Update documentation requirements

Fix typo for TFLOPs metric in MI250 architecture page: cherry pick to docs/6.1.0

.github/CODEOWNERS

@@ -0,0 +1,5 @@
+* @amd-aakash @jlgreathouse @samjwu @ROCm/rocm-documentation
+# Documentation files
+docs/* @ROCm/rocm-documentation
+*.md @ROCm/rocm-documentation
+*.rst @ROCm/rocm-documentation

.github/dependabot.yml

@@ -0,0 +1,13 @@
+# To get started with Dependabot version updates, you'll need to specify which
+# package ecosystems to update and where the package manifests are located.
+# Please see the documentation for all configuration options:
+# https://docs.github.com/github/administering-a-repository/configuration-options-for-dependency-updates
+
+version: 2
+updates:
+  - package-ecosystem: "pip" # See documentation for possible values
+    directory: "/docs/sphinx" # Location of package manifests
+    open-pull-requests-limit: 10
+    schedule:
+      interval: "daily"
+    versioning-strategy: increase

.github/workflows/issue_retrieval.yml

@@ -0,0 +1,22 @@
+name: Issue retrieval
+
+on:
+  issues:
+    types: [opened]
+
+jobs:
+  auto-retrieve:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Generate a token
+        id: generate_token
+        uses: actions/create-github-app-token@v1
+        with:
+          app_id: ${{ secrets.ACTION_APP_ID }}
+          private_key: ${{ secrets.ACTION_PEM }}
+      - name: 'Retrieve Issue'
+        uses: abhimeda/rocm_issue_management@main
+        with:
+          authentication-token: ${{ steps.generate_token.outputs.token }}
+          github-organization: 'ROCm'
+          project-num: '6'

.github/workflows/linting.yml

@@ -0,0 +1,20 @@
+name: Linting
+
+on:
+  push:
+    branches: 
+    - develop
+    - main
+    - 'docs/*'
+    - 'roc**'
+  pull_request:
+    branches: 
+    - develop
+    - main
+    - 'docs/*'
+    - 'roc**'
+
+jobs:
+  call-workflow-passing-data:
+    name: Documentation
+    uses: ROCm/rocm-docs-core/.github/workflows/linting.yml@develop

.gitignore

@@ -0,0 +1,19 @@
+.venv
+.vscode
+build
+
+# documentation artifacts
+_build/
+_images/
+_static/
+_templates/
+_toc.yml
+docBin/
+_doxygen/
+_readthedocs/
+
+# avoid duplicating contributing.md due to conf.py
+docs/CHANGELOG.md
+docs/contribute/index.md
+docs/about/release-notes.md
+docs/about/CHANGELOG.md

.markdownlint-cli2.yaml

@@ -0,0 +1,17 @@
+config:
+  default: true
+  MD004:
+    style: asterisk
+  MD013: false
+  MD026:
+    punctuation: '.,;:!'
+  MD029:
+    style: ordered
+  MD033: false
+  MD034: false
+  MD041: false
+  MD051: false
+ignores:
+  - CHANGELOG.md
+  - "{,docs/}{RELEASE,release}.md"
+  - tools/autotag/templates/**/*.md

.readthedocs.yaml

@@ -0,0 +1,21 @@
+# Read the Docs configuration file
+# See https://docs.readthedocs.io/en/stable/config-file/v2.html for details
+
+version: 2
+
+build:
+   os: ubuntu-22.04
+   tools:
+      python: "3.10"
+   apt_packages:
+     - "doxygen"
+     - "graphviz" # For dot graphs in doxygen
+
+python:
+   install:
+   - requirements: docs/sphinx/requirements.txt
+
+sphinx:
+   configuration: docs/conf.py
+
+formats: []

.wordlist.txt

@@ -0,0 +1,688 @@
+AAC
+ABI
+ACE
+ACEs
+AccVGPR
+AccVGPRs
+ALU
+AMD
+AMDGPU
+AMDGPUs
+AMDMIGraphX
+AMI
+AOCC
+AOMP
+APIC
+APIs
+APU
+ASIC
+ASICs
+ASan
+ASAN
+ASm
+ATI
+AddressSanitizer
+AlexNet
+Arb
+BLAS
+BMC
+BitCode
+Blit
+Bluefield
+CCD
+CDNA
+CIFAR
+CLI
+CLion
+CMake
+CMakeLists
+CMakePackage
+CP
+CPC
+CPF
+CPP
+CPU
+CPUs
+CSC
+CSE
+CSV
+CSn
+CTests
+CU
+CUDA
+CUs
+CXX
+Cavium
+CentOS
+ChatGPT
+CoRR
+Codespaces
+Commitizen
+CommonMark
+Concretized
+Conda
+ConnectX
+DGEMM
+DKMS
+DL
+DMA
+DNN
+DNNL
+DPM
+DRI
+DW
+DWORD
+Dask
+DataFrame
+DataLoader
+DataParallel
+DeepSpeed
+Dependabot
+DevCap
+Dockerfile
+Doxygen
+ELMo
+ENDPGM
+EPYC
+ESXi
+FFT
+FFTs
+FFmpeg
+FHS
+FMA
+FP
+Filesystem
+Flang
+Fortran
+Fuyu
+GALB
+GCD
+GCDs
+GCN
+GDB
+GDDR
+GDR
+GDS
+GEMM
+GEMMs
+GFortran
+GiB
+GIM
+GL
+GLXT
+GMI
+GPG
+GPR
+GPT
+GPU
+GPU's
+GPUs
+GRBM
+GenAI
+GenZ
+GitHub
+Gitpod
+HBM
+HCA
+HIPCC
+HIPExtension
+HIPIFY
+HPC
+HPCG
+HPE
+HPL
+HSA
+HWE
+Haswell
+Higgs
+Hyperparameters
+ICV
+IDE
+IDEs
+IMDb
+IOMMU
+IOP
+IOPM
+IOV
+IRQ
+ISA
+ISV
+ISVs
+ImageNet
+InfiniBand
+Inlines
+IntelliSense
+Intersphinx
+Intra
+Ioffe
+JSON
+Jupyter
+KFD
+KiB
+KVM
+Keras
+Khronos
+LAPACK
+LCLK
+LDS
+LLM
+LLMs
+LLVM
+LM
+LSan
+LSAN
+LTS
+LoRA
+MEM
+MERCHANTABILITY
+MFMA
+MiB
+MIGraphX
+MIOpen
+MIOpenGEMM
+MIVisionX
+MLM
+MMA
+MMIO
+MMIOH
+MNIST
+MPI
+MSVC
+MVAPICH
+MVFFR
+Makefile
+Makefiles
+Matplotlib
+Megatron
+Mellanox
+Mellanox's
+Meta's
+MirroredStrategy
+Multicore
+Multithreaded
+MyEnvironment
+MyST
+NBIO
+NBIOs
+NIC
+NICs
+NLI
+NLP
+NPS
+NSP
+NUMA
+NVCC
+NVIDIA
+NVPTX
+NaN
+Nano
+Navi
+Noncoherently
+NousResearch's
+NumPy
+OAM
+OAMs
+OCP
+OEM
+OFED
+OMP
+OMPI
+OMPT
+OMPX
+ONNX
+OSS
+OSU
+OpenCL
+OpenCV
+OpenFabrics
+OpenGL
+OpenMP
+OpenSSL
+OpenVX
+PCI
+PCIe
+PEFT
+PIL
+PILImage
+PRNG
+PRs
+PaLM
+Pageable
+PeerDirect
+Perfetto
+PipelineParallel
+PnP
+PowerShell
+PyPi
+PyTorch
+Qcycles
+RAII
+RCCL
+RDC
+RDMA
+RDNA
+RHEL
+ROC
+ROCProfiler
+ROCTracer
+ROCclr
+ROCdbgapi
+ROCgdb
+ROCk
+ROCm
+ROCmCC
+ROCmSoftwarePlatform
+ROCmValidationSuite
+ROCr
+RST
+RW
+Radeon
+RelWithDebInfo
+Req
+Rickle
+RoCE
+Ryzen
+SALU
+SBIOS
+SCA
+SDK
+SDMA
+SDRAM
+SENDMSG
+SGPR
+SGPRs
+SHA
+SIGQUIT
+SIMD
+SIMDs
+SKU
+SKUs
+SLES
+SMEM
+SMI
+SMT
+SPI
+SQs
+SRAM
+SRAMECC
+SVD
+SWE
+SerDes
+Shlens
+Skylake
+Softmax
+Spack
+Supermicro
+Szegedy
+TCA
+TCC
+TCI
+TCIU
+TCP
+TCR
+TF
+TFLOPS
+TPU
+TPUs
+TensorBoard
+TensorFlow
+TensorParallel
+ToC
+TorchAudio
+TorchMIGraphX
+TorchScript
+TorchServe
+TorchVision
+TransferBench
+TrapStatus
+UAC
+UC
+UCC
+UCX
+UIF
+USM
+UTCL
+UTIL
+Uncached
+Unhandled
+VALU
+VBIOS
+VGPR
+VGPRs
+VM
+VMEM
+VMWare
+VRAM
+VSIX
+VSkipped
+Vanhoucke
+Vulkan
+WGP
+WGPs
+WX
+WikiText
+Wojna
+Workgroups
+Writebacks
+XCD
+XCDs
+XGBoost
+XGBoost's
+XGMI
+XT
+XTX
+Xeon
+Xilinx
+Xnack
+Xteam
+YAML
+YML
+YModel
+ZeRO
+ZenDNN
+accuracies
+activations
+addr
+alloc
+allocator
+allocators
+amdgpu
+api
+atmi
+atomics
+autogenerated
+avx
+awk
+backend
+backends
+benchmarking
+bfloat
+bilinear
+bitsandbytes
+blit
+boson
+bosons
+buildable
+bursty
+bzip
+cacheable
+cd
+centos
+centric
+changelog
+chiplet
+cmake
+cmd
+coalescable
+codename
+collater
+comgr
+completers
+composable
+concretization
+config
+conformant
+convolutional
+convolves
+cpp
+csn
+cuBLAS
+cuFFT
+cuLIB
+cuRAND
+cuSOLVER
+cuSPARSE
+dataset
+datasets
+dataspace
+datatype
+datatypes
+dbgapi
+de
+deallocation
+denoise
+denoised
+denoises
+denormalize
+deserializers
+detections
+dev
+devicelibs
+devsel
+dimensionality
+disambiguates
+distro
+el
+embeddings
+enablement
+endpgm
+encodings
+env
+epilog
+etcetera
+ethernet
+exascale
+executables
+ffmpeg
+filesystem
+fortran
+galb
+gcc
+gdb
+gfortran
+gfx
+githooks
+github
+gnupg
+grayscale
+gzip
+heterogenous
+hipBLAS
+hipBLASLt
+hipCUB
+hipFFT
+hipLIB
+hipRAND
+hipSOLVER
+hipSPARSE
+hipSPARSELt
+hipTensor
+hipamd
+hipblas
+hipcub
+hipfft
+hipfort
+hipify
+hipsolver
+hipsparse
+hpp
+hsa
+hsakmt
+hyperparameter
+ib_core
+inband
+incrementing
+inferencing
+inflight
+init
+initializer
+inlining
+installable
+interprocedural
+intra
+invariants
+invocating
+ipo
+kdb
+latencies
+libfabric
+libjpeg
+libs
+linearized
+linter
+linux
+llvm
+localscratch
+logits
+lossy
+macOS
+matchers
+microarchitecture
+migraphx
+miopen
+miopengemm
+mivisionx
+mkdir
+mlirmiopen
+mtypes
+mvffr
+namespace
+namespaces
+numref
+ocl
+opencl
+opencv
+openmp
+openssl
+optimizers
+os
+pageable
+parallelization
+parameterization
+passthrough
+perfcounter
+performant
+perl
+pragma
+pre
+prebuilt
+precompiled
+prefetch
+prefetchable
+preprocess
+preprocessed
+preprocessing
+prequantized
+prerequisites
+profiler
+protobuf
+pseudorandom
+py
+quasirandom
+queueing
+rccl
+rdc
+reStructuredText
+reformats
+repos
+representativeness
+req
+resampling
+rescaling
+reusability
+roadmap
+roc
+rocAL
+rocALUTION
+rocBLAS
+rocFFT
+rocLIB
+rocMLIR
+rocPRIM
+rocRAND
+rocSOLVER
+rocSPARSE
+rocThrust
+rocWMMA
+rocalution
+rocblas
+rocclr
+rocfft
+rocm
+rocminfo
+rocprim
+rocprof
+rocprofiler
+rocr
+rocrand
+rocsolver
+rocsparse
+rocthrust
+roctracer
+runtime
+runtimes
+sL
+scalability
+scalable
+sendmsg
+serializers
+shader
+sharding
+sigmoid
+sm
+smi
+softmax
+spack
+src
+stochastically
+strided
+subdirectory
+subexpression
+subfolder
+subfolders
+supercomputing
+tensorfloat
+th
+tokenization
+tokenize
+tokenized
+tokenizer
+tokenizes
+toolchain
+toolchains
+toolset
+toolsets
+torchvision
+tqdm
+tracebacks
+txt
+uarch
+uncached
+uncorrectable
+uninstallation
+unsqueeze
+unstacking
+unswitching
+untrusted
+untuned
+upvote
+USM
+UTCL
+UTIL
+utils
+vL
+variational
+vdi
+vectorizable
+vectorization
+vectorize
+vectorized
+vectorizer
+vectorizes
+vjxb
+walkthrough
+walkthroughs
+wavefront
+wavefronts
+whitespaces
+workgroup
+workgroups
+writeback
+writebacks
+wrreq
+wzo
+xargs
+xz
+yaml
+ysvmadyb
+zypper

CMakeLists.txt

@@ -0,0 +1,40 @@
+# MIT License
+#
+# Copyright (c) 2023 Advanced Micro Devices, Inc. All rights reserved.
+#
+# 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.
+
+cmake_minimum_required(VERSION 3.18.0)
+
+project(ROCm VERSION 5.7.1 LANGUAGES NONE)
+
+option(BUILD_DOCS "Build ROCm documentation" ON)
+
+include(GNUInstallDirs)
+
+# Adding default path cmake modules
+list(APPEND CMAKE_MODULE_PATH "${CMAKE_CURRENT_SOURCE_DIR}/cmake/Modules")
+
+# Handle dependencies
+include(Dependencies)
+
+# Build docs
+if(BUILD_DOCS)
+  add_subdirectory(docs)
+endif()

CONTRIBUTING.md

@@ -0,0 +1,94 @@
+<head>
+  <meta charset="UTF-8">
+  <meta name="description" content="Contributing to ROCm">
+  <meta name="keywords" content="ROCm, contributing, contribute, maintainer, contributor">
+</head>
+
+# Contribute to ROCm
+
+AMD values and encourages contributions to our code and documentation. If you want to contribute
+to our ROCm repositories, first review the following guidance. For documentation-specific information,
+see [Contributing to ROCm docs](https://rocm.docs.amd.com/en/latest/contribute/contributing.html).
+
+ROCm is a software stack made up of a collection of drivers, development tools, and APIs that enable
+GPU programming from low-level kernel to end-user applications. Because some of our components
+are inherited from external projects (such as
+[LLVM](https://github.com/ROCm/llvm-project) and
+[Kernel driver](https://github.com/ROCm/ROCK-Kernel-Driver)), these use
+project-specific contribution guidelines and workflow. Refer to their repositories for more information.
+All other ROCm components follow the workflow described in the following sections.
+
+## Development workflow
+
+ROCm uses GitHub to host code, collaborate, and manage version control. We use pull requests (PRs)
+for all changes within our repositories. We use
+[GitHub issues](https://github.com/ROCm/ROCm/issues) to track known issues, such as
+bugs.
+
+### Issue tracking
+
+Before filing a new issue, search the
+[existing issues](https://github.com/ROCm/ROCm/issues) to make sure your issue isn't
+already listed.
+
+General issue guidelines:
+
+* Use your best judgement for issue creation. If your issue is already listed, upvote the issue and
+  comment or post to provide additional details, such as how you reproduced this issue.
+* If you're not sure if your issue is the same, err on the side of caution and file your issue.
+  You can add a comment to include the issue number (and link) for the similar issue. If we evaluate
+  your issue as being the same as the existing issue, we'll close the duplicate.
+* If your issue doesn't exist, use the issue template to file a new issue.
+  * When filing an issue, be sure to provide as much information as possible, including script output so
+    we can collect information about your configuration. This helps reduce the time required to
+    reproduce your issue.
+  * Check your issue regularly, as we may require additional information to successfully reproduce the
+    issue.
+
+### Pull requests
+
+When you create a pull request, you should target the default branch.  Our repositories typically use the **develop** branch as the default integration branch.
+
+When creating a PR, use the following process. Note that each repository may include additional,
+project-specific steps. Refer to each repository's PR process for any additional steps.
+
+* Identify the issue you want to fix
+* Target the default branch (usually the **develop** branch) for integration
+* Ensure your code builds successfully
+* Each component has a suite of test cases to run; include the log of the successful test run in your PR
+* Do not break existing test cases
+* New functionality is only merged with new unit tests
+  * If your PR includes a new feature, you must provide an application or test so we can ensure that the
+    feature works and continues to be valid in the future
+* Tests must have good code coverage
+* Submit your PR and work with the reviewer or maintainer to get your PR approved
+* Once approved, the PR is brought onto internal CI systems and may be merged into the component
+  during our release cycle, as coordinated by the maintainer
+* We'll inform you once your change is committed
+
+:::{important}
+By creating a PR, you agree to allow your contribution to be licensed under the
+terms of the LICENSE.txt file in the corresponding repository. Different repositories may use different
+licenses.
+:::
+
+You can look up each license on the [ROCm licensing](https://rocm.docs.amd.com/en/latest/about/license.html) page.
+
+### New feature development
+
+Use the [GitHub Discussion forum](https://github.com/ROCm/ROCm/discussions)
+(Ideas category) to propose new features. Our maintainers are happy to provide direction and
+feedback on feature development.
+
+### Documentation
+
+Submit ROCm documentation changes to our
+[documentation repository](https://github.com/ROCm/ROCm). You must update
+documentation related to any new feature or API contribution.
+
+Note that each ROCm project uses its own repository for documentation.
+
+## Future development workflow
+
+The current ROCm development workflow is GitHub-based. If, in the future, we change this platform,
+the tools and links may change. In this instance, we will update contribution guidelines accordingly.

