE_DEV_KERNARG` Forces kernel arguments to be stored in device memory to reduce latency. Can improve performance by 2-3 µs for some kernels. | `1` | 0: Disable 1: Enable |
| `HIP_LAUNCH_BLOCKING` Used for serialization on kernel execution. | `0` | 0: Disable. Kernel executes normally. 1: Enable. Serializes kernel enqueue, behaves the same as `AMD_SERIALIZE_KERNEL`. |
| `HIP_VISIBLE_DEVICES` (or `CUDA_VISIBLE_DEVICES`) Only devices whose index is present in the sequence are visible to HIP | Unset by default. | 0,1,2: Depending on the number of devices on the system. |
| `GPU_DUMP_CODE_OBJECT` Dump code object. | `0` | 0: Disable 1: Enable |
| `AMD_SERIALIZE_KERNEL` Serialize kernel enqueue. | `0` | 0: Disable 1: Wait for completion before enqueue. 2: Wait for completion after enqueue. 3: Both |
| `AMD_SERIALIZE_COPY` Serialize copies | `0` | 0: Disable 1: Wait for completion before enqueue. 2: Wait for completion after enqueue. 3: Both |
| `GPU_MAX_HW_QUEUES` The maximum number of hardware queues allocated per device. | `4` | The recommended maximum value for this setting is 4. The variable controls how many independent hardware queues HIP runtime can create per process, per device. If an application allocates more HIP streams than this number, then HIP runtime reuses the same hardware queues for the new streams in a round-robin manner. Note that this maximum number does not apply to hardware queues that are created for CU-masked HIP streams, or cooperative queues for HIP Cooperative Groups (single queue per device). |

### Memory management related variables

The memory management related environment variables in HIP are collected in the
following table. The `HIP_HOST_COHERENT` variable linked at the following
pages:

- [Coherence control](https://rocm.docs.amd.com/projects/HIP/en/latest/how-to/hip_runtime_api/memory_management/coherence_control.html#coherence-control)
- [Memory allocation flags](https://rocm.docs.amd.com/projects/HIP/en/latest/how-to/hip_runtime_api/memory_management/host_memory.html#memory-allocation-flags)

### Other useful variables

The following table lists environment variables that are useful but relate to
different features in HIP.

## ROCR-Runtime environment variables

The following table lists the [ROCR-Runtime](https://rocm.docs.amd.com/projects/ROCR-Runtime/en/latest/index.html)
environment variables:

| Environment variable | Default value | Value |
|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|--------------------|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| `ROCR_VISIBLE_DEVICES` Specifies a list of device indices or UUIDs to be exposed to the applications. | None | `0,GPU-DEADBEEFDEADBEEF` |
| `HSA_NO_SCRATCH_RECLAIM` Controls whether scratch memory allocations are permanently assigned to queues or can be reclaimed based on usage thresholds. | `0` | 0: Disable. When dispatches need scratch memory that are lower than the threshold, the memory will be permanently assigned to the queue. For dispatches that exceed the threshold, a scratch-use-once mechanism will be used, resulting in the memory to be unassigned after the dispatch. 1: Enable. If a kernel dispatch needs scratch memory, runtime will allocate and permanently assign device memory to the queue handling the dispatch, even if the amount of scratch memory exceeds the default threshold. This memory will not be available to other queues or processes until this process exits. |
| `HSA_SCRATCH_SINGLE_LIMIT` Specifies the threshold for the amount of scratch memory allocated and reclaimed in kernel dispatches. Enabling `HSA_NO_SCRATCH_RECLAIM` circumvents `HSA_SCRATCH_SINGLE_LIMIT`, and treats `HSA_SCRATCH_SINGLE_LIMIT` as the maximum value. **NOTE:** In the 7.0 release the developer can use the HIP enumerator `hipExtLimitScratchCurrent` to programmatically change the default scratch memory allocation size. For more information, see [Global enums and defines](https://rocm.docs.amd.com/projects/HIP/en/latest/reference/hip_runtime_api/global_defines_enums_structs_files.html). | `146800640` | 0 to 4GB per XCC |
| `HSA_SCRATCH_SINGLE_LIMIT_ASYNC` On GPUs that support asynchronous scratch reclaim, this variable is used instead of `HSA_SCRATCH_SINGLE_LIMIT` to specify the threshold for scratch memory allocation. | `3221225472` (3GB) | 0 to 4GB per XCC |
| `HSA_ENABLE_SCRATCH_ASYNC_RECLAIM` Controls asynchronous scratch memory reclamation on supported GPUs. When enabled, if a device memory allocation fails, ROCr will attempt to reclaim scratch memory assigned to all queues and retry the allocation. | `1` | 0: Disable asynchronous scratch reclaim. 1: Enable asynchronous scratch reclaim on supported GPUs. |
| `HSA_XNACK` Enables XNACK. | None | 1: Enable |
| `HSA_CU_MASK` Sets the mask on a lower level of queue creation in the driver. This mask is also applied to the queues being profiled. | None | `1:0-8` |
| `HSA_ENABLE_SDMA` Enables the use of direct memory access (DMA) engines in all copy directions (Host-to-Device, Device-to-Host, Device-to-Device), when using any of the following APIs: `hsa_memory_copy`, `hsa_amd_memory_fill`, `hsa_amd_memory_async_copy`, `hsa_amd_memory_async_copy_on_engine`. | `1` | 0: Disable 1: Enable |
| `HSA_ENABLE_PEER_SDMA` **Note**: This environment variable is ignored if `HSA_ENABLE_SDMA` is set to 0. Enables the use of DMA engines for Device-to-Device copies, when using any of the following APIs: `hsa_memory_copy`, `hsa_amd_memory_async_copy`, `hsa_amd_memory_async_copy_on_engine`. | `1` | 0: Disable 1: Enable |
| `HSA_ENABLE_MWAITX` When mwaitx is enabled, on AMD CPUs, runtime will hint to the CPU to go into lower power-states when doing busy loops by using the mwaitx instruction. | `1` | 0: Disable 1: Enable |
| `HSA_OVERRIDE_CPU_AFFINITY_DEBUG` Controls whether ROCm helper threads inherit the parent process’s CPU affinity mask. | `1` | 0: Enable inheritance. Helper threads use the parent process’s core affinity mask, which should be set with enough cores for all threads. 1: Disable inheritance. Helper threads spawn on all available cores, ignoring the parent’s affinity settings, which may affect performance in certain environments. |
| `HSA_ENABLE_DEBUG` Enables additional debug information and validation in the runtime. | `0` | 0: Disable debug mode. 1: Enable debug mode with additional validation and logging. |
| `HSA_HOTSWAP_DISABLE` Disables HotSwap code object rewriting. HotSwap is enabled by default on supported GPU and ASIC revisions. | `0` | 0, `false`, `off`, `no`, `n`, or `f`: Allow HotSwap on supported GPU and ASIC revisions. Any other non-empty value: Disable HotSwap and use the original code object load path. |
| `HSA_HOTSWAP_VERBOSE` Enables HotSwap diagnostic logging to stderr. | `0` | 0, `false`, `off`, `no`, `n`, or `f`: Disable HotSwap diagnostic logging. Any other non-empty value: Enable HotSwap diagnostic logging. |
| `HSA_HOTSWAP_STRICT_MODE` Enables the opt-in HotSwap strict mask workaround request for non-A0 `gfx1250` targets. | `0` | 0, `false`, `off`, `no`, `n`, or `f`: Disable strict mask workaround requests. Any other non-empty value: Request COMGR strict mask workarounds for non-A0 `gfx1250` HotSwap rewrites. |
| `AMD_COMGR_HOTSWAP_ENTRY_TRAMPOLINES` Controls whether ROCr requests COMGR entry-trampoline HotSwap rewriting for gfx12.5 targets. Disabled by default. | `0` | `1`, `true`, `on`, `yes`, or any other truthy value: Enable entry-trampoline rewrites for gfx125\* and `gfx12-5-generic` targets. Unset, empty, 0, `false`, `off`, `no`, `n`, or `f`: Disable entry-trampoline rewrites. |
| `HSA_ENABLE_DXG_DETECTION` Controls detection of the DXG driver (/dev/dxg) on WSL2. | `1` | 0: Disable DXG detection. 1: Enable DXG detection, allowing ROCr to detect that it is running in WSL2. |

### Hardware Debugging Environment Variables

The following environment variables are intended for experienced users who are debugging hardware-specific issues.
These settings may impact performance and stability and should only be used when troubleshooting specific hardware problems.

## HIPCC environment variables

This topic provides descriptions of the HIPCC environment variables.

| Environment variable | Value |
|---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| `HIP_PLATFORM` The platform targeted by HIP. If `HIP_PLATFORM` isn’t set, then [HIPCC](https://rocm.docs.amd.com/projects/HIPCC/en/latest/index.html) attempts to auto-detect the platform based on whether the `nvcc` tool is found. | `amd`, `nvidia` |
| `HIP_PATH` The path of the HIP SDK on Microsoft Windows for AMD platforms. | Default: `C:/hip` |
| `ROCM_PATH` The path of the installed ROCm software stack on Linux for AMD platforms. | Default: `/opt/rocm` |
| `CUDA_PATH` Path to the CUDA SDK, which is only used for NVIDIA platforms. | Default: `/usr/local/cuda` |
| `HIP_CLANG_PATH` Path to the clang, which is only used for AMD platforms. | Default: `ROCM_PATH/llvm/bin` or `HIP_PATH/../llvm/bin"` |
| `HIP_LIB_PATH` The HIP device library installation path. | Default: `HIP_PATH/lib` |
| `HIP_DEVICE_LIB_PATH` The HIP device library installation path. | |
| `HIPCC_COMPILE_FLAGS_APPEND` Append extra flags as compilation options to `hipcc`. | |
| `HIPCC_LINK_FLAGS_APPEND` Append extra flags as compilation options to `hipcc`. | |
| `HIPCC_VERBOSE` Outputs detailed information on subcommands executed during compilation. | 1: Displays the command to `clang++` or `nvcc` with all options (`hipcc-cmd`). 2: Displays all relevant environment variables and their values. 4: Displays only the arguments passed to the `hipcc` command (`hipcc_args`). 5: Displays both the command to `clang++` or `nvcc` and `hipcc` arguments (`hipcc-cmd` and `hipcc-args`). 6: Displays all relevant environment variables and their values, along with the arguments to the `hipcc` command. 7: Displays all of the above: `hipcc-cmd`, `hipcc-args`, and environment variables. |

## Environment variables in ROCm libraries

Many ROCm libraries define environment variables for specific tuning, debugging,
or behavioral control. The table below provides an overview and links to further
documentation.

| Library | Purpose of Environment Variables |
|------------------------------------------------------------------------------------------------------------|----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| [hipBLASLt](https://rocm.docs.amd.com/projects/hipBLASLt/en/latest/reference/env-variables.html) | Manage logging, debugging, offline tuning, and stream-K configuration for hipBLASLt. |
| hipFile | Control compatibility modes, supported file systems, and statistics collection. |
| [hipSPARSELt](https://rocm.docs.amd.com/projects/hipSPARSELt/en/latest/reference/env-variables.html) | Control logging, debugging and performance monitoring of hipSPARSELt. |
| [rocBLAS](https://rocm.docs.amd.com/projects/rocBLAS/en/latest/reference/env-variables.html) | Performance tuning, kernel selection, logging, and debugging for BLAS operations. |
| rocSHMEM | Control the behavior of rocSHMEM. |
| [rocSOLVER](https://rocm.docs.amd.com/projects/rocSOLVER/en/latest/reference/env_variables.html) | Control logging of rocSOLVER. |
| [rocSPARSE](https://rocm.docs.amd.com/projects/rocSPARSE/en/latest/reference/env_variables.html) | Control logging of rocSPARSE. |
| [MIGraphX](https://rocm.docs.amd.com/projects/AMDMIGraphX/en/latest/reference/MIGraphX-dev-env-vars.html) | Control debugging, testing, and model performance tuning options for MIGraphX. |
| [MIOpen](https://rocm.docs.amd.com/projects/MIOpen/en/latest/reference/env_variables.html) | Control MIOpen logging and debugging, find mode and algorithm behavior and others. |
| [MIVisionX](https://rocm.docs.amd.com/projects/MIVisionX/en/latest/reference/MIVisionX-env-variables.html) | Control core OpenVX, GPU/device and debugging/profiling, stitching and chroma key configurations, file I/O operations, model deployment, and neural network parameters of MIVisionX. |
| [RCCL](https://rocm.docs.amd.com/projects/rccl/en/latest/api-reference/env-variables.html) | Control the logging, debugging, compiler and assembly behavior, and cache of RPP. |
| [RPP](https://rocm.docs.amd.com/projects/rpp/en/latest/reference/rpp-env-variables.html) | Logging, debugging, compiler and assembly management, and cache control in RPP |
| [Tensile](https://rocm.docs.amd.com/projects/Tensile/en/latest/src/reference/environment-variables.html) | Enable testing, debugging, and experimental features for Tensile clients and applications |