GOVERNANCE.md

@@ -0,0 +1,60 @@
+<head>
+  <meta charset="UTF-8">
+  <meta name="description" content="ROCm governance model">
+  <meta name="keywords" content="ROCm, governance">
+</head>
+
+# Governance model
+
+ROCm is a software stack made up of a collection of drivers, development tools, and APIs that enable
+GPU programming from the low-level kernel to end-user applications.
+
+Components of ROCm that are inherited from external projects (such as
+[LLVM](https://github.com/ROCm/llvm-project) and
+[Kernel driver](https://github.com/ROCm/ROCK-Kernel-Driver)) follow their own
+governance model and code of conduct. All other components of ROCm are governed by this
+document.
+
+## Governance
+
+ROCm is led and managed by AMD.
+
+We welcome contributions from the community. Our maintainers review all proposed changes to
+ROCm.
+
+## Roles
+
+* **Maintainers** are responsible for their designated component and repositories.
+* **Contributors** provide input and suggest changes to existing components.
+
+### Maintainers
+
+Maintainers are appointed by AMD. They are able to approve changes and can commit to our
+repositories. They must use pull requests (PRs) for all changes.
+
+You can find the list of maintainers in the CODEOWNERS file of each repository. Code owners differ
+between repositories.
+
+### Contributors
+
+If you're not a maintainer, you're a contributor. We encourage the ROCm community to contribute in
+several ways:
+
+* Help other community members by posting questions or solutions on our
+  [GitHub discussion forums](https://github.com/ROCm/ROCm/discussions)
+* Notify us of a bugs by filing an issue report on
+  [GitHub Issues](https://github.com/ROCm/ROCm/issues)
+* Improve our documentation by submitting a PR to our
+  [repository](https://github.com/ROCm/ROCm/)
+* Improve the code base (for smaller or contained changes) by submitting a PR to the component
+* Suggest larger features by adding to the *Ideas* category in the
+  [GitHub discussion forum](https://github.com/ROCm/ROCm/discussions)
+
+For more information, refer to our [contribution guidelines](CONTRIBUTING.md).
+
+## Code of conduct
+
+To engage with any AMD ROCm component that is hosted on GitHub, you must abide by the
+[GitHub community guidelines](https://docs.github.com/en/site-policy/github-terms/github-community-guidelines)
+and the
+[GitHub community code of conduct](https://docs.github.com/en/site-policy/github-terms/github-community-code-of-conduct).

LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2023 - 2024 Advanced Micro Devices, Inc. All rights reserved.
+
+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.