## Key single-variable details

This section provides detailed descriptions, in the standard format, for ROCm
libraries that feature a single, key environment variable (or a very minimal set)
which is documented directly on this page for convenience.

### rocALUTION

---

Source: https://rocm.docs.amd.com/en/latest/reference/environment-variables/setting-cus.html

# Set the number of compute units

![image](images/shared/compute-unit.png)

The GPU driver provides two environment variables to set the number of compute units (CUs) used:

- `HSA_CU_MASK`
- `ROC_GLOBAL_CU_MASK`

The `ROC_GLOBAL_CU_MASK` variable sets the CU mask on queues created by HIP or OpenCL runtimes. The `HSA_CU_MASK` variable sets the mask on a lower level of queue creation in the driver. It also sets the mask on the queues being profiled.

#### NOTE
When using GPUs to accelerate compute workloads, it sometimes becomes necessary to configure the hardware’s usage of compute units (CU). This is a more advanced option, so please read this page before experimentation.

The environment variables have the following syntax:

```default
ID = [0-9][0-9]*                         ex. base 10 numbers
ID_list = (ID | ID-ID)[, (ID | ID-ID)]*  ex. 0,2-4,7
GPU_list = ID_list                       ex. 0,2-4,7
CU_list = 0x[0-F]* | ID_list             ex. 0x337F OR 0,2-4,7
CU_Set = GPU_list : CU_list              ex. 0,2-4,7:0-15,32-47 OR 0,2-4,7:0x337F
HSA_CU_MASK = CU_Set [; CU_Set]*         ex. 0,2-4,7:0-15,32-47; 3-9:0x337F
```

The GPU indices are taken post `ROCR_VISIBLE_DEVICES` reordering. The listed or masked CUs are enabled for listed GPUs, and the others are disabled. Unlisted GPUs are not be affected, and their CUs are enabled.

The variable parsing stops when a syntax error occurs. The erroneous set and the following are ignored. Repeating GPU or CU IDs results in a syntax error. Specifying a mask with no usable CUs (CU_list is 0x0) results in a syntax error. To exclude GPU devices, use `ROCR_VISIBLE_DEVICES`.

#### NOTE
These environment variables only affect ROCm software, not graphics applications.

Not all CU configurations are valid on all devices. For example, on devices where two CUs can be combined into a WGP (for kernels running in WGP mode), it’s not valid to disable only a single CU in a WGP.

---

Source: https://rocm.docs.amd.com/en/latest/reference/hip-programming.html

# AMD GPU programming on ROCm

ROCm provides a robust environment for heterogeneous programs running on CPUs
and AMD GPUs. ROCm supports various programming languages and frameworks to
help developers access the power of AMD GPUs. The natively supported programming
languages are HIP and OpenCL, but HIP bindings are available for Python and
Fortran.

#### NOTE
* For a complete description of the HIP programming language, see the
  [HIP documentation](https://rocm.docs.amd.com/projects/HIP/en/latest/index.html).
* Developers who require a unified, book-style reference for ROCm and HIP can
  consult the [AMD ROCm Programming Guide](https://rocm-handbook.amd.com/projects/amd-rocm-programming-guide/en/latest/).
  It aggregates documentation from the ROCm portal and organizes it into a
  structured format optimized for in-depth study and offline access in both
  PDF and HTML formats.

HIP is an API based on C++ that provides a runtime and kernel language for GPU
programming and is the essential ROCm programming language. HIP enables
single-source C++ programming with support for templates, C++11 lambdas,
classes, and namespaces, letting developers create applications that run on
heterogeneous systems using both CPUs and AMD GPUs from a single code base.

HIP provides two components: code that runs on the host (CPU) and code that runs
on the device (GPU). Host code manages device buffers, moves data between host
and device, launches kernels, and handles streams, events, and synchronization.
The kernel language provides access to GPU-specific hardware capabilities for
massively parallel execution.

ROCm includes a complete toolchain: compilers ([clang](https://rocm.docs.amd.com/projects/llvm-project/en/latest/index.html),
[hipcc](https://rocm.docs.amd.com/projects/HIPCC/en/latest/index.html)), a code profiler
([rocprofv3](https://rocm.docs.amd.com/projects/rocprofiler-sdk/en/latest/how-to/using-rocprofv3.html)), and a debugger
([rocgdb](https://rocm.docs.amd.com/projects/ROCgdb/en/latest/index.html)). ROCm also provides libraries such as
[hipFFT](https://rocm.docs.amd.com/projects/hipFFT/en/latest/index.html) and [hipBLAS](https://rocm.docs.amd.com/projects/hipBLAS/en/latest/index.html) that offer API
compatibility with equivalent NVIDIA CUDA libraries, making it easier to
integrate into existing workflows.

Developers with CUDA experience will find the HIP API familiar. HIP allows code
written for CUDA to be ported to AMD GPUs. [HIPIFY](https://rocm.docs.amd.com/projects/HIPIFY/en/latest/index.html), based
on the Clang front-end and Perl, can convert CUDA API calls into the
corresponding HIP API calls. However, HIP is not a drop-in replacement for
CUDA, and some manual coding and performance tuning may be required when porting
existing projects to AMD GPUs.

Python bindings can be found at [ROCm/hip-python](https://github.com/ROCm/hip-python).
Python is popular in AI and machine learning applications due to the
availability of frameworks such as PyTorch and TensorFlow.

Fortran bindings can be found at [ROCm/hipfort](https://github.com/ROCm/hipfort).
It enables scientific, academic, and legacy applications, particularly those in
high-performance computing, to run on AMD GPUs via HIP.

OpenCL (Open Computing Language) is an open standard for cross-platform,
parallel programming of diverse processors. ROCm supports OpenCL for developers
who want to use standard frameworks across different hardware platforms,
including CPUs, GPUs, and APUs. For more information, see
[OpenCL](https://www.khronos.org/opencl/).

---

Source: https://rocm.docs.amd.com/en/latest/contribute/contributing.html

# Contributing to ROCm documentation

ROCm documentation is open source and available on GitHub. You can contribute to ROCm documentation by cloning the appropriate repository, making your changes, and opening a pull request.

#### NOTE
To provide feedback on the ROCm documentation without contributing to it, see [Providing feedback about the ROCm documentation](https://rocm.docs.amd.com/en/latest/contribute/feedback.html).

The documentation for ROCm and for all ROCm components is under their respective `docs` folders. The `docs` folders for all components across ROCm have the same structure:

| Sub-folder name   | Documentation type                                               |
|-------------------|------------------------------------------------------------------|
| `install`         | Installation instructions, build instructions, and prerequisites |
| `conceptual`      | Important concepts                                               |
| `how-to`          | How to implement specific use cases                              |
| `tutorials`       | Tutorials                                                        |
| `reference`       | API references and other reference resources                     |
| `sphinx`          | `_toc.yaml.in` file                                              |

ROCm stack documentation differs from this structure.

Most documentation topics are written in [reStructuredText (rst)](https://www.sphinx-doc.org/en/master/usage/restructuredtext/index.html), with some topics written in Markdown. Only use reStructuredText when adding new topics. Only use Markdown if the topic you are editing is already in Markdown.

To edit or add to the documentation, first clone the appropriate repository, ensuring that you follow the repository recommendations as outlined in their respective README files:

| Module                | Repository                                                    |
|-----------------------|---------------------------------------------------------------|
| ROCm stack            | [ROCm/ROCm](https://github.com/ROCm/ROCm)                     |
| ROCm libraries        | [ROCm/rocm-libraries](https://github.com/ROCm/rocm-libraries) |
| ROCm systems projects | [ROCm/rocm-systems](https://github.com/ROCm/rocm-systems)     |

#### NOTE
Individual components in the ROCm libraries and the ROCm systems projects repositories are located under the `projects` folder. Some components have their own individual repositories. Each component has a link to its GitHub location from its documentation.

Cut a local branch from the `develop` branch of the repository and make your changes. Your changes must adhere to the [Google developer documentation style guide](https://developers.google.com/style/highlights). If you’re adding a topic, provide a link to it from the `index.rst` and `_toc.yaml.in` files.

Build the documentation locally to verify your changes.

#### NOTE
If you’re making changes to the Doxygen comments within the source code, delete the `docs/doxygen/xml` and `docs/doxygen/html` folders between each build.

From within the `docs` directory, run:

```bash
python3 -m sphinx -T -E -b html -d _build/doctrees -D language=en . _build/html
```

The output will be saved to the `docs/_build` folder. Open `docs/_build/html/index.html` to view the documentation.

#### NOTE
If your build returns an error due to missing packages, run

`pip3 install -r sphinx/requirements.txt`

This command only needs to be run once.

Once you’ve verified your changes, push your branch and create a pull request. Your pull request will be reviewed by a member of the ROCm documentation team.

For information about how to clone a repository, or how to create and push a local branch, see the [GitHub documentation](https://docs.github.com/en).

For information about ROCm build tools, see [Documentation toolchain](https://rocm.docs.amd.com/en/latest/contribute/toolchain.html).

#### IMPORTANT
By creating a pull request (PR), you agree to allow your contribution to be licensed under the terms of the
LICENSE.txt file in the corresponding repository. Different repositories can use different licenses.

---

Source: https://rocm.docs.amd.com/en/latest/contribute/toolchain.html

# ROCm documentation toolchain

The ROCm documentation relies on several open source toolchains and sites.

## rocm-docs-core

[rocm-docs-core](https://github.com/ROCm/rocm-docs-core) is an AMD-maintained
project that applies customizations for the ROCm documentation. This project is the tool most ROCm repositories use as part of their documentation build pipeline. It is available as a [pip package on PyPI](https://pypi.org/project/rocm-docs-core/).

See the user and developer guides for rocm-docs-core at
[rocm-docs-core documentation](https://rocm.docs.amd.com/projects/rocm-docs-core/en/latest/index.html).

## Sphinx

[Sphinx](https://www.sphinx-doc.org/en/master/) is a documentation generator originally used for Python. It is now widely used in the open source community.

### Sphinx External ToC

[Sphinx External ToC](https://sphinx-external-toc.readthedocs.io/en/latest/intro.html) is a Sphinx extension used for ROCm documentation navigation. This tool generates a navigation menu on the left
based on a YAML file (`_toc.yml.in`) that contains the table of contents.

### Sphinx-book-theme

[Sphinx-book-theme](https://sphinx-book-theme.readthedocs.io/en/latest/) is a Sphinx theme that defines the base appearance for ROCm documentation. ROCm documentation applies some customization, such as a custom header and footer, on top of the Sphinx Book Theme.

### Sphinx Design

[Sphinx design](https://sphinx-design.readthedocs.io/en/latest/index.html) is a Sphinx extension that adds design functionality. ROCm documentation uses Sphinx Design for grids, cards, and synchronized tabs.

## Doxygen

[Doxygen](https://www.doxygen.nl/) is a documentation generator that extracts information from in-code comments. It is used for API documentation.

## Breathe

[Breathe](https://www.breathe-doc.org/) is a Sphinx plugin for integrating Doxygen content.

## Read the Docs

[Read the Docs](https://docs.readthedocs.io/en/stable/) is the service that builds and hosts the HTML version of the ROCm documentation.

---

Source: https://rocm.docs.amd.com/en/latest/contribute/building.html

# Building documentation

## GitHub

If you open a pull request and scroll down to the summary panel,
there is a commit status section. Next to the line
`docs/readthedocs.com:advanced-micro-devices-demo`, there is a `Details` link.
If you click this, it takes you to the Read the Docs build for your pull request.