README.md

@@ -1,608 +1,99 @@
-# AMD ROCm™ v4.0.1 Patch Release Notes 
-
-## SUPPORTED OPERATING SYSTEMS
-
-### List of Supported Operating Systems
-
-The AMD ROCm platform is designed to support the following operating systems:
-
-* Ubuntu 20.04.1 (5.4 and 5.6-oem) and 18.04.5 (Kernel 5.4)	
-* CentOS 7.9 (3.10.0-1127) & RHEL 7.9 (3.10.0-1160.6.1.el7) (Using devtoolset-7 runtime support)
-* CentOS 8.3 (4.18.0-193.el8) and RHEL 8.3 (4.18.0-193.1.1.el8) (devtoolset is not required)
-* SLES 15 SP2
-
-### SUPPORT FOR RHEL V8.3
-
-This patch release extends support for RHEL v8.3. 
-
-### FRESH INSTALLATION OF AMD ROCM V4.0.1 RECOMMENDED
-
-A fresh and clean installation of AMD ROCm v4.0.1 is recommended. An upgrade from previous releases to AMD ROCm v4.0.1 is not supported. For more information, refer to the AMD ROCm Installation Guide.
-
-**Note**: AMD ROCm release v3.3 or prior releases are not fully compatible with AMD ROCm v3.5 and higher versions. You must perform a fresh ROCm installation if you want to upgrade from AMD ROCm v3.3 or older to 3.5 or higher versions and vice-versa.
-
-**Note**: *render* group is required only for Ubuntu v20.04. For all other ROCm supported operating systems, continue to use video group. 
-
-* For ROCm v3.5 and releases thereafter, the clinfo path is changed to /opt/rocm/opencl/bin/clinfo. 
-
-* For ROCm v3.3 and older releases, the clinfo path remains /opt/rocm/opencl/bin/x86_64/clinfo. 
- 
-### ROCM MULTI-VERSION INSTALLATION UPDATE
-
-With the AMD ROCm v4.0.1 release, the following ROCm multi-version installation changes apply:
-
-The meta packages rocm-dkms<version> are now deprecated for multi-version ROCm installs.  For example, rocm-dkms3.7.0, rocm-dkms3.8.0.
- 
-* Multi-version installation of ROCm should be performed by installing rocm-dev<version> using each of the desired ROCm versions. For example, rocm-dev3.7.0, rocm-dev3.8.0, rocm-dev3.9.0.   
-
-* Version files must be created for each multi-version rocm <= 4.0.1
-
- * Command: echo <version> | sudo tee /opt/rocm-<version>/.info/version
-
- * Example: echo 4.0.1 | sudo tee /opt/rocm-4.0.1/.info/version
-
-* The rock-dkms loadable kernel modules should be installed using a single rock-dkms package. 
-
-* ROCm v3.9 and above will not set any ldconfig entries for ROCm libraries for multi-version installation.  Users must set LD_LIBRARY_PATH to load the ROCm library version of choice.
-
-**NOTE**: The single version installation of the ROCm stack remains the same. The rocm-dkms package can be used for single version installs and is not deprecated at this time.
-
-
-
-# AMD ROCm™ Release Notes v4.0
-
-This page describes the features, fixed issues, and information about downloading and installing the ROCm software.
-It also covers known issues in this release.
-
-- [Supported Operating Systems and Documentation Updates](#Supported-Operating-Systems-and-Documentation-Updates)
-  * [Supported Operating Systems](#Supported-Operating-Systems)
-  * [ROCm Installation Updates](#ROCm-Installation-Updates)
-  * [AMD ROCm Documentation Updates](#AMD-ROCm-Documentation-Updates)
-   
-- [What\'s New in This Release](#Whats-New-in-This-Release)
-  * [INTRODUCING AMD INSTINCT MI100](#INTRODUCING-AMD-INSTINCT-MI100)
-  * [RAS Enhancements](#RAS-Enhancements)
-  * [Using CMake with AMD ROCm](#Using-CMake-with-AMD-ROCm)
-  * [AMD ROCm and Mesa Multimedia](#AMD-ROCm-and-Mesa-Multimedia)
-  * [ROCm System Management Information](#ROCm-System-Management-Information)
-  * [AMD GPU Debugger Enhancements](#AMD-GPU-Debugger-Enhancements)
-
-    
-- [Known Issues](#Known-Issues)
-
-- [Deprecations](#Deprecations)
-
-  * [Compiler Generated Code Object Version 2 Deprecation ](#Compiler-Generated-Code-Object-Version-2-Deprecation)
-  * [ROCr Runtime Deprecations](#ROCr-Runtime-Deprecations)
-  * [AOMP Deprecation](#AOMP-Deprecation)
-  
-
-- [Deploying ROCm](#Deploying-ROCm)
- 
-- [Hardware and Software Support](#Hardware-and-Software-Support)
-
-- [Machine Learning and High Performance Computing Software Stack for AMD GPU](#Machine-Learning-and-High-Performance-Computing-Software-Stack-for-AMD-GPU)
-  * [ROCm Binary Package Structure](#ROCm-Binary-Package-Structure)
-  * [ROCm Platform Packages](#ROCm-Platform-Packages)
-  
-
-
-# Supported Operating Systems 
-
-## List of Supported Operating Systems
-
-The AMD ROCm platform is designed to support the following operating systems:
-
-* Ubuntu 20.04.1 (5.4 and 5.6-oem) and 18.04.5 (Kernel 5.4)	
-
-* CentOS 7.8 (3.10.0-1127) & RHEL 7.9 (3.10.0-1160.6.1.el7) (Using devtoolset-7 runtime support)
-
-* CentOS 8.2 (4.18.0-193.el8) and RHEL 8.2 (4.18.0-193.1.1.el8) (devtoolset is not required)
-
-* SLES 15 SP2
-
- 
-# ROCm Installation Updates
-
-## Fresh Installation of AMD ROCm v4.0 Recommended
-
-A fresh and clean installation of AMD ROCm v4.0 is recommended. An upgrade from previous releases to AMD ROCm v4.0 is not supported.
-
-For more information, refer to the AMD ROCm Installation Guide at:
-https://rocmdocs.amd.com/en/latest/Installation_Guide/Installation-Guide.html
-
-**Note**: AMD ROCm release v3.3 or prior releases are not fully compatible with AMD ROCm v3.5 and higher versions. You must perform a fresh ROCm installation if you want to upgrade from AMD ROCm v3.3 or older to 3.5 or higher versions and vice-versa.
-
-**Note**: *render group* is required only for Ubuntu v20.04. For all other ROCm supported operating systems, continue to use *video group*.
-
-* For ROCm v3.5 and releases thereafter,the *clinfo* path is changed to -  */opt/rocm/opencl/bin/clinfo*.
-
-* For ROCm v3.3 and older releases, the *clinfo* path remains unchanged -  */opt/rocm/opencl/bin/x86_64/clinfo*.
-
-**Note**: After an operating system upgrade, AMD ROCm may upgrade automatically and result in an error. This is because AMD ROCm does not support upgrades currently. You must uninstall and reinstall AMD ROCm after an operating system upgrade.
-
-
-## ROCm MultiVersion Installation Update
-
-With the AMD ROCm v4.0 release, the following ROCm multi-version installation changes apply:
-
-The meta packages rocm-dkms<version> are now deprecated for multi-version ROCm installs. For example, rocm-dkms3.7.0, rocm-dkms3.8.0.
- 
-* Multi-version installation of ROCm should be performed by installing rocm-dev<version> using   each of the desired ROCm versions. For example, rocm-dev3.7.0, rocm-dev3.8.0, rocm-dev3.9.0.   
-* Version files must be created for each multi-version rocm <= 4.0.0
-
-    * command: echo <version> | sudo tee /opt/rocm-<version>/.info/version
-
-    * example: echo 4.0.0 | sudo tee /opt/rocm-4.0.0/.info/version
-
-* The rock-dkms loadable kernel modules should be installed using a single rock-dkms package. 
-
-* ROCm v3.9 and above will not set any *ldconfig* entries for ROCm libraries for multi-version installation.  Users must set *LD_LIBRARY_PATH* to load the ROCm library version of choice.
-
-
-**NOTE**: The single version installation of the ROCm stack remains the same. The rocm-dkms package can be used for single version installs and is not deprecated at this time.
-
-
-
-# AMD ROCm Documentation Updates
-
-## AMD ROCm Installation Guide 
-
-The AMD ROCm Installation Guide in this release includes:
-
-* Supported Environments
-
-* Installation Instructions for v4.0
-
-* HIP Installation Instructions 
-
-* AMD ROCm and Mesa Multimedia Installation 
-
-* Using CMake with AMD ROCm 
-
-For more information, refer to the ROCm documentation website at:
-
-https://rocmdocs.amd.com/en/latest/
-
-
-## AMD ROCm - HIP Documentation Updates
-
-* HIP Programming Guide v4.0 
-
-https://github.com/RadeonOpenCompute/ROCm/blob/master/HIP_Programming_Guide_v4.0.pdf
-
-* HIP API Guide v4.0
-
-https://github.com/RadeonOpenCompute/ROCm/blob/master/HIP-API_Guide_v4.0.pdf
-
-* HIP FAQ  
-
-For more information, refer to
-
-https://rocmdocs.amd.com/en/latest/Programming_Guides/HIP-FAQ.html#hip-faq
-
-
-## ROCm SMI API Documentation Updates 
-
-* xGMI API
-
-For more information, refer to the ROCm SMI API Guide at,
-
-https://github.com/RadeonOpenCompute/ROCm/blob/master/ROCm_SMI_API_Guide_v4.0.pdf
-
-
-## General AMD ROCm Documentation Links
-
-Access the following links for more information:
-
-* For AMD ROCm documentation, see 
-
-  https://rocmdocs.amd.com/en/latest/
-
-* For installation instructions on supped platforms, see
-
-  https://rocmdocs.amd.com/en/latest/Installation_Guide/Installation-Guide.html
-
-* For AMD ROCm binary structure, see
-
-  https://rocmdocs.amd.com/en/latest/Installation_Guide/Software-Stack-for-AMD-GPU.html
-  
-
-* For AMD ROCm Release History, see
-
- https://rocmdocs.amd.com/en/latest/Current_Release_Notes/ROCm-Version-History.html
-
-
-
-# What\'s New in This Release
-
-## INTRODUCING AMD INSTINCT MI100
-
-The AMD Instinct™ MI100 accelerator is the world’s fastest HPC GPU, and a culmination of the AMD CDNA architecture, with all-new Matrix Core Technology, and AMD ROCm™ open ecosystem to deliver new levels of performance, portability, and productivity. AMD CDNA is an all-new GPU architecture from AMD to drive accelerated computing into the era of exascale computing. The new architecture augments scalar and vector processing with new Matrix Core Engines and adds Infinity Fabric™ technology to scale up to larger systems. The open ROCm ecosystem puts customers in control and is a robust, mature platform that is easy to develop for and capable of running the most critical applications. The overall result is that the MI100 is the first GPU to break the 10TFLOP/s FP64 barrier designed as the steppingstone to the next generation of Exascale systems that will deliver pioneering discoveries in machine learning and scientific computing.
-
-
-### Key Features of AMD Instinct™ MI100 
-
-Important features of the AMD Instinct™ MI100 accelerator include:
-
-* Extended matrix core engine with Matrix Fused Multiply-Add (MFMA) for mixed-precision arithmetic and operates on KxN matrices (FP32, FP16, BF16, Int8) 
-
-* Added native support for the bfloat16 data type
-
-* 3 Infinity fabric connections per GPU enable a fully connected group of 4 GPUs in a ‘hive’ 
-
-![Screenshot](https://github.com/Rmalavally/ROCm/blob/master/images/keyfeatures.PNG)
-
-
-### Matrix Core Engines and GFX908 Considerations
-
-The AMD CDNA architecture builds on GCN’s foundation of scalars and vectors and adds matrices while simultaneously adding support for new numerical formats for machine learning and preserving backward compatibility for any software written for the GCN architecture. These Matrix Core Engines add a new family of wavefront-level instructions, the Matrix Fused MultiplyAdd or MFMA. The MFMA family performs mixed-precision arithmetic and operates on KxN matrices using four different types of input data: 8-bit integers (INT8), 16-bit half-precision FP (FP16), 16-bit brain FP (bf16), and 32-bit single-precision (FP32). All MFMA instructions produce either a 32-bit integer (INT32) or FP32 output, which reduces the likelihood of overflowing during the final accumulation stages of matrix multiplication.
-
-On nodes with gfx908, MFMA instructions are available to substantially speed up matrix operations. This hardware feature is used only in matrix multiplications functions in rocBLAS and supports only three base types f16_r, bf16_r, and f32_r. 
-
-* For half precision (f16_r and bf16_r) GEMM, use the function rocblas_gemm_ex, and set the compute_type parameter to f32_r.
-
-* For single precision (f32_r) GEMM, use the function rocblas_sgemm.
-
-* For single precision complex (f32_c) GEMM, use the function rocblas_cgemm.
-
-
-### References
-* For more information about bfloat16, see 
-
-https://rocblas.readthedocs.io/en/master/usermanual.html
-
-* For more details about AMD Instinct™ MI100 accelerator key features, see 
-
-https://www.amd.com/system/files/documents/instinct-mi100-brochure.pdf
-
-* For more information about the AMD Instinct MI100 accelerator, refer to the following sources:
-
-  - AMD CDNA whitepaper at https://www.amd.com/system/files/documents/amd-cdna-whitepaper.pdf
-  
-  - MI100 datasheet at https://www.amd.com/system/files/documents/instinct-mi100-brochure.pdf
-
-* AMD Instinct MI100/CDNA1 Shader Instruction Set Architecture (Dec. 2020) – This document describes the current environment, organization, and program state of AMD CDNA “Instinct MI100” devices. It details the instruction set and the microcode formats native to this family of processors that are accessible to programmers and compilers.
-
-https://developer.amd.com/wp-content/resources/CDNA1_Shader_ISA_14December2020.pdf
-
-
-## RAS Enhancements
-
-RAS (Reliability, Availability, and Accessibility) features provide help with data center GPU management. It is a method provided to users to track and manage data points via options implemented in the ROCm-SMI Command Line Interface (CLI) tool. 
-
-For more information about rocm-smi, see 
-
-https://github.com/RadeonOpenCompute/ROC-smi 
-
-The command options are wrappers of the system calls into the device driver interface as described here:
-
-https://dri.freedesktop.org/docs/drm/gpu/amdgpu.html#amdgpu-ras-support
-
-
-
-## Using CMake with AMD ROCm
-
-Most components in AMD ROCm support CMake 3.5 or higher out-of-the-box and do not require any special Find modules. A Find module is often used downstream to find the files by guessing locations of files with platform-specific hints. Typically, the Find module is required when the upstream is not built with CMake or the package configuration files are not available.
-
-AMD ROCm provides the respective config-file packages, and this enables find_package to be used directly. AMD ROCm does not require any Find module as the config-file packages are shipped with the upstream projects.
-
-For more information, see 
-
-https://rocmdocs.amd.com/en/latest/Installation_Guide/Using-CMake-with-AMD-ROCm.html
-
-
-## AMD ROCm and Mesa Multimedia
-
-AMD ROCm extends support to Mesa Multimedia. Mesa is an open-source software implementation of OpenGL, Vulkan, and other graphics API specifications. Mesa translates these specifications to vendor-specific graphics hardware drivers.
-
-For detailed installation instructions, refer to
-
-https://rocmdocs.amd.com/en/latest/Installation_Guide/Mesa-Multimedia-Installation.html
-
-
-## ROCm System Management Information 
-
-The following enhancements are made to ROCm System Management Interface (SMI).
-
-### Support for Printing PCle Information on AMD Instinct™100
-
-AMD ROCm extends support for printing PCle information on AMD Instinct MI100. 
-
-To check the pp_dpm_pcie file, use *"rocm-smi --showclocks"*.
-
-*/opt/rocm-4.0.0-6132/bin/rocm_smi.py  --showclocks*
-
-![Screenshot](https://github.com/Rmalavally/ROCm/blob/master/images/SMI.PNG)
-
-
-### New API for xGMI 
-
-Rocm_smi_lib now provides an API that exposes xGMI (inter-chip Global Memory Interconnect) throughput from one node to another. 
-
-Refer to the rocm_smi_lib API documentation for more details. 
-
-https://github.com/RadeonOpenCompute/ROCm/blob/master/ROCm_SMI_API_Guide_v4.0.pdf
-
-
-
-
-## AMD GPU Debugger Enhancements
-
-In this release, AMD GPU Debugger has the following enhancements:
-
-* ROCm v4.0 ROCgdb is based on gdb 10.1
-
-* Extended support for AMD Instinct™ MI100 
-
-
-
-
-# Known Issues 
-
-The following are the known issues in this release.
-
-## Upgrade to AMD ROCm v4.0 Not Supported
-
-An upgrade from previous releases to AMD ROCm v4.0 is not supported. A fresh and clean installation of AMD ROCm v4.0 is recommended. 
-
-
-
-# Deprecations
-
-This section describes deprecations and removals in AMD ROCm.
-
-## Compiler Generated Code Object Version 2 Deprecation 
-
-**WARNING**
-
-Compiler-generated code object version 2 is no longer supported and will be removed shortly. AMD ROCm users must plan for the code object version 2 deprecation immediately. 
-
-Support for loading code object version 2 is also being deprecated with no announced removal release.
-
-## ROCr Runtime Deprecations
-
-The following ROCr Runtime enumerations, functions, and structs are deprecated in the AMD ROCm v4.0 release.
-
-### Deprecated ROCr Runtime Functions
-
-* hsa_isa_get_info
-
-* hsa_isa_compatible
-
-* hsa_executable_create
-
-* hsa_executable_get_symbol
-
-* hsa_executable_iterate_symbols
-
-* hsa_code_object_serialize
-
-* hsa_code_object_deserialize
-
-* hsa_code_object_destroy
-
-* hsa_code_object_get_info
-
-* hsa_executable_load_code_object
-
-* hsa_code_object_get_symbol
-
-* hsa_code_object_get_symbol_from_name
-
-* hsa_code_symbol_get_info
-
-* hsa_code_object_iterate_symbols
-
-
-### Deprecated ROCr Runtime Enumerations
-
-* HSA_ISA_INFO_CALL_CONVENTION_COUNT
-
-* HSA_ISA_INFO_CALL_CONVENTION_INFO_WAVEFRONT_SIZE
-
-* HSA_ISA_INFO_CALL_CONVENTION_INFO_WAVEFRONTS_PER_COMPUTE_UNIT
-
-* HSA_EXECUTABLE_SYMBOL_INFO_MODULE_NAME_LENGTH
-
-* HSA_EXECUTABLE_SYMBOL_INFO_MODULE_NAME
-
-* HSA_EXECUTABLE_SYMBOL_INFO_AGENT
-
-* HSA_EXECUTABLE_SYMBOL_INFO_VARIABLE_ALLOCATION
-
-* HSA_EXECUTABLE_SYMBOL_INFO_VARIABLE_SEGMENT
-
-* HSA_EXECUTABLE_SYMBOL_INFO_VARIABLE_ALIGNMENT
-
-* HSA_EXECUTABLE_SYMBOL_INFO_VARIABLE_SIZE
-
-* HSA_EXECUTABLE_SYMBOL_INFO_VARIABLE_IS_CONST
-
-* HSA_EXECUTABLE_SYMBOL_INFO_KERNEL_CALL_CONVENTION
-
-* HSA_EXECUTABLE_SYMBOL_INFO_INDIRECT_FUNCTION_CALL_CONVENTION
-
-   - hsa_code_object_type_t
-   
-   - hsa_code_object_info_t
-   
-   - hsa_code_symbol_info_t
-
-
-### Deprecated ROCr Runtime Structs
-
-* hsa_code_object_t
-
-* hsa_callback_data_t
-
-* hsa_code_symbol_t
-
-
-
-## AOMP Deprecation
-
-As of AMD ROCm v4.0, AOMP (aomp-amdgpu) is deprecated. OpenMP support has moved to the openmp-extras auxiliary package, which leverages the ROCm compiler on LLVM 12.
-
-For more information, refer to 
-
-https://rocmdocs.amd.com/en/latest/Programming_Guides/openmp_support.html
-
-
-# Deploying ROCm
-
-AMD hosts both Debian and RPM repositories for the ROCm v4.0.0 packages. 
-
-For more information on ROCM installation on all platforms, see
-
-https://rocmdocs.amd.com/en/latest/Installation_Guide/Installation-Guide.html
-
-
-## Machine Learning and High Performance Computing Software Stack for AMD GPU
-
-For an updated version of the software stack for AMD GPU, see
-
-https://rocmdocs.amd.com/en/latest/Installation_Guide/Installation-Guide.html#software-stack-for-amd-gpu
-
-
-
-# Hardware and Software Support
-ROCm is focused on using AMD GPUs to accelerate computational tasks such as machine learning, engineering workloads, and scientific computing.
-In order to focus our development efforts on these domains of interest, ROCm supports a targeted set of hardware configurations which are detailed further in this section.
-
-#### Supported GPUs
-Because the ROCm Platform has a focus on particular computational domains, we offer official support for a selection of AMD GPUs that are designed to offer good performance and price in these domains.
-
-**Note:** The integrated GPUs of Ryzen are not officially supported targets for ROCm.
-
-ROCm officially supports AMD GPUs that use following chips:
-
-* GFX9 GPUs
-
-  - "Vega 10" chips, such as on the AMD Radeon RX Vega 64 and Radeon Instinct MI25
-  
-  - "Vega 7nm" chips, such as on the Radeon Instinct MI50, Radeon Instinct MI60 or AMD Radeon VII, 
-
-* CDNA GPUs
-
-  - MI100 chips such as on the AMD Instinct™ MI100
-
-
-ROCm is a collection of software ranging from drivers and runtimes to libraries and developer tools.
-Some of this software may work with more GPUs than the "officially supported" list above, though AMD does not make any official claims of support for these devices on the ROCm software platform.
-The following list of GPUs are enabled in the ROCm software, though full support is not guaranteed:
-
-  * GFX8 GPUs
-    * "Polaris 11" chips, such as on the AMD Radeon RX 570 and Radeon Pro WX 4100
-    * "Polaris 12" chips, such as on the AMD Radeon RX 550 and Radeon RX 540
-  * GFX7 GPUs
-    * "Hawaii" chips, such as the AMD Radeon R9 390X and FirePro W9100
-
-As described in the next section, GFX8 GPUs require PCI Express 3.0 (PCIe 3.0) with support for PCIe atomics. This requires both CPU and motherboard support. GFX9 GPUs require PCIe 3.0 with support for PCIe atomics by default, but they can operate in most cases without this capability.
-
-The integrated GPUs in AMD APUs are not officially supported targets for ROCm.
-As described [below](#limited-support), "Carrizo", "Bristol Ridge", and "Raven Ridge" APUs are enabled in our upstream drivers and the ROCm OpenCL runtime.
-However, they are not enabled in the HIP runtime, and may not work due to motherboard or OEM hardware limitations.
-As such, they are not yet officially supported targets for ROCm.
-
-For a more detailed list of hardware support, please see [the following documentation](https://en.wikipedia.org/wiki/List_of_AMD_graphics_processing_units).
-
-#### Supported CPUs
-As described above, GFX8 GPUs require PCIe 3.0 with PCIe atomics in order to run ROCm.
-In particular, the CPU and every active PCIe point between the CPU and GPU require support for PCIe 3.0 and PCIe atomics.
-The CPU root must indicate PCIe AtomicOp Completion capabilities and any intermediate switch must indicate PCIe AtomicOp Routing capabilities.
-
-Current CPUs which support PCIe Gen3 + PCIe Atomics are:
-
-  * AMD Ryzen CPUs
-  * The CPUs in AMD Ryzen APUs
-  * AMD Ryzen Threadripper CPUs
-  * AMD EPYC CPUs
-  * Intel Xeon E7 v3 or newer CPUs
-  * Intel Xeon E5 v3 or newer CPUs
-  * Intel Xeon E3 v3 or newer CPUs
-  * Intel Core i7 v4, Core i5 v4, Core i3 v4 or newer CPUs (i.e. Haswell family or newer)
-  * Some Ivy Bridge-E systems
-
-Beginning with ROCm 1.8, GFX9 GPUs (such as Vega 10) no longer require PCIe atomics.
-We have similarly opened up more options for number of PCIe lanes.
-GFX9 GPUs can now be run on CPUs without PCIe atomics and on older PCIe generations, such as PCIe 2.0.
-This is not supported on GPUs below GFX9, e.g. GFX8 cards in the Fiji and Polaris families.
-
-If you are using any PCIe switches in your system, please note that PCIe Atomics are only supported on some switches, such as Broadcom PLX.
-When you install your GPUs, make sure you install them in a PCIe 3.1.0 x16, x8, x4, or x1 slot attached either directly to the CPU's Root I/O controller or via a PCIe switch directly attached to the CPU's Root I/O controller.
-
-In our experience, many issues stem from trying to use consumer motherboards which provide physical x16 connectors that are electrically connected as e.g. PCIe 2.0 x4, PCIe slots connected via the Southbridge PCIe I/O controller, or PCIe slots connected through a PCIe switch that does
-not support PCIe atomics.
-
-If you attempt to run ROCm on a system without proper PCIe atomic support, you may see an error in the kernel log (`dmesg`):
-```
-kfd: skipped device 1002:7300, PCI rejects atomics
-```
-
-Experimental support for our Hawaii (GFX7) GPUs (Radeon R9 290, R9 390, FirePro W9100, S9150, S9170)
-does not require or take advantage of PCIe Atomics. However, we still recommend that you use a CPU
-from the list provided above for compatibility purposes.
-
-#### Not supported or limited support under ROCm
-
-##### Limited support
-
-* ROCm 2.9.x should support PCIe 2.0 enabled CPUs such as the AMD Opteron, Phenom, Phenom II, Athlon, Athlon X2, Athlon II and older Intel Xeon and Intel Core Architecture and Pentium CPUs. However, we have done very limited testing on these configurations, since our test farm has been catering to CPUs listed above. This is where we need community support. _If you find problems on such setups, please report these issues_.
-* Thunderbolt 1, 2, and 3 enabled breakout boxes should now be able to work with ROCm. Thunderbolt 1 and 2 are PCIe 2.0 based, and thus are only supported with GPUs that do not require PCIe 3.1.0 atomics (e.g. Vega 10). However, we have done no testing on this configuration and would need community support due to limited access to this type of equipment.
-* AMD "Carrizo" and "Bristol Ridge" APUs are enabled to run OpenCL, but do not yet support HIP or our libraries built on top of these compilers and runtimes.
-  * As of ROCm 2.1, "Carrizo" and "Bristol Ridge" require the use of upstream kernel drivers.
-  * In addition, various "Carrizo" and "Bristol Ridge" platforms may not work due to OEM and ODM choices when it comes to key configurations parameters such as inclusion of the required CRAT tables and IOMMU configuration parameters in the system BIOS.
-  * Before purchasing such a system for ROCm, please verify that the BIOS provides an option for enabling IOMMUv2 and that the system BIOS properly exposes the correct CRAT table. Inquire with your vendor about the latter.
-* AMD "Raven Ridge" APUs are enabled to run OpenCL, but do not yet support HIP or our libraries built on top of these compilers and runtimes.
-  * As of ROCm 2.1, "Raven Ridge" requires the use of upstream kernel drivers.
-  * In addition, various "Raven Ridge" platforms may not work due to OEM and ODM choices when it comes to key configurations parameters such as inclusion of the required CRAT tables and IOMMU configuration parameters in the system BIOS.
-  * Before purchasing such a system for ROCm, please verify that the BIOS provides an option for enabling IOMMUv2 and that the system BIOS properly exposes the correct CRAT table. Inquire with your vendor about the latter.
-
-##### Not supported
-
-* "Tonga", "Iceland", "Vega M", and "Vega 12" GPUs are not supported in ROCm 2.9.x
-* We do not support GFX8-class GPUs (Fiji, Polaris, etc.) on CPUs that do not have PCIe 3.0 with PCIe atomics.
-  * As such, we do not support AMD Carrizo and Kaveri APUs as hosts for such GPUs.
-  * Thunderbolt 1 and 2 enabled GPUs are not supported by GFX8 GPUs on ROCm. Thunderbolt 1 & 2 are based on PCIe 2.0.
-
-#### ROCm support in upstream Linux kernels
-
-As of ROCm 1.9.0, the ROCm user-level software is compatible with the AMD drivers in certain upstream Linux kernels.
-As such, users have the option of either using the ROCK kernel driver that are part of AMD's ROCm repositories or using the upstream driver and only installing ROCm user-level utilities from AMD's ROCm repositories.
-
-These releases of the upstream Linux kernel support the following GPUs in ROCm:
- * 4.17: Fiji, Polaris 10, Polaris 11
- * 4.18: Fiji, Polaris 10, Polaris 11, Vega10
- * 4.20: Fiji, Polaris 10, Polaris 11, Vega10, Vega 7nm
-
-The upstream driver may be useful for running ROCm software on systems that are not compatible with the kernel driver available in AMD's repositories.
-For users that have the option of using either AMD's or the upstreamed driver, there are various tradeoffs to take into consideration:
-
-|   | Using AMD's `rock-dkms` package | Using the upstream kernel driver |
-| ---- | ------------------------------------------------------------| ----- |
-| Pros | More GPU features, and they are enabled earlier | Includes the latest Linux kernel features |
-|      | Tested by AMD on supported distributions | May work on other distributions and with custom kernels |
-|      | Supported GPUs enabled regardless of kernel version | |
-|      | Includes the latest GPU firmware | |
-| Cons | May not work on all Linux distributions or versions | Features and hardware support varies depending on kernel version |
-|      | Not currently supported on kernels newer than 5.4 | Limits GPU's usage of system memory to 3/8 of system memory (before 5.6). For 5.6 and beyond, both DKMS and upstream kernels allow use of 15/16 of system memory. |
-|      |   | IPC and RDMA capabilities are not yet enabled |
-|      |   | Not tested by AMD to the same level as `rock-dkms` package |
-|      |   | Does not include most up-to-date firmware |
-
-
-# Disclaimer
-
-AMD®, the AMD Arrow logo, AMD Instinct™, Radeon™, ROCm® and combinations thereof are trademarks of Advanced Micro Devices, Inc.
-
-Linux® is the registered trademark of Linus Torvalds in the U.S. and other countries.
-
-PCIe® is a registered trademark of PCI-SIG Corporation. Other product names used in this publication are for identification purposes only and may be trademarks of their respective companies.  
-
-Google®  is a registered trademark of Google LLC.
-
-Ubuntu and the Ubuntu logo are registered trademarks of Canonical Ltd.
-
-Other product names used in this publication are for identification purposes only and may be trademarks of their respective companies.
-
+# AMD ROCm Software
+
+ROCm is an open-source stack, composed primarily of open-source software, designed for graphics
+processing unit (GPU) computation. ROCm consists of a collection of drivers, development tools, and
+APIs that enable GPU programming from low-level kernel to end-user applications.
+
+With ROCm, you can customize your GPU software to meet your specific needs. You can develop,
+collaborate, test, and deploy your applications in a free, open source, integrated, and secure software
+ecosystem. ROCm is particularly well-suited to GPU-accelerated high-performance computing (HPC),
+artificial intelligence (AI), scientific computing, and computer aided design (CAD).
+
+ROCm is powered by AMD’s
+[Heterogeneous-computing Interface for Portability (HIP)](https://github.com/ROCm/HIP),
+an open-source software C++ GPU programming environment and its corresponding runtime. HIP
+allows ROCm developers to create portable applications on different platforms by deploying code on a
+range of platforms, from dedicated gaming GPUs to exascale HPC clusters.
+
+ROCm supports programming models, such as OpenMP and OpenCL, and includes all necessary open
+source software compilers, debuggers, and libraries. ROCm is fully integrated into machine learning
+(ML) frameworks, such as PyTorch and TensorFlow.
+
+## Getting the ROCm Source Code
+
+AMD ROCm is built from open source software. It is, therefore, possible to modify the various components of ROCm by downloading the source code and rebuilding the components. The source code for ROCm components can be cloned from each of the GitHub repositories using git.  For easy access to download the correct versions of each of these tools, the ROCm repository contains a repo manifest file called [default.xml](./default.xml). You can use this manifest file to download the source code for ROCm software.
+
+### Installing the repo tool
+
+The repo tool from Google allows you to manage multiple git repositories simultaneously. Run the following commands to install the repo tool:
+
+```bash
+mkdir -p ~/bin/
+curl https://storage.googleapis.com/git-repo-downloads/repo > ~/bin/repo
+chmod a+x ~/bin/repo
+```
+
+**Note:** The ```~/bin/``` folder is used as an example. You can specify a different folder to install the repo tool into if you desire.
+
+### Installing git-lfs
+
+Some ROCm projects use the Git Large File Storage (LFS) format that may require you to install git-lfs. Refer to [Git Large File Storage](https://github.com/git-lfs/git-lfs/blob/main/INSTALLING.md) for more information. For example, to install git-lfs for Ubuntu, use the following command:
+
+```bash
+sudo apt-get install git-lfs
+```
+
+### Downloading the ROCm source code
+
+The following example shows how to use the repo tool to download the ROCm source code. If you choose a directory other than ~/bin/ to install the repo tool, you must use that chosen directory in the code as shown below:
+
+```bash
+mkdir -p ~/ROCm/
+cd ~/ROCm/
+~/bin/repo init -u http://github.com/ROCm/ROCm.git -b roc-6.0.x
+~/bin/repo sync
+```
+
+**Note:** Using this sample code will cause the repo tool to download the open source code associated with the specified ROCm release. Ensure that you have ssh-keys configured on your machine for your GitHub ID prior to the download as explained at [Connecting to GitHub with SSH](https://docs.github.com/en/authentication/connecting-to-github-with-ssh).
+
+### Building the ROCm source code
+
+Each ROCm component repository contains directions for building that component, such as the rocSPARSE documentation [Installation and Building for Linux](https://rocm.docs.amd.com/projects/rocSPARSE/en/latest/install/Linux_Install_Guide.html). Refer to the specific component documentation for instructions on building the repository.
+
+Each release of the ROCm software supports specific hardware and software configurations. Refer to [System requirements (Linux)](https://rocm.docs.amd.com/projects/install-on-linux/en/latest/reference/system-requirements.html) for the current supported hardware and OS.
+
+## ROCm documentation
+
+This repository contains the [manifest file](https://gerrit.googlesource.com/git-repo/+/HEAD/docs/manifest-format.md)
+for ROCm releases, changelogs, and release information.
+
+The `default.xml` file contains information for all repositories and the associated commit used to build
+the current ROCm release; `default.xml` uses the [Manifest Format repository](https://gerrit.googlesource.com/git-repo/).
+
+Source code for our documentation is located in the `/docs` folder of most ROCm repositories. The
+`develop` branch of our repositories contains content for the next ROCm release.
+
+The ROCm documentation homepage is [rocm.docs.amd.com](https://rocm.docs.amd.com).
+
+### Building the documentation
+
+For a quick-start build, use the following code. For more options and detail, refer to
+[Building documentation](./docs/contribute/building.md).
+
+```bash
+cd docs
+pip3 install -r sphinx/requirements.txt
+python3 -m sphinx -T -E -b html -d _build/doctrees -D language=en . _build/html
+```
+
+Alternatively, CMake build is supported.
+
+```bash
+cmake -B build
+cmake --build build --target=doc
+```
+
+## Older ROCm releases
+
+For release information for older ROCm releases, refer to the
+[CHANGELOG](./CHANGELOG.md).