![GitHub PR commit status](data/contribute/commit-status.png)

If you don’t see this line, click `Show all checks` to get an itemized view.

## Command line

You can build our documentation via the command line using Python.

See the `build.tools.python` setting in the [Read the Docs configuration file](https://github.com/ROCm/ROCm/blob/develop/.readthedocs.yaml) for the Python version used by Read the Docs to build documentation.

See the [Python requirements file](https://github.com/ROCm/ROCm/blob/develop/docs/sphinx/requirements.txt) for Python packages needed to build the documentation.

Use the Python Virtual Environment (`venv`) and run the following commands from the project root:

**Linux and WSL**

```sh
python3 -mvenv .venv

.venv/bin/python -m pip install -r docs/sphinx/requirements.txt
.venv/bin/python -m sphinx -T -E -b html -d _build/doctrees -D language=en docs _build/html
```

**Windows**

```powershell
python -mvenv .venv

.venv\Scripts\python.exe -m pip install -r docs/sphinx/requirements.txt
.venv\Scripts\python.exe -m sphinx -T -E -b html -d _build/doctrees -D language=en docs _build/html
```

Navigate to `_build/html/index.html` and open this file in a web browser.

## Visual Studio Code

With the help of a few extensions, you can create a productive environment to author and test
documentation locally using Visual Studio (VS) Code. Follow these steps to configure VS Code:

1. Install the required extensions:
   * Python: `(ms-python.python)`
   * Live Server: `(ritwickdey.LiveServer)`
2. Add the following entries to `.vscode/settings.json`.
   ```json
     {
       "liveServer.settings.root": "/.vscode/build/html",
       "liveServer.settings.wait": 1000,
       "python.terminal.activateEnvInCurrentTerminal": true
     }
   ```

* `liveServer.settings.root`: Sets the root of the output website for live previews. Must be changed
     alongside the `tasks.json` command.
   * `liveServer.settings.wait`: Tells the live server to wait with the update in order to give Sphinx time to
     regenerate the site contents and not refresh before the build is complete.
   * `python.terminal.activateEnvInCurrentTerminal`: Activates the automatic virtual environment, so you
     can build the site from the integrated terminal.
3. Add the following tasks to `.vscode/tasks.json`.
   ```json
     {
       "version": "2.0.0",
       "tasks": [
         {
           "label": "Build Docs",
           "type": "process",
           "windows": {
             "command": "${workspaceFolder}/.venv/Scripts/python.exe"
           },
           "command": "${workspaceFolder}/.venv/bin/python3",
           "args": [
             "-m",
             "sphinx",
             "-j",
             "auto",
             "-T",
             "-b",
             "html",
             "-d",
             "${workspaceFolder}/.vscode/build/doctrees",
             "-D",
             "language=en",
             "${workspaceFolder}/docs",
             "${workspaceFolder}/.vscode/build/html"
           ],
           "problemMatcher": [
             {
               "owner": "sphinx",
               "fileLocation": "absolute",
               "pattern": {
                 "regexp": "^(?:.*\\.{3}\\s+)?(\\/[^:]*|[a-zA-Z]:\\\\[^:]*):(\\d+):\\s+(WARNING|ERROR):\\s+(.*)$",
                 "file": 1,
                 "line": 2,
                 "severity": 3,
                 "message": 4
               }
             },
             {
             "owner": "sphinx",
               "fileLocation": "absolute",
               "pattern": {
                 "regexp": "^(?:.*\\.{3}\\s+)?(\\/[^:]*|[a-zA-Z]:\\\\[^:]*):{1,2}\\s+(WARNING|ERROR):\\s+(.*)$",
                 "file": 1,
                 "severity": 2,
                 "message": 3
               }
             }
           ],
           "group": {
             "kind": "build",
             "isDefault": true
           }
         }
       ]
     }
   ```

> Implementation detail: two problem matchers were needed to be defined,
   > because VS Code doesn’t tolerate some problem information being potentially
   > absent. While a single regex could match all types of errors, if a capture
   > group remains empty (the line number doesn’t show up in all warning/error
   > messages) but the `pattern` references said empty capture group, VS Code
   > discards the message completely.
4. Configure the Python virtual environment (`venv`).

From the Command Palette, run `Python: Create Environment`. Select `venv` environment and
   `docs/sphinx/requirements.txt`.
5. Build the docs.

Launch the default build task using one of the following options:
   * A hotkey (the default is `Ctrl+Shift+B`)
   * Issuing the `Tasks: Run Build Task` from the Command Palette
6. Open the live preview.

Navigate to the site output within VS Code: right-click on `.vscode/build/html/index.html` and
   select `Open with Live Server`. The contents should update on every rebuild without having to
   refresh the browser.

---

Source: https://rocm.docs.amd.com/en/latest/contribute/feedback.html

# Providing feedback about the ROCm documentation

Feedback about the ROCm documentation is welcome. You can provide feedback about the ROCm documentation either through GitHub Discussions or GitHub Issues.

## Participating in discussions through GitHub Discussions

You can ask questions, view announcements, suggest new features, and communicate with other members of the community through [GitHub Discussions](https://github.com/ROCm/TheRock/discussions).

## Submitting issues through GitHub Issues

You can submit issues through [GitHub Issues](https://github.com/ROCm/TheRock/issues).

When creating a new issue, follow the following guidelines:

1. Always do a search to see if the same issue already exists. If the issue already exists, upvote it, and comment or post to provide any additional details you might have.
2. If you find an issue that is similar to your issue, log your issue, then add a comment that includes a link to the similar issue, as well as its issue number.
3. Always provide as much information as possible. This helps reduce the time required to reproduce the issue.

After creating your issue, make sure to check it regularly for any requests for additional information.

For information about contributing content to the ROCm documentation, see [Contributing to the ROCm documentation](https://rocm.docs.amd.com/en/latest/contribute/contributing.html).

---

Source: https://rocm.docs.amd.com/en/latest/about/license.html

# ROCm licenses

MIT License

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the “Software”), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED “AS IS”, WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.

#### NOTE
The preceding license applies to the [ROCm repository](https://github.com/ROCm/ROCm), which
primarily contains documentation. For licenses related to other ROCm components, refer to the
following section.

## ROCm component licenses

ROCm is released by Advanced Micro Devices, Inc. (AMD) and is licensed per component separately.
The following table is a list of ROCm components with links to their respective license
terms. These components may include third party components subject to
additional licenses. Please review individual repositories for more information.

| Component                                                                                                | License                                                                                                                |
|----------------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------------------------------|
| [AMD Compute Language Runtime (CLR)](https://github.com/ROCm/rocm-systems/tree/develop/projects/clr)     | [MIT](https://github.com/ROCm/rocm-systems/blob/develop/projects/clr/LICENSE.md)                                       |
| [AMD SMI](https://github.com/ROCm/amdsmi)                                                                | [MIT](https://github.com/ROCm/amdsmi/blob/amd-staging/LICENSE)                                                         |
| [aomp](https://github.com/ROCm/aomp/)                                                                    | [Apache 2.0](https://github.com/ROCm/aomp/blob/aomp-dev/LICENSE)                                                       |
| [aomp-extras](https://github.com/ROCm/aomp-extras/)                                                      | [MIT](https://github.com/ROCm/aomp-extras/blob/aomp-dev/LICENSE)                                                       |
| [AQLprofile](https://github.com/ROCm/rocm-systems/tree/develop/projects/aqlprofile/)                     | [MIT](https://github.com/ROCm/rocm-systems/blob/develop/projects/aqlprofile/LICENSE.md)                                |
| [Code Object Manager (Comgr)](https://github.com/ROCm/llvm-project/tree/amd-staging/amd/comgr)           | [The University of Illinois/NCSA](https://github.com/ROCm/llvm-project/blob/amd-staging/amd/comgr/LICENSE.txt)         |
| [Composable Kernel](https://github.com/ROCm/rocm-libraries/tree/develop/projects/composablekernel)       | [MIT](https://github.com/ROCm/rocm-libraries/tree/develop/projects/composablekernel/LICENSE)                           |
| [half](https://github.com/ROCm/half/)                                                                    | [MIT](https://github.com/ROCm/half/blob/rocm/LICENSE.txt)                                                              |
| [HIP](https://github.com/ROCm/rocm-systems/tree/develop/projects/hip/)                                   | [MIT](https://github.com/ROCm/rocm-systems/blob/develop/projects/hip/LICENSE.md)                                       |
| [hipamd](https://github.com/ROCm/rocm-systems/tree/develop/projects/clr/hipamd/)                         | [MIT](https://github.com/ROCm/rocm-systems/blob/develop/projects/clr/hipamd/LICENSE.md)                                |
| [hipBLAS](https://github.com/ROCm/rocm-libraries/tree/develop/projects/hipblas/)                         | [MIT](https://github.com/ROCm/rocm-libraries/blob/develop/projects/hipblas/LICENSE.md)                                 |
| [hipBLASLt](https://github.com/ROCm/rocm-libraries/tree/develop/projects/hipblaslt/)                     | [MIT](https://github.com/ROCm/rocm-libraries/blob/develop/projects/hipblaslt/LICENSE.md)                               |
| [HIPCC](https://github.com/ROCm/llvm-project/tree/amd-staging/amd/hipcc)                                 | [MIT](https://github.com/ROCm/llvm-project/blob/amd-staging/amd/hipcc/LICENSE.txt)                                     |
| [hipCUB](https://github.com/ROCm/rocm-libraries/tree/develop/projects/hipcub/)                           | [Custom](https://github.com/ROCm/rocm-libraries/blob/develop/projects/hipcub/LICENSE.txt)                              |
| [hipFFT](https://github.com/ROCm/rocm-libraries/tree/develop/projects/hipfft/)                           | [MIT](https://github.com/ROCm/rocm-libraries/blob/develop/projects/hipfft/LICENSE.md)                                  |
| [hipFile](https://github.com/ROCm/rocm-systems/tree/develop/projects/hipfile/)                           | [MIT](https://github.com/ROCm/rocm-systems/blob/develop/projects/hipfile/LICENSE.md)                                   |
| [hipfort](https://github.com/ROCm/hipfort/)                                                              | [MIT](https://github.com/ROCm/hipfort/blob/develop/LICENSE)                                                            |
| [HIPIFY](https://github.com/ROCm/HIPIFY/)                                                                | [MIT](https://github.com/ROCm/HIPIFY/blob/amd-staging/LICENSE.txt)                                                     |
| [hipRAND](https://github.com/ROCm/rocm-libraries/tree/develop/projects/hiprand/)                         | [MIT](https://github.com/ROCm/rocm-libraries/blob/develop/projects/hiprand/LICENSE.md)                                 |
| [hipSOLVER](https://github.com/ROCm/rocm-libraries/tree/develop/projects/hipsolver/)                     | [MIT](https://github.com/ROCm/rocm-libraries/blob/develop/projects/hipsolver/LICENSE.md)                               |
| [hipSPARSE](https://github.com/ROCm/rocm-libraries/tree/develop/projects/hipsparse/)                     | [MIT](https://github.com/ROCm/rocm-libraries/blob/develop/projects/hipsparse/LICENSE.md)                               |
| [hipSPARSELt](https://github.com/ROCm/rocm-libraries/tree/develop/projects/hipsparselt/)                 | [MIT](https://github.com/ROCm/rocm-libraries/blob/develop/projects/hipsparselt/LICENSE.md)                             |
| [hipTensor](https://github.com/ROCm/rocm-libraries/tree/develop/projects/hiptensor/)                     | [MIT](https://github.com/ROCm/rocm-libraries/blob/develop/projects/hiptensor/LICENSE)                                  |
| [llvm-project](https://github.com/ROCm/llvm-project/)                                                    | [Apache](https://github.com/ROCm/llvm-project/blob/amd-staging/LICENSE.TXT)                                            |
| [llvm-project/flang](https://github.com/ROCm/llvm-project/tree/amd-staging/flang)                        | [Apache 2.0](https://github.com/ROCm/llvm-project/blob/amd-staging/flang/LICENSE.TXT)                                  |
| [MIGraphX](https://github.com/ROCm/AMDMIGraphX/)                                                         | [MIT](https://github.com/ROCm/AMDMIGraphX/blob/develop/LICENSE)                                                        |
| [MIOpen](https://github.com/ROCm/rocm-libraries/tree/develop/projects/miopen/)                           | [MIT](https://github.com/ROCm/rocm-libraries/blob/develop/projects/miopen/LICENSE.md)                                  |
| [MIVisionX](https://github.com/ROCm/MIVisionX/)                                                          | [MIT](https://github.com/ROCm/MIVisionX/blob/develop/LICENSE.txt)                                                      |
| [rocAL](https://github.com/ROCm/rocAL)                                                                   | [MIT](https://github.com/ROCm/rocAL/blob/develop/LICENSE.txt)                                                          |
| [rocALUTION](https://github.com/ROCm/rocALUTION/)                                                        | [MIT](https://github.com/ROCm/rocALUTION/blob/develop/LICENSE.md)                                                      |
| [rocBLAS](https://github.com/ROCm/rocm-libraries/tree/develop/projects/rocblas/)                         | [MIT](https://github.com/ROCm/rocm-libraries/blob/develop/projects/rocblas/LICENSE.md)                                 |