RELEASE.md

@@ -0,0 +1,252 @@
+# ROCm 6.1 release highlights
+<!-- Disable lints since this is an auto-generated file.    -->
+<!-- markdownlint-disable blanks-around-headers             -->
+<!-- markdownlint-disable no-duplicate-header               -->
+<!-- markdownlint-disable no-blanks-blockquote              -->
+<!-- markdownlint-disable ul-indent                         -->
+<!-- markdownlint-disable no-trailing-spaces                -->
+
+<!-- spellcheck-disable -->
+
+The ROCm™ 6.1 release consists of new features and fixes to improve the stability and
+performance of AMD Instinct™ MI300 GPU applications. Notably, we've added:
+
+* Full support for Ubuntu 22.04.4.
+
+* **rocDecode**, a new ROCm component that provides high-performance video decode support for
+  AMD GPUs. With rocDecode, you can decode compressed video streams while keeping the resulting
+  YUV frames in video memory. With decoded frames in video memory, you can run video
+  post-processing using ROCm HIP, avoiding unnecessary data copies via the PCIe bus.
+
+  To learn more, refer to the rocDecode 
+  [documentation](https://rocm.docs.amd.com/projects/rocDecode/en/latest/).
+
+## OS and GPU support changes
+
+ROCm 6.1 adds the following operating system support:
+
+* MI300A: Ubuntu 22.04.4 and RHEL 9.3
+* MI300X: Ubuntu 22.04.4
+
+Future releases will add additional operating systems to match the general offering. For older
+generations of supported AMD Instinct products, we’ve added Ubuntu 22.04.4 support.
+
+```{tip}
+To view the complete list of supported GPUs and operating systems, refer to the system requirements
+page for
+[Linux](https://rocm.docs.amd.com/projects/install-on-linux/en/latest/reference/system-requirements.html)
+and
+[Windows](https://rocm.docs.amd.com/projects/install-on-windows/en/latest/reference/system-requirements.html).
+```
+
+## Installation packages
+
+This release includes a new set of packages for every module (all libraries and binaries default to
+`DT_RPATH`). Package names have the suffix `rpath`; for example, the `rpath` variant of `rocminfo` is
+`rocminfo-rpath`.
+
+```{warning}
+The new `rpath` packages will conflict with the default packages; they are meant to be used only in
+environments where legacy `DT_RPATH` is the preferred form of linking (instead of `DT_RUNPATH`). We
+do **not** recommend installing both sets of packages.
+```
+
+## ROCm components
+
+The following sections highlight select component-specific changes. For additional details, refer to the
+[Changelog](https://rocm.docs.amd.com/en/develop/about/CHANGELOG.html).
+
+### AMD System Management Interface (SMI) Tool
+
+* **New monitor command for GPU metrics**.
+  Use the monitor command to customize, capture, collect, and observe GPU metrics on
+  target devices.
+
+* **Integration with E-SMI**.
+  The EPYC™ System Management Interface In-band Library is a Linux C-library that provides in-band
+  user space software APIs to monitor and control your CPU’s power, energy, performance, and other
+  system management functionality. This integration enables access to CPU metrics and telemetry
+  through the AMD SMI API and CLI tools.
+
+### Composable Kernel (CK)
+
+* **New architecture support**.
+  CK now supports to the following architectures to enable efficient image denoising on the following
+  AMD GPUs: gfx1030, gfx1100, gfx1031, gfx1101, gfx1032, gfx1102, gfx1034, gfx1103, gfx1035,
+  gfx1036
+
+* **FP8 rounding logic is replaced with stochastic rounding**.
+  Stochastic rounding mimics a more realistic data behavior and improves model convergence.
+
+### HIP
+
+* **New environment variable to enable kernel run serialization**.
+  The default `HIP_LAUNCH_BLOCKING` value is `0` (disable); which causes kernels to run as defined in
+  the queue. When set to `1` (enable), the HIP runtime serializes the kernel queue, which behaves the
+  same as `AMD_SERIALIZE_KERNEL`.
+
+### hipBLASLt
+
+* **New GemmTuning extension parameter** GemmTuning allows you to set a split-k value for each solution, which is more feasible for
+  performance tuning.
+
+### hipFFT
+
+* **New multi-GPU support for single-process transforms** Multiple GPUs can be used to perform a transform in a single process. Note that this initial
+  implementation is a functional preview.
+
+### HIPIFY
+
+* **Skipped code blocks**: Code blocks that are skipped by the preprocessor are no longer hipified under the
+  `--default-preprocessor` option. To hipify everything, despite conditional preprocessor directives
+  (`#if`, `#ifdef`, `#ifndef`, `#elif`, or `#else`), don't use the `--default-preprocessor` or `--amap` options.
+
+### hipSPARSELt
+
+* **Structured sparsity matrix support extensions**
+  Structured sparsity matrices help speed up deep-learning workloads. We now support `B` as the
+  sparse matrix and `A` as the dense matrix in Sparse Matrix-Matrix Multiplication (SPMM). Prior to this
+  release, we only supported sparse (matrix A) x dense (matrix B) matrix multiplication. Structured
+  sparsity matrices help speed up deep learning workloads.
+
+### hipTensor
+
+* **4D tensor permutation and contraction support**.
+  You can now perform tensor permutation on 4D tensors and 4D contractions for F16, BF16, and
+  Complex F32/F64 datatypes.
+
+### MIGraphX
+
+* **Improved performance for transformer-based models**.
+  We added support for FlashAttention, which benefits models like BERT, GPT, and Stable Diffusion.
+
+* **New Torch-MIGraphX driver**.
+  This driver calls MIGraphX directly from PyTorch. It provides an `mgx_module` object that you can
+  invoke like any other Torch module, but which utilizes the MIGraphX inference engine internally.
+  Torch-MIGraphX supports FP32, FP16, and INT8 datatypes.
+
+  * **FP8 support**. We now offer functional support for inference in the FP8E4M3FNUZ datatype. You
+  can load an ONNX model in FP8E4M3FNUZ using C++ or Python APIs, or `migraphx-driver`.
+  You can quantize a floating point model to FP8 format by using the `--fp8` flag with `migraphx-driver`.
+  To accelerate inference, MIGraphX uses hardware acceleration on MI300 for FP8 by leveraging FP8
+  support in various backend kernel libraries.
+
+### MIOpen
+
+* **Improved performance for inference and convolutions**.
+  Inference support now provided for Find 2.0 fusion plans. Additionally, we've enhanced the Number of
+  samples, Height, Width, and Channels (NHWC) convolution kernels for heuristics. NHWC stores data
+  in a format where the height and width dimensions come first, followed by channels.
+
+### OpenMP
+
+* **Implicit Zero-copy is triggered automatically in XNACK-enabled MI300A systems**.
+  Implicit Zero-copy behavior in `non unified_shared_memory` programs is triggered automatically in
+  XNACK-enabled MI300A systems (for example, when using the `HSA_XNACK=1` environment
+  variable). OpenMP supports the 'requires `unified_shared_memory`' directive to support programs
+  that don’t want to copy data explicitly between the CPU and GPU. However, this requires that you add
+  these directives to every translation unit of the program.
+
+* **New MI300 FP atomics**. Application performance can now improve by leveraging fast floating-point atomics on MI300 (gfx942).
+  
+
+### RCCL
+
+* **NCCL 2.18.6 compatibility**.
+  RCCL is now compatible with NCCL 2.18.6, which includes increasing the maximum IB network interfaces to 32 and fixing network device ordering when creating communicators with only one GPU
+  per node.
+
+* **Doubled simultaneous communication channels**.
+  We improved MI300X performance by increasing the maximum number of simultaneous
+  communication channels from 32 to 64.
+
+### rocALUTION
+
+* **New multiple node and GPU support**.
+  Unsmoothed and smoothed aggregations and Ruge-Stueben AMG now work with multiple nodes
+  and GPUs. For more information, refer to the 
+  [API documentation](https://rocm.docs.amd.com/projects/rocALUTION/en/latest/usermanual/solvers.html#unsmoothed-aggregation-amg).
+
+### rocDecode
+
+* **New ROCm component**.
+  rocDecode ROCm's newest component, providing high-performance video decode support for AMD
+  GPUs. To learn more, refer to the 
+  [documentation](https://rocm.docs.amd.com/projects/rocDecode/en/latest/).
+
+### ROCm Compiler
+
+* **Combined projects**. ROCm Device-Libs, ROCm Compiler Support, and hipCC are now located in
+  the `llvm-project/amd` subdirectory of AMD's fork of the LLVM project. Previously, these projects
+  were maintained in separate repositories. Note that the projects themselves will continue to be
+  packaged separately.
+
+* **Split the 'rocm-llvm' package**. This package has been split into a required and an optional package: 
+
+  * **rocm-llvm(required)**: A package containing the essential binaries needed for compilation.
+  
+  * **rocm-llvm-dev(optional)**: A package containing binaries for compiler and application developers.
+    
+
+### ROCm Data Center Tool (RDC)
+
+* **C++ upgrades**.
+  RDC was upgraded from C++11 to C++17 to enable a more modern C++ standard when writing RDC plugins.
+
+### ROCm Performance Primitives (RPP)
+
+* **New backend support**.
+  Audio processing support added for the `HOST` backend and 3D Voxel kernels support
+  for the `HOST` and `HIP` backends.
+
+### ROCm Validation Suite
+
+* **New datatype support**.
+Added BF16 and FP8 datatypes based on General Matrix Multiply(GEMM) operations in the GPU Stress Test (GST) module. This provides additional performance benchmarking and stress testing based on the newly supported datatypes.
+
+### rocSOLVER
+
+* **New EigenSolver routine**.
+Based on the Jacobi algorithm, a new EigenSolver routine was added to the library. This routine computes the eigenvalues and eigenvectors of a matrix with improved performance.
+
+### ROCTracer
+
+* **New versioning and callback enhancements**.
+Improved to match versioning changes in HIP Runtime and supports runtime API callbacks and activity record logging. The APIs of different runtimes at different levels are considered different API domains with assigned domain IDs.
+
+## Upcoming changes
+
+* ROCm SMI will be deprecated in a future release. We advise **migrating to AMD SMI** now to
+  prevent future workflow disruptions.
+
+* hipCC supports, by default, the following compiler invocation flags:
+
+  * `-mllvm -amdgpu-early-inline-all=true`
+  * `-mllvm -amdgpu-function-calls=false`
+
+  In a future ROCm release, hipCC will no longer support these flags. It will, instead, use the Clang
+  defaults:
+
+  * `-mllvm -amdgpu-early-inline-all=false`
+  * `-mllvm -amdgpu-function-calls=true`
+
+  To evaluate the impact of this change, include `--hipcc-func-supp` in your hipCC invocation.
+
+  For information on these flags, and the differences between hipCC and Clang, refer to
+  [ROCm Compiler Interfaces](https://rocm.docs.amd.com/en/latest/reference/rocmcc.html#rocm-compiler-interfaces).
+
+*  Future ROCm releases will not provide `clang-ocl`. For more information, refer to the
+  [`clang-ocl` README](https://github.com/ROCm/clang-ocl).
+
+* The following operating systems will be supported in a future ROCm release. They are currently
+  only available in beta.
+
+  * RHEL 9.4
+  * RHEL 8.10
+  * SLES 15 SP6
+
+* As of ROCm 6.2, we’ve planned for **end-of-support** for:
+
+  * Ubuntu 20.04.5
+  * SLES 15 SP4
+  * RHEL/CentOS 7.9

cmake/Modules/Dependencies.cmake

@@ -0,0 +1,47 @@
+# MIT License
+#
+# Copyright (c) 2023 Advanced Micro Devices, Inc. All rights reserved.
+#
+# 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.
+
+# ###########################
+# ROCm dependencies
+# ###########################
+
+include(FetchContent)
+
+if(BUILD_DOCS)
+  find_package(ROCM 0.11.0 CONFIG QUIET PATHS "${ROCM_PATH}") # First version with Sphinx doc gen improvement
+  if(NOT ROCM_FOUND)
+    message(STATUS "ROCm CMake not found. Fetching...")
+    set(rocm_cmake_tag
+      "c044bb52ba85058d28afe2313be98d9fed02e293" # develop@2023.09.12. (move to 6.0 tag when released)
+      CACHE STRING "rocm-cmake tag to download")
+    FetchContent_Declare(
+      rocm-cmake
+      GIT_REPOSITORY https://github.com/ROCm/rocm-cmake.git
+      GIT_TAG        ${rocm_cmake_tag}
+      SOURCE_SUBDIR "DISABLE ADDING TO BUILD" # We don't really want to consume the build and test targets of ROCm CMake.
+    )
+    FetchContent_MakeAvailable(rocm-cmake)
+    find_package(ROCM CONFIG REQUIRED NO_DEFAULT_PATH PATHS "${rocm-cmake_SOURCE_DIR}")
+  else()
+    find_package(ROCM 0.11.0 CONFIG REQUIRED PATHS "${ROCM_PATH}")
+  endif()
+endif()

default.xml

@@ -1,73 +1,71 @@
 <?xml version="1.0" encoding="UTF-8"?>
 <manifest>
-    <remote name="roc-github"
-fetch="http://github.com/RadeonOpenCompute/" />
-    <remote name="rocm-devtools"
-fetch="https://github.com/ROCm-Developer-Tools/" />
-    <remote name="rocm-swplat"
-fetch="https://github.com/ROCmSoftwarePlatform/" />
-    <remote name="gpuopen-libs"
-fetch="https://github.com/GPUOpen-ProfessionalCompute-Libraries/" />
-    <remote name="gpuopen-tools"
-fetch="https://github.com/GPUOpen-Tools/" />
-    <remote name="KhronosGroup"
-fetch="https://github.com/KhronosGroup/" />
-    <default revision="refs/tags/rocm-4.0.0"
-remote="roc-github"
-sync-c="true"
-sync-j="4" />
-<!--list of projects for ROCM-->
-    <project name="ROCK-Kernel-Driver" revision="refs/tags/rocm-4.0.1" />
-    <project name="ROCT-Thunk-Interface" />
+    <remote name="rocm-org" fetch="https://github.com/ROCm/" />
+    <remote name="KhronosGroup" fetch="https://github.com/KhronosGroup/" />
+    <default revision="refs/tags/rocm-6.1.0"
+     remote="rocm-org"
+     sync-c="true"
+     sync-j="4" />
+<!--list of projects for ROCm-->
+    <project path="ROCm-OpenCL-Runtime/api/opencl/khronos/icd" name="OpenCL-ICD-Loader" remote="KhronosGroup" />
+    <project name="ROCK-Kernel-Driver" />
     <project name="ROCR-Runtime" />
-    <project name="ROC-smi" />
-    <project name="rocm_smi_lib" />
-    <project name="rocm-cmake" />
-    <project name="rocminfo" />
-    <project name="rocprofiler" remote="rocm-devtools" />
-    <project name="roctracer" remote="rocm-devtools" />
-    <project name="ROCm-OpenCL-Runtime" />
-    <project path="ROCm-OpenCL-Runtime/api/opencl/khronos/icd" name="OpenCL-ICD-Loader" remote="KhronosGroup" revision="6c03f8b58fafd9dd693eaac826749a5cfad515f8" />
+    <project name="ROCT-Thunk-Interface" />
+    <project name="amdsmi" />
     <project name="clang-ocl" />
-<!--HIP Projects-->
-    <project name="HIP" remote="rocm-devtools" />
-    <project name="HIP-Examples" remote="rocm-devtools" />
-    <project name="ROCclr" remote="rocm-devtools" />
-    <project name="HIPIFY" remote="rocm-devtools" />
-<!-- The following projects are all associated with the AMDGPU LLVM compiler -->
-    <project name="llvm-project" revision="refs/tags/rocm-4.0.1" />
-    <project name="ROCm-Device-Libs" />
-    <project name="atmi" />
-    <project name="ROCm-CompilerSupport" />
-    <project name="rocr_debug_agent" remote="rocm-devtools" />
+    <project name="rdc" />
     <project name="rocm_bandwidth_test" />
-    <project name="half" remote="rocm-swplat" revision="37742ce15b76b44e4b271c1e66d13d2fa7bd003e" />
-    <project name="RCP" remote="gpuopen-tools" revision="3a49405a1500067c49d181844ec90aea606055bb" />
+    <project name="rocm_smi_lib" />
+    <project name="rocm-core" />
+    <project name="rocminfo" />
+    <project name="rocprofiler" />
+    <project name="rocprofiler-register" />
+    <project name="roctracer" />
+<!--HIP Projects-->
+    <project name="HIP" />
+    <project name="HIP-Examples" />
+    <project name="HIPIFY" />
+    <project name="clr" />
+    <project name="hipother" />
+<!-- The following projects are all associated with the AMDGPU LLVM compiler -->
+    <project name="half" />
+    <project name="llvm-project" />
 <!-- gdb projects -->
-    <project name="ROCgdb" remote="rocm-devtools" />
-    <project name="ROCdbgapi" remote="rocm-devtools" />
+    <project name="ROCdbgapi" />
+    <project name="ROCgdb" />
+    <project name="rocr_debug_agent" />
 <!-- ROCm Libraries -->
-    <project name="rdc" remote="roc-github" />
-    <project name="rocBLAS" remote="rocm-swplat" />
-    <project name="hipBLAS" remote="rocm-swplat" />
-    <project name="rocFFT" remote="rocm-swplat" />
-    <project name="rocRAND" remote="rocm-swplat" />
-    <project name="rocSPARSE" remote="rocm-swplat" />
-    <project name="rocSOLVER" remote="rocm-swplat" />
-    <project name="hipSPARSE" remote="rocm-swplat" />
-    <project name="rocALUTION" remote="rocm-swplat" />
-    <project name="MIOpenGEMM" remote="rocm-swplat" />
-    <project name="MIOpen" remote="rocm-swplat"  revision="refs/tags/rocm-4.0.1" />
-    <project name="rccl" remote="rocm-swplat" />
-    <project name="MIVisionX" remote="gpuopen-libs" />
-    <project name="rocThrust" remote="rocm-swplat" />
-    <project name="hipCUB" remote="rocm-swplat" />
-    <project name="rocPRIM" remote="rocm-swplat" />
-    <project name="hipfort" remote="rocm-swplat" />
-    <project name="AMDMIGraphX" remote="rocm-swplat" />
-    <project name="ROCmValidationSuite" remote="rocm-devtools" revision="refs/tags/rocm-4.0.1" />
+    <project groups="mathlibs" name="AMDMIGraphX" />
+    <project groups="mathlibs" name="MIOpen" />
+    <project groups="mathlibs" name="MIVisionX" />
+    <project groups="mathlibs" name="ROCmValidationSuite" />
+    <project groups="mathlibs" name="Tensile" />
+    <project groups="mathlibs" name="composable_kernel" />
+    <project groups="mathlibs" name="hipBLAS" />
+    <project groups="mathlibs" name="hipBLASLt" />
+    <project groups="mathlibs" name="hipCUB" />
+    <project groups="mathlibs" name="hipFFT" />
+    <project groups="mathlibs" name="hipRAND" />
+    <project groups="mathlibs" name="hipSOLVER" />
+    <project groups="mathlibs" name="hipSPARSE" />
+    <project groups="mathlibs" name="hipSPARSELt" />
+    <project groups="mathlibs" name="hipTensor" />
+    <project groups="mathlibs" name="hipfort" />
+    <project groups="mathlibs" name="rccl" />
+    <project groups="mathlibs" name="rocALUTION" />
+    <project groups="mathlibs" name="rocBLAS" />
+    <project groups="mathlibs" name="rocDecode" />
+    <project groups="mathlibs" name="rocFFT" />
+    <project groups="mathlibs" name="rocPRIM" />
+    <project groups="mathlibs" name="rocRAND" />
+    <project groups="mathlibs" name="rocSOLVER" />
+    <project groups="mathlibs" name="rocSPARSE" />
+    <project groups="mathlibs" name="rocThrust" />
+    <project groups="mathlibs" name="rocWMMA" />
+    <project groups="mathlibs" name="rocm-cmake" />
+    <project groups="mathlibs" name="rpp" />
 <!-- Projects for OpenMP-Extras -->
-    <project name="aomp" path="openmp-extras/aomp" remote="rocm-devtools" />
-    <project name="aomp-extras" path="openmp-extras/aomp-extras" remote="rocm-devtools" />
-    <project name="flang" path="openmp-extras/flang" remote="rocm-devtools" />
+    <project name="aomp" path="openmp-extras/aomp" />
+    <project name="aomp-extras" path="openmp-extras/aomp-extras" />
+    <project name="flang" path="openmp-extras/flang" />
 </manifest>

docs/CMakeLists.txt

@@ -0,0 +1,33 @@
+# MIT License
+#
+# Copyright (c) 2023 Advanced Micro Devices, Inc. All rights reserved.
+#
+# 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.
+
+include(ROCMSphinxDoc)
+
+rocm_add_sphinx_doc(
+    "${CMAKE_CURRENT_SOURCE_DIR}"
+  OUTPUT_DIR html
+  BUILDER html
+)
+  
+install(
+  DIRECTORY "${CMAKE_CURRENT_BINARY_DIR}/html"
+  DESTINATION "${CMAKE_INSTALL_DOCDIR}")

docs/about/compatibility/openmp.md

@@ -0,0 +1,483 @@
+<head>
+  <meta charset="UTF-8">
+  <meta name="description" content="OpenMP support in ROCm">
+  <meta name="keywords" content="OpenMP, LLVM, OpenMP toolchain">
+</head>
+
+# OpenMP support in ROCm
+
+## Introduction
+
+The ROCm™ installation includes an LLVM-based implementation that fully supports
+the OpenMP 4.5 standard and a subset of OpenMP 5.0, 5.1, and 5.2 standards.
+Fortran, C/C++ compilers, and corresponding runtime libraries are included.
+Along with host APIs, the OpenMP compilers support offloading code and data onto
+GPU devices. This document briefly describes the installation location of the
+OpenMP toolchain, example usage of device offloading, and usage of `rocprof`
+with OpenMP applications. The GPUs supported are the same as those supported by
+this ROCm release. See the list of supported GPUs for {doc}`Linux<rocm-install-on-linux:reference/system-requirements>` and
+{doc}`Windows<rocm-install-on-windows:reference/system-requirements>`.
+
+The ROCm OpenMP compiler is implemented using LLVM compiler technology.
+The following image illustrates the internal steps taken to translate a user’s application into an executable that can offload computation to the AMDGPU. The compilation is a two-pass process. Pass 1 compiles the application to generate the CPU code and Pass 2 links the CPU code to the AMDGPU device code.
+
+![OpenMP toolchain](../../data/reference/openmp/openmp-toolchain.svg "OpenMP toolchain")
+
+### Installation
+
+The OpenMP toolchain is automatically installed as part of the standard ROCm
+installation and is available under `/opt/rocm-{version}/llvm`. The
+sub-directories are:
+
+* bin: Compilers (`flang` and `clang`) and other binaries.
+* examples: The usage section below shows how to compile and run these programs.
+* include: Header files.
+* lib: Libraries including those required for target offload.
+* lib-debug: Debug versions of the above libraries.
+
+## OpenMP: usage
+
+The example programs can be compiled and run by pointing the environment
+variable `ROCM_PATH` to the ROCm install directory.
+
+**Example:**
+
+```bash
+export ROCM_PATH=/opt/rocm-{version}
+cd $ROCM_PATH/share/openmp-extras/examples/openmp/veccopy
+sudo make run
+```
+
+:::{note}
+`sudo` is required since we are building inside the `/opt` directory.
+Alternatively, copy the files to your home directory first.
+:::
+
+The above invocation of Make compiles and runs the program. Note the options
+that are required for target offload from an OpenMP program:
+
+```bash
+-fopenmp --offload-arch=<gpu-arch>
+```
+
+:::{note}
+The compiler also accepts the alternative offloading notation:
+
+```bash
+-fopenmp -fopenmp-targets=amdgcn-amd-amdhsa -Xopenmp-target=amdgcn-amd-amdhsa -march=<gpu-arch>
+```
+
+:::
+
+Obtain the value of `gpu-arch` by running the following command:
+
+```bash
+% /opt/rocm-{version}/bin/rocminfo | grep gfx
+```
+
+[//]: # (dated link below, needs updating)
+
+See the complete list of compiler command-line references
+[here](https://github.com/ROCm/llvm-project/blob/amd-stg-open/clang/docs/CommandGuide/clang.rst).
+
+### Using `rocprof` with OpenMP
+
+The following steps describe a typical workflow for using `rocprof` with OpenMP
+code compiled with AOMP:
+
+1. Run `rocprof` with the program command line:
+
+    ```bash
+    % rocprof <application> <args>
+    ```
+
+    This produces a `results.csv` file in the user’s current directory that
+    shows basic stats such as kernel names, grid size, number of registers used,
+    etc. The user can choose to specify the preferred output file name using the
+    o option.
+
+2. Add options for a detailed result:
+
+   ```bash
+   --stats: % rocprof --stats <application> <args>
+   ```
+
+   The stats option produces timestamps for the kernels. Look into the output
+   CSV file for the field, `DurationNs`, which is useful in getting an
+   understanding of the critical kernels in the code.
+
+   Apart from `--stats`, the option `--timestamp` on produces a timestamp for
+   the kernels.
+
+3. After learning about the required kernels, the user can take a detailed look
+   at each one of them. `rocprof` has support for hardware counters: a set of
+   basic and a set of derived ones. See the complete list of counters using
+   options --list-basic and --list-derived. `rocprof` accepts either a text or
+   an XML file as an input.
+
+For more details on `rocprof`, refer to the {doc}`ROCProfilerV1 User Manual <rocprofiler:rocprofv1>`.
+
+### Using tracing options
+
+**Prerequisite:** When using the `--sys-trace` option, compile the OpenMP
+program with:
+
+```bash
+    -Wl,-rpath,/opt/rocm-{version}/lib -lamdhip64
+```
+
+The following tracing options are widely used to generate useful information:
+
+* **`--hsa-trace`**: This option is used to get a JSON output file with the HSA
+  API execution traces and a flat profile in a CSV file.
+
+* **`--sys-trace`**: This allows programmers to trace both HIP and HSA calls.
+  Since this option results in loading ``libamdhip64.so``, follow the
+  prerequisite as mentioned above.
+
+A CSV and a JSON file are produced by the above trace options. The CSV file
+presents the data in a tabular format, and the JSON file can be visualized using
+Google Chrome at chrome://tracing/ or [Perfetto](https://perfetto.dev/).
+Navigate to Chrome or Perfetto and load the JSON file to see the timeline of the
+HSA calls.
+
+For more details on tracing, refer to the {doc}`ROCProfilerV1 User Manual <rocprofiler:rocprofv1>`.
+
+### Environment variables
+
+:::{table}
+:widths: auto
+| Environment Variable        | Purpose                  |
+| --------------------------- | ---------------------------- |
+| `OMP_NUM_TEAMS`             | To set the number of teams for kernel launch, which is otherwise chosen by the implementation by default. You can set this number (subject to implementation limits) for performance tuning. |
+| `LIBOMPTARGET_KERNEL_TRACE` | To print useful statistics for device operations. Setting it to 1 and running the program emits the name of every kernel launched, the number of teams and threads used, and the corresponding register usage. Setting it to 2 additionally emits timing information for kernel launches and data transfer operations between the host and the device. |
+| `LIBOMPTARGET_INFO`         | To print informational messages from the device runtime as the program executes. Setting it to a value of 1 or higher, prints fine-grain information and setting it to -1 prints complete information. |
+| `LIBOMPTARGET_DEBUG`        | To get detailed debugging information about data transfer operations and kernel launch when using a debug version of the device library. Set this environment variable to 1 to get the detailed information from the library. |
+| `GPU_MAX_HW_QUEUES`         | To set the number of HSA queues in the OpenMP runtime. The HSA queues are created on demand up to the maximum value as supplied here. The queue creation starts with a single initialized queue to avoid unnecessary allocation of resources. The provided value is capped if it exceeds the recommended, device-specific value. |
+| `LIBOMPTARGET_AMDGPU_MAX_ASYNC_COPY_BYTES` | To set the threshold size up to which data transfers are initiated asynchronously. The default threshold size is 1*1024*1024 bytes (1MB). |
+| `OMPX_FORCE_SYNC_REGIONS` | To force the runtime to execute all operations synchronously, i.e., wait for an operation to complete immediately. This affects data transfers and kernel execution. While it is mainly designed for debugging, it may have a minor positive effect on performance in certain situations. |
+:::
+
+## OpenMP: features
+
+The OpenMP programming model is greatly enhanced with the following new features
+implemented in the past releases.
+
+(openmp_usm)=
+
+### Asynchronous behavior in OpenMP target regions
+
+* Controlling Asynchronous Behavior
+
+The OpenMP offloading runtime executes in an asynchronous fashion by default, allowing multiple data transfers to start concurrently. However, if the data to be transferred becomes larger than the default threshold of 1MB, the runtime falls back to a synchronous data transfer. The buffers that have been locked already are always executed asynchronously.
+You can overrule this default behavior by setting `LIBOMPTARGET_AMDGPU_MAX_ASYNC_COPY_BYTES` and `OMPX_FORCE_SYNC_REGIONS`. See the [Environment Variables](#environment-variables) table for details.
+
+* Multithreaded Offloading on the Same Device
+
+The `libomptarget` plugin for GPU offloading allows creation of separate configurable HSA queues per chiplet, which enables two or more threads to concurrently offload to the same device.
+
+* Parallel Memory Copy Invocations
+
+Implicit asynchronous execution of single target region enables parallel memory copy invocations.
+
+### Unified shared memory
+
+Unified Shared Memory (USM) provides a pointer-based approach to memory
+management. To implement USM, fulfill the following system requirements along
+with Xnack capability.
+
+#### Prerequisites
+
+* Linux Kernel versions above 5.14
+* Latest KFD driver packaged in ROCm stack
+* Xnack, as USM support can only be tested with applications compiled with Xnack
+  capability
+
+#### Xnack capability
+
+When enabled, Xnack capability allows GPU threads to access CPU (system) memory,
+allocated with OS-allocators, such as `malloc`, `new`, and `mmap`. Xnack must be
+enabled both at compile- and run-time. To enable Xnack support at compile-time,
+use:
+
+```bash
+--offload-arch=gfx908:xnack+
+```
+
+Or use another functionally equivalent option Xnack-any:
+
+```bash
+--offload-arch=gfx908
+```
+
+To enable Xnack functionality at runtime on a per-application basis,
+use environment variable:
+
+```bash
+HSA_XNACK=1
+```
+
+When Xnack support is not needed:
+
+* Build the applications to maximize resource utilization using:
+
+```bash
+--offload-arch=gfx908:xnack-
+```
+
+* At runtime, set the `HSA_XNACK` environment variable to 0.
+
+#### Unified shared memory pragma
+
+This OpenMP pragma is available on MI200 through `xnack+` support.
+
+```bash
+omp requires unified_shared_memory
+```
+
+As stated in the OpenMP specifications, this pragma makes the map clause on
+target constructs optional. By default, on MI200, all memory allocated on the
+host is fine grain. Using the map clause on a target clause is allowed, which
+transforms the access semantics of the associated memory to coarse grain.
+
+```bash
+A simple program demonstrating the use of this feature is:
+$ cat parallel_for.cpp
+#include <stdlib.h>
+#include <stdio.h>
+
+#define N 64
+#pragma omp requires unified_shared_memory
+int main() {
+  int n = N;
+  int *a = new int[n];
+  int *b = new int[n];
+
+  for(int i = 0; i < n; i++)
+    b[i] = i;
+
+  #pragma omp target parallel for map(to:b[:n])
+  for(int i = 0; i < n; i++)
+    a[i] = b[i];
+
+  for(int i = 0; i < n; i++)
+    if(a[i] != i)
+      printf("error at %d: expected %d, got %d\n", i, i+1, a[i]);
+
+  return 0;
+}
+$ clang++ -O2 -target x86_64-pc-linux-gnu -fopenmp --offload-arch=gfx90a:xnack+ parallel_for.cpp
+$ HSA_XNACK=1 ./a.out
+```
+
+In the above code example, pointer “a” is not mapped in the target region, while
+pointer “b” is. Both are valid pointers on the GPU device and passed by-value to
+the kernel implementing the target region. This means the pointer values on the
+host and the device are the same.
+
+The difference between the memory pages pointed to by these two variables is
+that the pages pointed by “a” are in fine-grain memory, while the pages pointed
+to by “b” are in coarse-grain memory during and after the execution of the
+target region. This is accomplished in the OpenMP runtime library with calls to
+the ROCr runtime to set the pages pointed by “b” as coarse grain.
+
+### OMPT target support
+
+The OpenMP runtime in ROCm implements a subset of the OMPT device APIs, as
+described in the OpenMP specification document. These APIs allow first-party
+tools to examine the profile and kernel traces that execute on a device. A tool
+can register callbacks for data transfer and kernel dispatch entry points or use
+APIs to start and stop tracing for device-related activities such as data
+transfer and kernel dispatch timings and associated metadata. If device tracing
+is enabled, trace records for device activities are collected during program
+execution and returned to the tool using the APIs described in the
+specification.
+
+The following example demonstrates how a tool uses the supported OMPT target
+APIs. The `README` in `/opt/rocm/llvm/examples/tools/ompt` outlines the steps to
+be followed, and the provided example can be run as shown below:
+
+```bash
+cd $ROCM_PATH/share/openmp-extras/examples/tools/ompt/veccopy-ompt-target-tracing
+sudo make run
+```
+
+The file `veccopy-ompt-target-tracing.c` simulates how a tool initiates device
+activity tracing. The file `callbacks.h` shows the callbacks registered and
+implemented by the tool.
+
+### Floating point atomic operations
+
+The MI200-series GPUs support the generation of hardware floating-point atomics
+using the OpenMP atomic pragma. The support includes single- and
+double-precision floating-point atomic operations. The programmer must ensure
+that the memory subjected to the atomic operation is in coarse-grain memory by
+mapping it explicitly with the help of map clauses when not implicitly mapped by
+the compiler as per the [OpenMP
+specifications](https://www.openmp.org/specifications/). This makes these
+hardware floating-point atomic instructions “fast,” as they are faster than
+using a default compare-and-swap loop scheme, but at the same time “unsafe,” as
+they are not supported on fine-grain memory. The operation in
+`unified_shared_memory` mode also requires programmers to map the memory
+explicitly when not implicitly mapped by the compiler.
+
+To request fast floating-point atomic instructions at the file level, use
+compiler flag `-munsafe-fp-atomics` or a hint clause on a specific pragma:
+
+```bash
+double a = 0.0;
+#pragma omp atomic hint(AMD_fast_fp_atomics)
+a = a + 1.0;
+```
+
+:::{note}
+`AMD_unsafe_fp_atomics` is an alias for `AMD_fast_fp_atomics`, and
+`AMD_safe_fp_atomics` is implemented with a compare-and-swap loop.
+:::
+
+To disable the generation of fast floating-point atomic instructions at the file
+level, build using the option `-msafe-fp-atomics` or use a hint clause on a
+specific pragma:
+
+```bash
+double a = 0.0;
+#pragma omp atomic hint(AMD_safe_fp_atomics)
+a = a + 1.0;
+```
+
+The hint clause value always has a precedence over the compiler flag, which
+allows programmers to create atomic constructs with a different behavior than
+the rest of the file.
+
+See the example below, where the user builds the program using
+`-msafe-fp-atomics` to select a file-wide “safe atomic” compilation. However,
+the fast atomics hint clause over variable “a” takes precedence and operates on
+“a” using a fast/unsafe floating-point atomic, while the variable “b” in the
+absence of a hint clause is operated upon using safe floating-point atomics as
+per the compiler flag.
+
+```bash
+double a = 0.0;.
+#pragma omp atomic hint(AMD_fast_fp_atomics)
+a = a + 1.0;
+
+double b = 0.0;
+#pragma omp atomic
+b = b + 1.0;
+```
+
+### AddressSanitizer tool
+
+AddressSanitizer (ASan) is a memory error detector tool utilized by applications to
+detect various errors ranging from spatial issues such as out-of-bound access to
+temporal issues such as use-after-free. The AOMP compiler supports ASan for AMD
+GPUs with applications written in both HIP and OpenMP.
+
+**Features supported on host platform (Target x86_64):**
+
+* Use-after-free
+* Buffer overflows
+* Heap buffer overflow
+* Stack buffer overflow
+* Global buffer overflow
+* Use-after-return
+* Use-after-scope
+* Initialization order bugs
+
+**Features supported on AMDGPU platform (`amdgcn-amd-amdhsa`):**
+
+* Heap buffer overflow
+* Global buffer overflow
+
+**Software (kernel/OS) requirements:** Unified Shared Memory support with Xnack
+capability. See the section on [Unified Shared Memory](#unified-shared-memory)
+for prerequisites and details on Xnack.
+
+**Example:**
+
+* Heap buffer overflow
+
+```bash
+void  main() {
+.......  // Some program statements
+.......  // Some program statements
+#pragma omp target map(to : A[0:N], B[0:N]) map(from: C[0:N])
+{
+#pragma omp parallel for
+    for(int i =0 ; i < N; i++){
+    C[i+10] = A[i] + B[i];
+  }   // end of for loop
+}
+.......   // Some program statements
+}// end of main
+```
+
+See the complete sample code for heap buffer overflow
+[here](https://github.com/ROCm/aomp/blob/aomp-dev/examples/tools/asan/heap_buffer_overflow/openmp/vecadd-HBO.cpp).
+
+* Global buffer overflow
+
+```bash
+#pragma omp declare target
+   int A[N],B[N],C[N];
+#pragma omp end declare target
+void main(){
+......  // some program statements
+......  // some program statements
+#pragma omp target data map(to:A[0:N],B[0:N]) map(from: C[0:N])
+{
+#pragma omp target update to(A,B)
+#pragma omp target parallel for
+for(int i=0; i<N; i++){
+    C[i]=A[i*100]+B[i+22];
+} // end of for loop
+#pragma omp target update from(C)
+}
+........  // some program statements
+} // end of main
+```
+
+See the complete sample code for global buffer overflow
+[here](https://github.com/ROCm/aomp/blob/aomp-dev/examples/tools/asan/global_buffer_overflow/openmp/vecadd-GBO.cpp).
+
+### Clang compiler option for kernel optimization
+
+You can use the clang compiler option `-fopenmp-target-fast` for kernel optimization if certain constraints implied by its component options are satisfied. `-fopenmp-target-fast` enables the following options:
+
+* `-fopenmp-target-ignore-env-vars`: It enables code generation of specialized kernels including no-loop and Cross-team reductions.
+
+* `-fopenmp-assume-no-thread-state`: It enables the compiler to assume that no thread in a parallel region modifies an Internal Control Variable (`ICV`), thus potentially reducing the device runtime code execution.
+
+* `-fopenmp-assume-no-nested-parallelism`: It enables the compiler to assume that no thread in a parallel region encounters a parallel region, thus potentially reducing the device runtime code execution.
+
+* `-O3` if no `-O*` is specified by the user.
+
+### Specialized kernels
+
+Clang will attempt to generate specialized kernels based on compiler options and OpenMP constructs. The following specialized kernels are supported:
+
+* No-loop
+* Big-jump-loop
+* Cross-team reductions
+
+To enable the generation of specialized kernels, follow these guidelines:
+
+* Do not specify teams, threads, and schedule-related environment variables. The `num_teams` clause in an OpenMP target construct acts as an override and prevents the generation of the no-loop kernel. If the specification of `num_teams` clause is a user requirement then clang tries to generate the big-jump-loop kernel instead of the no-loop kernel.
+
+* Assert the absence of the teams, threads, and schedule-related environment variables by adding the command-line option `-fopenmp-target-ignore-env-vars`.
+
+* To automatically enable the specialized kernel generation, use `-Ofast` or `-fopenmp-target-fast` for compilation.
+
+* To disable specialized kernel generation, use `-fno-openmp-target-ignore-env-vars`.
+
+#### No-loop kernel generation
+
+The no-loop kernel generation feature optimizes the compiler performance by generating a specialized kernel for certain OpenMP target constructs such as target teams distribute parallel for. The specialized kernel generation feature assumes every thread executes a single iteration of the user loop, which leads the runtime to launch a total number of GPU threads equal to or greater than the iteration space size of the target region loop. This allows the compiler to generate code for the loop body without an enclosing loop, resulting in reduced control-flow complexity and potentially better performance.
+
+#### Big-jump-loop kernel generation
+
+A no-loop kernel is not generated if the OpenMP teams construct uses a `num_teams` clause. Instead, the compiler attempts to generate a different specialized kernel called the big-jump-loop kernel. The compiler launches the kernel with a grid size determined by the number of teams specified by the OpenMP `num_teams` clause and the `blocksize` chosen either by the compiler or specified by the corresponding OpenMP clause.
+
+#### Cross-team optimized reduction kernel generation
+
+If the OpenMP construct has a reduction clause, the compiler attempts to generate optimized code by utilizing efficient cross-team communication. New APIs for cross-team reduction are implemented in the device runtime and are automatically generated by clang.