| [ROCdbgapi](https://github.com/ROCm/ROCdbgapi/)                                                          | [MIT](https://github.com/ROCm/ROCdbgapi/blob/amd-staging/LICENSE.txt)                                                  |
| [rocDecode](https://github.com/ROCm/rocm-systems/tree/develop/projects/rocdecode)                        | [MIT](https://github.com/ROCm/rocm-systems/tree/develop/projects/rocdecode/LICENSE)                                    |
| [rocFFT](https://github.com/ROCm/rocm-libraries/tree/develop/projects/rocfft/)                           | [MIT](https://github.com/ROCm/rocm-libraries/blob/develop/projects/rocfft/LICENSE.md)                                  |
| [ROCgdb](https://github.com/ROCm/ROCgdb/)                                                                | [GNU General Public License v3.0](https://github.com/ROCm/ROCgdb/blob/amd-staging/COPYING3)                            |
| [rocJPEG](https://github.com/ROCm/rocm-systems/tree/develop/projects/rocjpeg)                            | [MIT](https://github.com/ROCm/rocm-systems/tree/develop/projects/rocjpeg/LICENSE)                                      |
| [ROCK-Kernel-Driver](https://github.com/ROCm/ROCK-Kernel-Driver/)                                        | [GPL 2.0 WITH Linux-syscall-note](https://github.com/ROCm/ROCK-Kernel-Driver/blob/master/COPYING)                      |
| [rocminfo](https://github.com/ROCm/rocm-systems/tree/develop/projects/rocminfo/)                         | [The University of Illinois/NCSA](https://github.com/ROCm/rocm-systems/blob/develop/projects/rocminfo/License.txt)     |
| [ROCm Bandwidth Test](https://github.com/ROCm/rocm_bandwidth_test/)                                      | [MIT](https://github.com/ROCm/rocm_bandwidth_test/blob/master/LICENSE.txt)                                             |
| [ROCm CMake](https://github.com/ROCm/rocm-cmake/)                                                        | [MIT](https://github.com/ROCm/rocm-cmake/blob/develop/LICENSE)                                                         |
| [ROCm Communication Collectives Library (RCCL)](https://github.com/ROCm/rccl/)                           | [Custom](https://github.com/ROCm/rccl/blob/develop/LICENSE.txt)                                                        |
| [ROCm-Core](https://github.com/ROCm/rocm-systems/tree/develop/projects/rocm-core/)                       | [MIT](https://github.com/ROCm/rocm-systems/blob/develop/projects/rocm-core/LICENSE.md)                                 |
| [ROCm Compute Profiler](https://github.com/ROCm/rocm-systems/tree/develop/projects/rocprofiler-compute/) | [MIT](https://github.com/ROCm/rocm-systems/blob/develop/projects/rocprofiler-compute/LICENSE.md)                       |
| [ROCm Data Center (RDC)](https://github.com/ROCm/rocm-systems/tree/develop/projects/rdc/)                | [MIT](https://github.com/ROCm/rocm-systems/blob/develop/projects/rdc/LICENSE.md)                                       |
| [ROCm-Device-Libs](https://github.com/ROCm/llvm-project/tree/amd-staging/amd/device-libs)                | [The University of Illinois/NCSA](https://github.com/ROCm/llvm-project/blob/amd-staging/amd/device-libs/LICENSE.TXT)   |
| [ROCm-OpenCL-Runtime](https://github.com/ROCm/rocm-systems/tree/develop/projects/clr/opencl/)            | [MIT](https://github.com/ROCm/rocm-systems/blob/develop/projects/clr/opencl/LICENSE.md)                                |
| [ROCm Performance Primitives (RPP)](https://github.com/ROCm/rpp)                                         | [MIT](https://github.com/ROCm/rpp/blob/develop/LICENSE)                                                                |
| [ROCm SMI Lib](https://github.com/ROCm/rocm-systems/tree/develop/projects/rocm-smi-lib/)                 | [MIT](https://github.com/ROCm/rocm-systems/blob/develop/projects/rocm-smi-lib/LICENSE.md)                              |
| [ROCm Systems Profiler](https://github.com/ROCm/rocm-systems/tree/develop/projects/rocprofiler-systems/) | [MIT](https://github.com/ROCm/rocm-systems/blob/develop/projects/rocprofiler-systems/LICENSE.md)                       |
| [ROCm Validation Suite](https://github.com/ROCm/ROCmValidationSuite/)                                    | [MIT](https://github.com/ROCm/ROCmValidationSuite/blob/master/LICENSE)                                                 |
| [rocPRIM](https://github.com/ROCm/rocm-libraries/tree/develop/projects/rocprim/)                         | [MIT](https://github.com/ROCm/rocm-libraries/blob/develop/projects/rocprim/LICENSE.md)                                 |
| [ROCProfiler](https://github.com/ROCm/rocm-systems/tree/develop/projects/rocprofiler/)                   | [MIT](https://github.com/ROCm/rocm-systems/blob/develop/projects/rocprofiler/LICENSE.md)                               |
| [ROCprofiler-SDK](https://github.com/ROCm/rocm-systems/tree/develop/projects/rocprofiler-sdk/)           | [MIT](https://github.com/ROCm/rocm-systems/blob/develop/projects/rocprofiler-sdk/LICENSE.md)                           |
| [rocPyDecode](https://github.com/ROCm/rocPyDecode)                                                       | [MIT](https://github.com/ROCm/rocPyDecode/blob/develop/LICENSE.txt)                                                    |
| [rocRAND](https://github.com/ROCm/rocm-libraries/tree/develop/projects/rocrand/)                         | [MIT](https://github.com/ROCm/rocm-libraries/blob/develop/projects/rocrand/LICENSE.md)                                 |
| [ROCR Debug Agent](https://github.com/ROCm/rocr_debug_agent/)                                            | [The University of Illinois/NCSA](https://github.com/ROCm/rocr_debug_agent/blob/amd-staging/LICENSE.txt)               |
| [ROCR-Runtime](https://github.com/ROCm/rocm-systems/tree/develop/projects/rocr-runtime/)                 | [The University of Illinois/NCSA](https://github.com/ROCm/rocm-systems/blob/develop/projects/rocr-runtime/LICENSE.txt) |
| [rocSHMEM](https://github.com/ROCm/rocSHMEM/)                                                            | [MIT](https://github.com/ROCm/rocSHMEM/blob/develop/LICENSE.md)                                                        |
| [rocSOLVER](https://github.com/ROCm/rocm-libraries/tree/develop/projects/rocsolver/)                     | [BSD-2-Clause](https://github.com/ROCm/rocm-libraries/blob/develop/projects/rocsolver/LICENSE.md)                      |
| [rocSPARSE](https://github.com/ROCm/rocm-libraries/tree/develop/projects/rocsparse/)                     | [MIT](https://github.com/ROCm/rocm-libraries/blob/develop/projects/rocsparse/LICENSE.md)                               |
| [rocThrust](https://github.com/ROCm/rocm-libraries/tree/develop/projects/rocthrust/)                     | [Apache 2.0](https://github.com/ROCm/rocm-libraries/blob/develop/projects/rocthrust/LICENSE)                           |
| [ROCTracer](https://github.com/ROCm/rocm-systems/tree/develop/projects/roctracer/)                       | [MIT](https://github.com/ROCm/rocm-systems/blob/develop/projects/roctracer/LICENSE.md)                                 |
| [rocWMMA](https://github.com/ROCm/rocm-libraries/tree/develop/projects/rocwmma/)                         | [MIT](https://github.com/ROCm/rocm-libraries/blob/develop/projects/rocwmma/LICENSE.md)                                 |
| [Tensile](https://github.com/ROCm/rocm-libraries/tree/develop/shared/tensile/)                           | [MIT](https://github.com/ROCm/rocm-libraries/blob/develop/shared/tensile/LICENSE.md)                                   |
| [TransferBench](https://github.com/ROCm/TransferBench)                                                   | [MIT](https://github.com/ROCm/TransferBench/blob/develop/LICENSE.md)                                                   |