docs/about/license.md

@@ -0,0 +1,146 @@
+<head>
+  <meta charset="UTF-8">
+  <meta name="description" content="ROCm licensing terms">
+  <meta name="keywords" content="license, licensing terms">
+</head>
+
+# ROCm license
+
+```{include} ../../LICENSE
+```
+
+:::{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.
+
+<!-- spellcheck-disable -->
+| Component | License |
+|:---------------------|:-------------------------|
+| [HIP](https://github.com/ROCm/HIP/) | [MIT](https://github.com/ROCm/HIP/blob/develop/LICENSE.txt) |
+| [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) |
+| [HIPIFY](https://github.com/ROCm/HIPIFY/) | [MIT](https://github.com/ROCm/HIPIFY/blob/amd-staging/LICENSE.txt) |
+| [AMDMIGraphX](https://github.com/ROCm/AMDMIGraphX/) | [MIT](https://github.com/ROCm/AMDMIGraphX/blob/develop/LICENSE) |
+| [MIOpen](https://github.com/ROCm/MIOpen/) | [MIT](https://github.com/ROCm/MIOpen/blob/develop/LICENSE.txt) |
+| [MIVisionX](https://github.com/ROCm/MIVisionX/) | [MIT](https://github.com/ROCm/MIVisionX/blob/develop/LICENSE.txt) |
+| [AMD Common Language Runtime (CLR)](https://github.com/ROCm/clr) | [MIT](https://github.com/ROCm/clr/blob/develop/LICENCE) |
+| [ROCm-Core](https://github.com/ROCm/rocm-core) | [MIT](https://github.com/ROCm/rocm-core/blob/master/copyright) |
+| [hipamd](https://github.com/ROCm/clr/tree/develop/hipamd) | [MIT](https://github.com/ROCm/clr/blob/develop/hipamd/LICENSE.txt) |
+| [ROCm-OpenCL-Runtime](https://github.com/ROCm/clr/tree/develop/opencl) | [MIT](https://github.com/ROCm/clr/blob/develop/opencl/LICENSE.txt) |
+| [Tensile](https://github.com/ROCm/Tensile/) | [MIT](https://github.com/ROCm/Tensile/blob/develop/LICENSE.md) |
+| [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) |
+| [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) |
+| [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) |
+| [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) |
+| [clang-ocl](https://github.com/ROCm/clang-ocl/) | [MIT](https://github.com/ROCm/clang-ocl/blob/master/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) |
+| [ROCT-Thunk-Interface](https://github.com/ROCm/ROCT-Thunk-Interface/) | [MIT](https://github.com/ROCm/ROCT-Thunk-Interface/blob/master/LICENSE.md) |
+| [ROCR-Runtime](https://github.com/ROCm/ROCR-Runtime/) | [The University of Illinois/NCSA](https://github.com/ROCm/ROCR-Runtime/blob/master/LICENSE.txt) |
+| [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) |
+| [Composable Kernel](https://github.com/ROCm/composable_kernel) | [MIT](https://github.com/ROCm/composable_kernel/blob/develop/LICENSE) |
+| [half](https://github.com/ROCm/half/) | [MIT](https://github.com/ROCm/half/blob/rocm/LICENSE.txt) |
+| [hipBLAS](https://github.com/ROCm/hipBLAS/) | [MIT](https://github.com/ROCm/hipBLAS/blob/develop/LICENSE.md) |
+| [hipBLASLt](https://github.com/ROCm/hipBLASLt/) | [MIT](https://github.com/ROCm/hipBLASLt/blob/develop/LICENSE.md) |
+| [hipCUB](https://github.com/ROCm/hipCUB/) | [Custom](https://github.com/ROCm/hipCUB/blob/develop/LICENSE.txt) |
+| [hipFFT](https://github.com/ROCm/hipFFT/) | [MIT](https://github.com/ROCm/hipFFT/blob/develop/LICENSE.md) |
+| [hipFORT](https://github.com/ROCm/hipfort/) | [MIT](https://github.com/ROCm/hipfort/blob/develop/LICENSE) |
+| [hipRAND](https://github.com/ROCm/hipRAND/) | [MIT](https://github.com/ROCm/hipRAND/blob/develop/LICENSE.txt) |
+| [hipSOLVER](https://github.com/ROCm/hipSOLVER/) | [MIT](https://github.com/ROCm/hipSOLVER/blob/develop/LICENSE.md) |
+| [hipSPARSE](https://github.com/ROCm/hipSPARSE/) | [MIT](https://github.com/ROCm/hipSPARSE/blob/develop/LICENSE.md) |
+| [hipSPARSELt](https://github.com/ROCm/hipSPARSELt/) | [MIT](https://github.com/ROCm/hipSPARSELt/blob/develop/LICENSE.md) |
+| [hipTensor](https://github.com/ROCm/hipTensor) | [MIT](https://github.com/ROCm/hipTensor/blob/develop/LICENSE) |
+| [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/rocBLAS/) | [MIT](https://github.com/ROCm/rocBLAS/blob/develop/LICENSE.md) |
+| [rocDecode](https://github.com/ROCm/rocDecode) | [MIT](https://github.com/ROCm/rocDecode/blob/develop/LICENSE) |
+| [rocFFT](https://github.com/ROCm/rocFFT/) | [MIT](https://github.com/ROCm/rocFFT/blob/develop/LICENSE.md) |
+| [rocPRIM](https://github.com/ROCm/rocPRIM/) | [MIT](https://github.com/ROCm/rocPRIM/blob/develop/LICENSE.txt) |
+| [ROCm Performance Primitives (RPP)](https://github.com/ROCm/rpp) | [MIT](https://github.com/ROCm/rpp/blob/develop/LICENSE) |
+| [rocRAND](https://github.com/ROCm/rocRAND/) | [MIT](https://github.com/ROCm/rocRAND/blob/develop/LICENSE.txt) |
+| [rocSOLVER](https://github.com/ROCm/rocSOLVER/) | [BSD-2-Clause](https://github.com/ROCm/rocSOLVER/blob/develop/LICENSE.md) |
+| [rocSPARSE](https://github.com/ROCm/rocSPARSE/) | [MIT](https://github.com/ROCm/rocSPARSE/blob/develop/LICENSE.md) |
+| [rocThrust](https://github.com/ROCm/rocThrust/) | [Apache 2.0](https://github.com/ROCm/rocThrust/blob/develop/LICENSE) |
+| [rocWMMA](https://github.com/ROCm/rocWMMA/) | [MIT](https://github.com/ROCm/rocWMMA/blob/develop/LICENSE.md) |
+| [ROCm Communication Collectives Library (RCCL)](https://github.com/ROCm/rccl/) | [Custom](https://github.com/ROCm/rccl/blob/develop/LICENSE.txt) |
+| [ROCm Data Center (RDC)](https://github.com/ROCm/rdc/) | [MIT](https://github.com/ROCm/rdc/blob/develop/LICENSE) |
+| [ROCm CMake](https://github.com/ROCm/rocm-cmake/) | [MIT](https://github.com/ROCm/rocm-cmake/blob/develop/LICENSE) |
+| [ROCdbgapi](https://github.com/ROCm/ROCdbgapi/) | [MIT](https://github.com/ROCm/ROCdbgapi/blob/amd-staging/LICENSE.txt) |
+| [ROCgdb](https://github.com/ROCm/ROCgdb/) | [GNU General Public License v2.0](https://github.com/ROCm/ROCgdb/blob/amd-master/COPYING) |
+| [ROCm SMI Lib](https://github.com/ROCm/rocm_smi_lib/) | [MIT](https://github.com/ROCm/rocm_smi_lib/blob/develop/License.txt) |
+| [AMD SMI](https://github.com/ROCm/amdsmi) | [MIT](https://github.com/ROCm/amdsmi/blob/develop/LICENSE) |
+| [rocminfo](https://github.com/ROCm/rocminfo/) | [The University of Illinois/NCSA](https://github.com/ROCm/rocminfo/blob/amd-staging/License.txt) |
+| [ROCProfiler](https://github.com/ROCm/rocprofiler/) | [MIT](https://github.com/ROCm/rocprofiler/blob/amd-master/LICENSE) |
+| [ROCTracer](https://github.com/ROCm/roctracer/) | [MIT](https://github.com/ROCm/roctracer/blob/amd-master/LICENSE) |
+| [ROCm Bandwidth Test](https://github.com/ROCm/rocm_bandwidth_test/) | [The University of Illinois/NCSA](https://github.com/ROCm/rocm_bandwidth_test/blob/master/LICENSE.txt) |
+| [TransferBench](https://github.com/ROCm/TransferBench) | [MIT](https://github.com/ROCm/TransferBench/blob/develop/LICENSE.md) |
+| [ROCmValidationSuite](https://github.com/ROCm/ROCmValidationSuite/) | [MIT](https://github.com/ROCm/ROCmValidationSuite/blob/master/LICENSE) |
+| hsa-amd-aqlprofile | [AMD Software EULA](https://www.amd.com/en/legal/eula/amd-software-eula.html)
+
+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). Currently, only
+one component of ROCm, `rocm-llvm-alt` is governed by a proprietary license.
+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.
+```
+
+©2023 - 2024 Advanced Micro Devices, Inc. All rights reserved.
+
+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}
+AQL Profiler and AOCC CPU optimization are both provided in binary form, each
+subject to the license agreement enclosed in the directory for the binary available
+in `/opt/rocm/share/doc/hsa-amd-aqlprofile/EULA`. By using, installing,
+copying or distributing AQL Profiler and/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 the AQL Profiler and/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.

docs/compatibility/compatibility-matrix.rst

@@ -0,0 +1,127 @@
+.. meta::
+    :description: ROCm compatibility matrix
+    :keywords: AMD, GPU, architecture, hardware, compatibility, requirements
+
+**************************************************************************************
+Compatibility matrix
+**************************************************************************************
+
+Use this matrix to view the ROCm compatibility across successive major and minor releases.
+
+
+.. container:: format-big-table
+
+  .. csv-table:: 
+      :header: "ROCm Version", "6.1.0", "6.0.0"
+      :stub-columns: 1
+
+      :doc:`Operating Systems <rocm-install-on-linux:reference/system-requirements>`, "Ubuntu 22.04.4, 22.04.3","Ubuntu 22.04.4, 22.04.3"
+      ,"Ubuntu 20.04.6, 20.04.5","Ubuntu 20.04.6, 20.04.5"
+      ,"RHEL 9.4 [#red-hat94]_, 9.3, 9.2","RHEL 9.3, 9.2"
+      ,"RHEL 8.9, 8.8","RHEL 8.9, 8.8"
+      ,"SLES 15 SP5, SP4","SLES 15 SP5, SP4"
+      ,CentOS 7.9,CentOS 7.9
+      ,,
+      :doc:`GFX Architecture <rocm-install-on-linux:reference/system-requirements>`,CDNA3,CDNA3
+      ,CDNA2,CDNA2
+      ,CDNA,CDNA
+      ,RDNA3,RDNA3
+      ,RDNA2,RDNA2
+      ,,
+      :doc:`GFX Card <rocm-install-on-linux:reference/system-requirements>`,gfx1100,gfx1100
+      ,gfx1030,gfx1030
+      ,gfx942 [#]_, gfx942 [#]_
+      ,gfx90a,gfx90a
+      ,gfx908,gfx908
+      ,,
+      ECOSYSTEM SUPPORT:,,
+      :doc:`PyTorch <rocm-install-on-linux:how-to/3rd-party/pytorch-install>`,"2.1, 2.0, 1.13","2.1, 2.0, 1.13"
+      :doc:`TensorFlow <rocm-install-on-linux:how-to/3rd-party/tensorflow-install>`,"2.15, 2.14, 2.13","2.14, 2.13, 2.12"
+      :doc:`JAX <rocm-install-on-linux:how-to/3rd-party/jax-install>`,0.4.26,0.4.26
+      `ONNX Runtime <https://onnxruntime.ai/docs/build/eps.html#amd-migraphx>`_,1.17.3,1.14.1
+      ,,
+      3RD PARTY COMMUNICATION LIBS:,,
+      `UCC <https://github.com/ROCm/ucc>`_,>=1.2.0,>=1.2.0
+      `UCX <https://github.com/ROCm/ucx>`_,>=1.14.1,>=1.14.1
+      ,,
+      3RD PARTY ALGORITHM LIBS:,,
+      Thrust,2.1.0,2.0.1
+      CUB,2.1.0,2.0.1
+      ,,
+      ML & COMPUTER VISION LIBS:,,
+      :doc:`Composable Kernel <composable_kernel:index>`,1.1.0,1.1.0
+      :doc:`MIGraphX <amdmigraphx:index>`,2.9.0,2.8.0
+      :doc:`MIOpen <miopen:index>`,3.1.0,3.0.0
+      :doc:`MIVisionX <mivisionx:index>`,2.5.0,2.5.0
+      :doc:`rocDecode <rocdecode:index>`,0.5.0,N/A
+      :doc:`ROCm Performance Primitives (RPP) <rpp:index>`,1.5.0,1.4.0
+      ,,
+      COMMUNICATION:,,
+      :doc:`RCCL <rccl:index>`,2.18.6,2.18.3
+      ,,
+      MATH LIBS:,,
+      `half <https://github.com/ROCm/half>`_ ,1.12.0,1.12.0
+      :doc:`hipBLAS <hipblas:index>`,2.1.0,2.0.0
+      :doc:`hipBLASLt <hipblaslt:index>`,0.7.0,0.6.0
+      :doc:`hipFFT <hipfft:index>`,1.0.14,1.0.13
+      :doc:`hipFORT <hipfort:index>`,0.4.0,0.4.0
+      :doc:`hipRAND <hiprand:index>`,2.10.16,2.10.16
+      :doc:`hipSOLVER <hipsolver:index>`,2.1.0,2.0.0
+      :doc:`hipSPARSE <hipsparse:index>`,3.0.1,3.0.0
+      :doc:`hipSPARSELt <hipsparselt:index>`,0.1.0,0.1.0
+      :doc:`rocALUTION <rocalution:index>`,3.1.1,3.0.3
+      :doc:`rocBLAS <rocblas:index>`,4.1.0,4.0.0
+      :doc:`rocFFT <rocfft:index>`,1.0.27,1.0.23
+      :doc:`rocRAND <rocrand:index>`,3.0.1,2.10.17
+      :doc:`rocSOLVER <rocsolver:index>`,3.25.0,3.24.0
+      :doc:`rocSPARSE <rocsparse:index>`,3.1.2,3.0.2
+      :doc:`rocWMMA <rocwmma:index>`,1.4.0,1.3.0
+      `Tensile <https://github.com/ROCm/Tensile>`_,4.40.0,4.39.0
+      ,,
+      PRIMITIVES:,,
+      :doc:`hipCUB <hipcub:index>`,3.1.0,3.0.0
+      :doc:`hipTensor <hiptensor:index>`,1.2.0,1.1.0
+      :doc:`rocPRIM <rocprim:index>`,3.1.0,3.0.0
+      :doc:`rocThrust <rocthrust:index>`,3.0.1,3.0.0
+      ,,
+      SUPPORT LIBS:,,
+      `hipother <https://github.com/ROCm/hipother>`_,6.1.40091,6.0.32830
+      :doc:`ROCm CMake <rocmcmakebuildtools:index>`,0.12.0,0.11.0
+      `rocm-core <https://github.com/ROCm/rocm-core>`_,6.1.0,6.0.0
+      `ROCT-Thunk-Interface <https://github.com/ROCm/ROCT-Thunk-Interface>`_,20240125.3.30,20231016.2.245
+      ,,
+      TOOLS:,,
+      :doc:`AMD SMI <amdsmi:index>`,24.4.1,23.4.2
+      :doc:`HIPIFY <hipify:index>`,17.0.0,17.0.0
+      :doc:`ROCdbgapi <rocdbgapi:index>`,0.71.0,0.71.0
+      :doc:`rocminfo <rocminfo:index>`,1.0.0,1.0.0
+      :doc:`ROCProfiler <rocprofiler:index>`,2.0.60100,2.0.0
+      `rocprofiler-register <https://github.com/ROCm/rocprofiler-register>`_,0.3.0,N/A
+      :doc:`ROCTracer <roctracer:index>`,4.1.60100,4.1.0
+      :doc:`ROCm Bandwidth Test <rocm_bandwidth_test:index>`,1.4.0,1.4.0
+      :doc:`ROCm Data Center Tool <rdc:index>`,0.3.0,0.3.0
+      :doc:`ROCm Debugger (ROCgdb) <rocgdb:index>`,14.1.0,13.2.0
+      :doc:`ROCm SMI <rocm_smi_lib:index>`,7.0.0,6.0.0
+      :doc:`ROCm Validation Suite <rocmvalidationsuite:index>`,rocm-6.1.0,rocm-6.0.0
+      :doc:`ROCr Debug Agent <rocr_debug_agent:index>`,2.0.3,2.0.3
+      :doc:`TransferBench <transferbench:index>`,1.48,1.46
+      ,,
+      COMPILERS:,,
+      `clang-ocl <https://github.com/ROCm/clang-ocl>`_,0.5.0,0.5.0
+      `Flang <https://github.com/ROCm/flang>`_,17.0.0.24103,17.0.0.23483
+      `llvm-project <https://github.com/ROCm/llvm-project>`_,17.0.0.24103,17.0.0.23483
+      `OpenMP <https://github.com/ROCm/llvm-project/tree/amd-staging/openmp>`_,17.0.0.24103,17.0.0.23483
+      ,,
+      RUNTIMES:,,
+      :doc:`HIP <hip:index>`,6.1.40091,6.0.32830
+      `OpenCL Runtime <https://github.com/ROCm/clr/tree/develop/opencl>`_,2.0.0,2.0.0
+      :doc:`ROCR-Runtime <rocr-runtime:index>`,1.13.0,1.12.0
+
+
+.. rubric:: Footnotes
+
+.. [#red-hat94] **For ROCm 6.1** - RHEL 9.4 is supported only on AMD Instinct MI300A.
+.. [#] **For ROCm 6.1** - MI300A (gfx942) is supported on Ubuntu 22.04.4, RHEL 9.4, RHEL 9.3, RHEL 8.9, and SLES 15 SP5. MI300X (gfx942) is only supported on Ubuntu 22.04.4.
+.. [#] **For ROCm 6.0** - MI300A (gfx942) is supported on Ubuntu 22.04.3, RHEL 8.9 and SLES 15 SP5. MI300X (gfx942) is only supported on Ubuntu 22.04.3.
+
+

docs/compatibility/precision-support.rst

@@ -0,0 +1,565 @@
+.. meta::
+  :description: Supported data types in ROCm
+  :keywords: int8, float8, float8 (E4M3), float8 (E5M2), bfloat8, float16, half, bfloat16, tensorfloat32, float,
+   float32, float64, double, AMD, ROCm, AMDGPU
+
+*************************************************************
+Precision support
+*************************************************************
+
+Use the following sections to identify data types and HIP types ROCm™ supports.
+
+Integral types
+==========================================
+
+The signed and unsigned integral types that are supported by ROCm are listed in the following table,
+together with their corresponding HIP type and a short description.
+
+
+.. list-table::
+    :header-rows: 1
+    :widths: 15,35,50
+
+    *
+      - 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 that are supported by ROCm are listed in the following table, together with
+their corresponding HIP type and a short description.
+
+.. image:: ../data/about/compatibility/floating-point-data-types.png
+    :alt: Supported floating-point types
+
+.. list-table::
+    :header-rows: 1
+    :widths: 15,15,70
+
+    *
+      - Type name
+      - HIP type
+      - Description
+    *
+      - float8 (E4M3)
+      - ``-``
+      - An 8-bit floating-point number that mostly follows IEEE-754 conventions and **S1E4M3** bit layout, as described in `8-bit Numerical Formats for Deep Neural Networks <https://arxiv.org/abs/2206.02915>`_ , with expanded range and with no infinity or signed zero. NaN is represented as negative zero.
+    *
+      - float8 (E5M2)
+      - ``-``
+      - An 8-bit floating-point number mostly following IEEE-754 conventions and **S1E5M2** bit layout, as described in `8-bit Numerical Formats for Deep Neural Networks <https://arxiv.org/abs/2206.02915>`_ , with expanded range and with no infinity or signed zero. NaN is represented as negative zero.
+    *
+      - 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
+      - ``-``
+      - 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.
+  * The encodings for FP8 (E5M2) and FP8 (E4M3) that are natively supported by MI300 differ from the FP8 (E5M2) and FP8 (E4M3) encodings used in 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.
+
+ROCm support icons
+==========================================
+
+In the following sections, we use icons to represent the level of support. These icons, described in the
+following table, are also used on the library data type support pages.
+
+.. list-table::
+    :header-rows: 1
+
+    *
+      -  Icon
+      - Definition
+    *
+      - ❌
+      - 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.
+
+Hardware type support
+==========================================
+
+AMD GPU hardware support for data types is listed in the following tables.
+
+Compute units support
+-------------------------------------------------------------------------------
+
+The following table lists data type support for compute units.
+
+.. tab-set::
+
+  .. tab-item:: Integral types
+    :sync: integral-type
+
+    .. list-table::
+      :header-rows: 1
+
+      *
+        - Type name
+        - int8
+        - int16
+        - int32
+        - int64
+      *
+        - MI100
+        - ✅
+        - ✅
+        - ✅
+        - ✅
+      *
+        - MI200 series
+        - ✅
+        - ✅
+        - ✅
+        - ✅
+      *
+        - MI300 series
+        - ✅
+        - ✅
+        - ✅
+        - ✅
+
+  .. tab-item:: Floating-point types
+    :sync: floating-point-type
+
+    .. list-table::
+      :header-rows: 1
+
+      *
+        - Type name
+        - float8 (E4M3)
+        - float8 (E5M2)
+        - float16
+        - bfloat16
+        - tensorfloat32
+        - float32
+        - float64
+      *
+        - MI100
+        - ❌
+        - ❌
+        - ✅
+        - ✅
+        - ❌
+        - ✅
+        - ✅
+      *
+        - MI200 series
+        - ❌
+        - ❌
+        - ✅
+        - ✅
+        - ❌
+        - ✅
+        - ✅
+      *
+        - MI300 series
+        - ❌
+        - ❌
+        - ✅
+        - ✅
+        - ❌
+        - ✅
+        - ✅
+
+Matrix core support
+-------------------------------------------------------------------------------
+
+The following table lists data type support for AMD GPU matrix cores.
+
+.. tab-set::
+
+  .. tab-item:: Integral types
+    :sync: integral-type
+
+    .. list-table::
+      :header-rows: 1
+
+      *
+        - Type name
+        - int8
+        - int16
+        - int32
+        - int64
+      *
+        - MI100
+        - ✅
+        - ❌
+        - ❌
+        - ❌
+      *
+        - MI200 series
+        - ✅
+        - ❌
+        - ❌
+        - ❌
+      *
+        - MI300 series
+        - ✅
+        - ❌
+        - ❌
+        - ❌
+
+  .. tab-item:: Floating-point types
+    :sync: floating-point-type
+
+    .. list-table::
+      :header-rows: 1
+
+      *
+        - Type name
+        - float8 (E4M3)
+        - float8 (E5M2)
+        - float16
+        - bfloat16
+        - tensorfloat32
+        - float32
+        - float64
+      *
+        - MI100
+        - ❌
+        - ❌
+        - ✅
+        - ✅
+        - ❌
+        - ✅
+        - ❌
+      *
+        - MI200 series
+        - ❌
+        - ❌
+        - ✅
+        - ✅
+        - ❌
+        - ✅
+        - ✅
+      *
+        - MI300 series
+        - ✅
+        - ✅
+        - ✅
+        - ✅
+        - ✅
+        - ✅
+        - ✅
+
+Atomic operations support
+-------------------------------------------------------------------------------
+
+The following table lists data type support for atomic operations.
+
+.. tab-set::
+
+  .. tab-item:: Integral types
+    :sync: integral-type
+
+    .. list-table::
+      :header-rows: 1
+
+      *
+        - Type name
+        - int8
+        - int16
+        - int32
+        - int64
+      *
+        - MI100
+        - ❌
+        - ❌
+        - ✅
+        - ❌
+      *
+        - MI200 series
+        - ❌
+        - ❌
+        - ✅
+        - ✅
+      *
+        - MI300 series
+        - ❌
+        - ❌
+        - ✅
+        - ✅
+
+  .. tab-item:: Floating-point types
+    :sync: floating-point-type
+
+    .. list-table::
+      :header-rows: 1
+
+      *
+        - Type name
+        - float8 (E4M3)
+        - float8 (E5M2)
+        - float16
+        - bfloat16
+        - tensorfloat32
+        - float32
+        - float64
+      *
+        - MI100
+        - ❌
+        - ❌
+        - ✅
+        - ❌
+        - ❌
+        - ✅
+        - ❌
+      *
+        - MI200 series
+        - ❌
+        - ❌
+        - ✅
+        - ❌
+        - ❌
+        - ✅
+        - ✅
+      *
+        - MI300 series
+        - ❌
+        - ❌
+        - ✅
+        - ❌
+        - ❌
+        - ✅
+        - ✅
+
+.. note::
+
+  For cases that are not natively supported, you can emulate atomic operations using software.
+  Software-emulated atomic operations have 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. For a detailed
+description, refer to the corresponding library data type support page.
+
+.. tab-set::
+
+  .. tab-item:: Integral types
+    :sync: integral-type
+
+    .. list-table::
+      :header-rows: 1
+
+      *
+        - Library input/output data type name
+        - int8
+        - int16
+        - int32
+        - int64
+      *
+        - hipSPARSELt (:doc:`details <hipsparselt:reference/data-type-support>`)
+        - ✅/✅
+        - ❌/❌
+        - ❌/❌
+        - ❌/❌
+      *
+        - rocRAND (:doc:`details <rocrand:data-type-support>`)
+        - -/✅
+        - -/✅
+        - -/✅
+        - -/✅
+      *
+        - hipRAND (:doc:`details <hiprand:data-type-support>`)
+        - -/✅
+        - -/✅
+        - -/✅
+        - -/✅
+      *
+        - rocPRIM (:doc:`details <rocprim:reference/data-type-support>`)
+        - ✅/✅
+        - ✅/✅
+        - ✅/✅
+        - ✅/✅
+      *
+        - hipCUB (:doc:`details <hipcub:data-type-support>`)
+        - ✅/✅
+        - ✅/✅
+        - ✅/✅
+        - ✅/✅
+      *
+        - rocThrust (:doc:`details <rocthrust:data-type-support>`)
+        - ✅/✅
+        - ✅/✅
+        - ✅/✅
+        - ✅/✅
+
+  .. tab-item:: Floating-point types
+    :sync: floating-point-type
+
+    .. list-table::
+      :header-rows: 1
+
+      *
+        - Library input/output data type name
+        - float8 (E4M3)
+        - float8 (E5M2)
+        - float16
+        - bfloat16
+        - tensorfloat32
+        - float32
+        - float64
+      *
+        - hipSPARSELt (:doc:`details <hipsparselt:reference/data-type-support>`)
+        - ❌/❌
+        - ❌/❌
+        - ✅/✅
+        - ✅/✅
+        - ❌/❌
+        - ❌/❌
+        - ❌/❌
+      *
+        - rocRAND (:doc:`details <rocrand:data-type-support>`)
+        - -/❌
+        - -/❌
+        - -/✅
+        - -/❌
+        - -/❌
+        - -/✅
+        - -/✅
+      *
+        - hipRAND (:doc:`details <hiprand:data-type-support>`)
+        - -/❌
+        - -/❌
+        - -/✅
+        - -/❌
+        - -/❌
+        - -/✅
+        - -/✅
+      *
+        - rocPRIM (:doc:`details <rocprim:reference/data-type-support>`)
+        - ❌/❌
+        - ❌/❌
+        - ✅/✅
+        - ✅/✅
+        - ❌/❌
+        - ✅/✅
+        - ✅/✅
+      *
+        - hipCUB (:doc:`details <hipcub:data-type-support>`)
+        - ❌/❌
+        - ❌/❌
+        - ✅/✅
+        - ✅/✅
+        - ❌/❌
+        - ✅/✅
+        - ✅/✅
+      *
+        - rocThrust (:doc:`details <rocthrust:data-type-support>`)
+        - ❌/❌
+        - ❌/❌
+        - ⚠️/⚠️
+        - ⚠️/⚠️
+        - ❌/❌
+        - ✅/✅
+        - ✅/✅
+
+
+Libraries internal calculations type support
+-------------------------------------------------------------------------------
+
+The following tables list ROCm library support for specific internal data types. For a detailed
+description, refer to the corresponding library data type support page.
+
+.. tab-set::
+
+  .. tab-item:: Integral types
+    :sync: integral-type
+
+    .. list-table::
+      :header-rows: 1
+
+      *
+        - Library internal data type name
+        - int8
+        - int16
+        - int32
+        - int64
+      *
+        - hipSPARSELt (:doc:`details <hipsparselt:reference/data-type-support>`)
+        - ❌
+        - ❌
+        - ✅
+        - ❌
+
+
+  .. tab-item:: Floating-point types
+    :sync: floating-point-type
+
+    .. list-table::
+      :header-rows: 1
+
+      *
+        - Library internal data type name
+        - float8 (E4M3)
+        - float8 (E5M2)
+        - float16
+        - bfloat16
+        - tensorfloat32
+        - float32
+        - float64
+      *
+        - hipSPARSELt (:doc:`details <hipsparselt:reference/data-type-support>`)
+        - ❌
+        - ❌
+        - ❌
+        - ❌
+        - ❌
+        - ✅
+        - ❌