Open sourced ROCm components are released via public GitHub
repositories, packages on [https://repo.radeon.com](https://repo.radeon.com) and other distribution channels.
Proprietary products are only available on [https://repo.radeon.com](https://repo.radeon.com).
Proprietary components are organized in a proprietary subdirectory in the package
repositories to distinguish from open sourced packages.

#### NOTE
The following additional terms and conditions apply to your use of ROCm technical documentation.

The information presented in this document is for informational purposes only
and may contain technical inaccuracies, omissions, and typographical errors. The
information contained herein is subject to change and may be rendered inaccurate
for many reasons, including but not limited to product and roadmap changes,
component and motherboard version changes, new model and/or product releases,
product differences between differing manufacturers, software changes, BIOS
flashes, firmware upgrades, or the like. Any computer system has risks of
security vulnerabilities that cannot be completely prevented or mitigated. AMD
assumes no obligation to update or otherwise correct or revise this information.
However, AMD reserves the right to revise this information and to make changes
from time to time to the content hereof without obligation of AMD to notify any
person of such revisions or changes.

THIS INFORMATION IS PROVIDED “AS IS.” AMD MAKES NO REPRESENTATIONS OR WARRANTIES
WITH RESPECT TO THE CONTENTS HEREOF AND ASSUMES NO RESPONSIBILITY FOR ANY
INACCURACIES, ERRORS, OR OMISSIONS THAT MAY APPEAR IN THIS INFORMATION. AMD
SPECIFICALLY DISCLAIMS ANY IMPLIED WARRANTIES OF NON-INFRINGEMENT,
MERCHANTABILITY, OR FITNESS FOR ANY PARTICULAR PURPOSE. IN NO EVENT WILL AMD BE
LIABLE TO ANY PERSON FOR ANY RELIANCE, DIRECT, INDIRECT, SPECIAL, OR OTHER
CONSEQUENTIAL DAMAGES ARISING FROM THE USE OF ANY INFORMATION CONTAINED HEREIN,
EVEN IF AMD IS EXPRESSLY ADVISED OF THE POSSIBILITY OF SUCH DAMAGES.

AMD, the AMD Arrow logo, ROCm, and combinations thereof are trademarks of
Advanced Micro Devices, Inc. Other product names used in this publication are
for identification purposes only and may be trademarks of their respective
companies.

### Package licensing

#### ATTENTION
ROCprof Trace Decoder and AOCC CPU optimizations are provided in binary form, subject to the license agreement enclosed on [GitHub](https://github.com/ROCm/rocprof-trace-decoder/blob/amd-mainline/LICENSE) for ROCprof Trace Decoder, and [Developer Central](https://www.amd.com/en/developer/aocc.html) for AOCC. By using, installing,
copying or distributing ROCprof Trace Decoder or AOCC CPU Optimizations, you agree to
the terms and conditions of this license agreement. If you do not agree to the
terms of this agreement, do not install, copy or use ROCprof Trace Decoder or the
AOCC CPU Optimizations.

For the rest of the ROCm packages, you can find the licensing information at the
following location: `/opt/rocm/share/doc/<component-name>/` or in the locations
specified in the preceding table.

For example, you can fetch the licensing information of the `amd_comgr`
component (Code Object Manager) from the `/opt/rocm/share/doc/amd_comgr/LICENSE.txt` file.

Bearbeitung