docs/conceptual/More-about-how-ROCm-uses-PCIe-Atomics.rst

@@ -0,0 +1,156 @@
+.. meta::
+   :description: How ROCm uses PCIe atomics
+   :keywords: PCIe, PCIe atomics, atomics, BAR memory, AMD, ROCm
+
+*****************************************************************************
+How ROCm uses PCIe atomics
+*****************************************************************************
+
+ROCm PCIe feature and overview of BAR memory
+================================================================
+
+ROCm is an extension of HSA platform architecture, so it shares the queuing model, memory model,
+signaling and synchronization protocols. Platform atomics are integral to perform queuing and
+signaling memory operations where there may be multiple-writers across CPU and GPU agents.
+
+The full list of HSA system architecture platform requirements are here:
+`HSA Sys Arch Features <http://hsafoundation.com/wp-content/uploads/2021/02/HSA-SysArch-1.2.pdf>`_.
+
+AMD ROCm Software uses the new PCI Express 3.0 (Peripheral Component Interconnect Express [PCIe]
+3.0) features for atomic read-modify-write transactions which extends inter-processor synchronization
+mechanisms to IO to support the defined set of HSA capabilities needed for queuing and signaling
+memory operations.
+
+The new PCIe atomic operations operate as completers for ``CAS`` (Compare and Swap), ``FetchADD``,
+``SWAP`` atomics. The atomic operations are initiated by the I/O device which support 32-bit, 64-bit and
+128-bit operand which target address have to be naturally aligned to operation sizes.
+
+For ROCm the Platform atomics are used in ROCm in the following ways:
+
+  * Update HSA queue's read_dispatch_id: 64 bit atomic add used by the command processor on the
+    GPU agent to update the packet ID it processed.
+  * Update HSA queue's write_dispatch_id: 64 bit atomic add used by the CPU and GPU agent to
+    support multi-writer queue insertions.
+  * Update HSA Signals -- 64bit atomic ops are used for CPU & GPU synchronization.
+
+The PCIe 3.0 atomic operations feature allows atomic transactions to be requested by, routed through
+and completed by PCIe components. Routing and completion does not require software support.
+Component support for each is detectable via the Device Capabilities 2 (DevCap2) register. Upstream
+bridges need to have atomic operations routing enabled or the atomic operations will fail even though
+PCIe endpoint and PCIe I/O devices has the capability to atomic operations.
+
+To do atomic operations routing capability between two or more Root Ports, each associated Root Port
+must indicate that capability via the atomic operations routing supported bit in the DevCap2 register.
+
+If your system has a PCIe Express Switch it needs to support atomic operations routing. Atomic
+operations requests are permitted only if a component's ``DEVCTL2.ATOMICOP_REQUESTER_ENABLE``
+field is set. These requests can only be serviced if the upstream components support atomic operation
+completion and/or routing to a component which does. Atomic operations routing support=1, routing
+is supported; atomic operations routing support=0, routing is not supported.
+
+An atomic operation is a non-posted transaction supporting 32-bit and 64-bit address formats, there
+must be a response for Completion containing the result of the operation. Errors associated with the
+operation (uncorrectable error accessing the target location or carrying out the atomic operation) are
+signaled to the requester by setting the Completion Status field in the completion descriptor, they are
+set to to Completer Abort (CA) or Unsupported Request (UR).
+
+To understand more about how PCIe atomic operations work, see
+`PCIe atomics <https://pcisig.com/specifications/pciexpress/specifications/ECN_Atomic_Ops_080417.pdf>`_
+
+`Linux Kernel Patch to pci_enable_atomic_request <https://patchwork.kernel.org/project/linux-pci/patch/1443110390-4080-1-git-send-email-jay@jcornwall.me/>`_
+
+There are also a number of papers which talk about these new capabilities:
+
+  * `Atomic Read Modify Write Primitives by Intel <https://www.intel.es/content/dam/doc/white-paper/atomic-read-modify-write-primitives-i-o-devices-paper.pdf>`_
+  * `PCI express 3 Accelerator White paper by Intel <https://www.intel.sg/content/dam/doc/white-paper/pci-express3-accelerator-white-paper.pdf>`_
+  * `PCIe Generation 4 Base Specification includes atomic operations <https://astralvx.com/storage/2020/11/PCI_Express_Base_4.0_Rev0.3_February19-2014.pdf>`_
+  * `Xilinx PCIe Ultrascale White paper <https://docs.xilinx.com/v/u/8OZSA2V1b1LLU2rRCDVGQw>`_
+
+Other I/O devices with PCIe atomics support:
+
+  * Mellanox ConnectX-5 InfiniBand Card
+  * Cray Aries Interconnect
+  * Xilinx 7 Series Devices
+
+Future bus technology with richer I/O atomics operation Support
+
+  * GenZ
+
+New PCIe Endpoints with support beyond AMD Ryzen and EPYC CPU; Intel Haswell or newer CPUs
+with PCIe Generation 3.0 support.
+
+  * Mellanox Bluefield SOC
+  * Cavium Thunder X2
+
+In ROCm, we also take advantage of PCIe ID based ordering technology for P2P when the GPU
+originates two writes to two different targets:
+
+* Write to another GPU memory
+* Write to system memory to indicate transfer complete
+
+They are routed off to different ends of the computer but we want to make sure the write to system
+memory to indicate transfer complete occurs AFTER P2P write to GPU has complete.
+
+BAR memory overview
+----------------------------------------------------------------------------------------------------
+On a Xeon E5 based system in the BIOS we can turn on above 4GB PCIe addressing, if so he need to set
+memory-mapped input/output (MMIO) base address (MMIOH base) and range (MMIO high size) in the BIOS.
+
+In the Supermicro system in the system bios you need to see the following
+
+  * Advanced->PCIe/PCI/PnP configuration-\> Above 4G Decoding = Enabled
+  * Advanced->PCIe/PCI/PnP Configuration-\>MMIOH Base = 512G
+  * Advanced->PCIe/PCI/PnP Configuration-\>MMIO High Size = 256G
+
+When we support Large Bar Capability there is a Large Bar VBIOS which also disable the IO bar.
+
+For GFX9 and Vega10 which have Physical Address up 44 bit and 48 bit Virtual address.
+
+  * BAR0-1 registers: 64bit, prefetchable, GPU memory. 8GB or 16GB depending on Vega10 SKU. Must
+    be placed < 2^44 to support P2P  	access from other Vega10.
+  * BAR2-3 registers: 64bit, prefetchable, Doorbell. Must be placed \< 2^44 to support P2P access from
+    other Vega10.
+  * BAR4 register: Optional, not a boot device.
+  * BAR5 register: 32bit, non-prefetchable, MMIO. Must be placed \< 4GB.
+
+Here is how our base address register (BAR) works on GFX 8 GPUs with 40 bit Physical Address Limit ::
+
+  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]
+
+Legend:
+
+1 : GPU Frame Buffer BAR -- In this example it happens to be 256M, but typically this will be size of the
+GPU memory (typically 4GB+). This BAR has to be placed \< 2^40 to allow peer-to-peer access from
+other GFX8 AMD GPUs. For GFX9 (Vega GPU) the BAR has to be placed \< 2^44 to allow peer-to-peer
+access from other GFX9 AMD GPUs.
+
+2 : Doorbell BAR -- The size of the BAR is typically will be \< 10MB (currently fixed at 2MB) for this
+generation GPUs. This BAR has to be placed \< 2^40 to allow peer-to-peer access from other current
+generation AMD GPUs.
+
+3 : IO BAR -- This is for legacy VGA and boot device support, but since this the GPUs in this project are
+not VGA devices (headless), this is not a concern even if the SBIOS does not setup.
+
+4 : MMIO BAR -- This is required for the AMD Driver SW to access the configuration registers. Since the
+reminder of the BAR available is only 1 DWORD (32bit), this is placed \< 4GB. This is fixed at 256KB.
+
+5 : Expansion ROM -- This is required for the AMD Driver SW to access the GPU video-bios. This is
+currently fixed at 128KB.
+
+For more information, you can review
+`Overview of Changes to PCI Express 3.0 <https://www.mindshare.com/files/resources/PCIe%203-0.pdf>`_.

docs/conceptual/ai-migraphx-optimization.md

@@ -0,0 +1,333 @@
+<head>
+  <meta charset="UTF-8">
+  <meta name="description" content="Inference optimization with MIGraphX">
+  <meta name="keywords" content="Inference optimization, MIGraphX, deep-learning, MIGraphX
+  installation, AMD, ROCm">
+</head>
+
+# Inference optimization with MIGraphX
+
+The following sections cover inferencing and introduces [MIGraphX](https://rocm.docs.amd.com/projects/AMDMIGraphX/en/latest/).
+
+## Inference
+
+The inference is where capabilities learned during deep-learning training are put to work. It refers to using a fully trained neural network to make conclusions (predictions) on unseen data that the model has never interacted with before. Deep-learning inferencing is achieved by feeding new data, such as new images, to the network, giving the Deep Neural Network a chance to classify the image.
+
+Taking our previous example of MNIST, the DNN can be fed new images of handwritten digit images, allowing the neural network to classify digits. A fully trained DNN should make accurate predictions about what an image represents, and inference cannot happen without training.
+
+## MIGraphX introduction
+
+MIGraphX is a graph compiler focused on accelerating the machine-learning inference that can target AMD GPUs and CPUs. MIGraphX accelerates the machine-learning models by leveraging several graph-level transformations and optimizations. These optimizations include:
+
+* Operator fusion
+* Arithmetic simplifications
+* Dead-code elimination
+* Common subexpression elimination (CSE)
+* Constant propagation
+
+After doing all these transformations, MIGraphX emits code for the AMD GPU by calling to MIOpen or rocBLAS or creating HIP kernels for a particular operator. MIGraphX can also target CPUs using DNNL or ZenDNN libraries.
+
+MIGraphX provides easy-to-use APIs in C++ and Python to import machine models in ONNX or TensorFlow. Users can compile, save, load, and run these models using the MIGraphX C++ and Python APIs. Internally, MIGraphX parses ONNX or TensorFlow models into internal graph representation where each operator in the model gets mapped to an operator within MIGraphX. Each of these operators defines various attributes such as:
+
+* Number of arguments
+* Type of arguments
+* Shape of arguments
+
+After optimization passes, all these operators get mapped to different kernels on GPUs or CPUs.
+
+After importing a model into MIGraphX, the model is represented as `migraphx::program`. `migraphx::program` is made up of `migraphx::module`. The program can consist of several modules, but it always has one main_module. Modules are made up of `migraphx::instruction_ref`. Instructions contain the `migraphx::op` and arguments to the operator.  
+
+## Installing MIGraphX
+
+There are three options to get started with MIGraphX installation. MIGraphX depends on ROCm libraries; assume that the machine has ROCm installed.
+
+### Option 1: installing binaries
+
+To install MIGraphX on Debian-based systems like Ubuntu, use the following command:
+
+```bash
+sudo apt update && sudo apt install -y migraphx
+```
+
+The header files and libraries are installed under `/opt/rocm-\<version\>`, where \<version\> is the ROCm version.
+
+### Option 2: building from source
+
+There are two ways to build the MIGraphX sources.
+
+* [Use the ROCm build tool](https://github.com/ROCm/AMDMIGraphX#use-the-rocm-build-tool-rbuild) - This approach uses `[rbuild](https://github.com/ROCm/rbuild)` to install the prerequisites and build the libraries with just one command.
+
+  or
+
+* [Use CMake](https://github.com/ROCm/AMDMIGraphX#use-cmake-to-build-migraphx) - This approach uses a script to install the prerequisites, then uses CMake to build the source.
+
+For detailed steps on building from source and installing dependencies, refer to the following `README` file:
+
+[https://github.com/ROCm/AMDMIGraphX#building-from-source](https://github.com/ROCm/AMDMIGraphX#building-from-source)
+
+### Option 3: use docker
+
+To use Docker, follow these steps:
+
+1. The easiest way to set up the development environment is to use Docker. To build Docker from scratch, first clone the MIGraphX repository by running:
+
+    ```bash
+    git clone --recursive https://github.com/ROCm/AMDMIGraphX
+    ```
+
+2. The repository contains a Dockerfile from which you can build a Docker image as:
+
+    ```bash
+    docker build -t migraphx .
+    ```
+
+3. Then to enter the development environment, use Docker run:
+
+    ```bash
+    docker run --device='/dev/kfd' --device='/dev/dri' -v=`pwd`:/code/AMDMIGraphX -w /code/AMDMIGraphX --group-add video -it migraphx
+    ```
+
+The Docker image contains all the prerequisites required for the installation, so users can go to the folder `/code/AMDMIGraphX` and follow the steps mentioned in [Option 2: Building from Source](#option-2-building-from-source).
+
+## MIGraphX example
+
+MIGraphX provides both C++ and Python APIs. The following sections show examples of both using the Inception v3 model. To walk through the examples, fetch the Inception v3 ONNX model by running the following:
+
+```py
+import torch
+import torchvision.models as models
+inception = models.inception_v3(pretrained=True)
+torch.onnx.export(inception,torch.randn(1,3,299,299), "inceptioni1.onnx")
+```
+
+This will create `inceptioni1.onnx`, which can be imported in MIGraphX using C++ or Python API.
+
+### MIGraphX Python API
+
+Follow these steps:
+
+1. To import the MIGraphX module in Python script, set `PYTHONPATH` to the MIGraphX libraries installation. If binaries are installed using steps mentioned in [Option 1: Installing Binaries](#option-1-installing-binaries), perform the following action:
+
+    ```bash
+    export PYTHONPATH=$PYTHONPATH:/opt/rocm/
+    ```
+
+2. The following script shows the usage of Python API to import the ONNX model, compile it, and run inference on it. Set `LD_LIBRARY_PATH` to `/opt/rocm/` if required.
+
+    ```py
+    # import migraphx and numpy
+    import migraphx
+    import numpy as np
+    # import and parse inception model
+    model = migraphx.parse_onnx("inceptioni1.onnx")
+    # compile model for the GPU target
+    model.compile(migraphx.get_target("gpu"))
+    # optionally print compiled model
+    model.print()
+    # create random input image
+    input_image = np.random.rand(1, 3, 299, 299).astype('float32')
+    # feed image to model, 'x.1` is the input param name
+    results = model.run({'x.1': input_image})
+    # get the results back
+    result_np = np.array(results[0])
+    # print the inferred class of the input image
+    print(np.argmax(result_np))
+    ```
+
+    Find additional examples of Python API in the `/examples` directory of the MIGraphX repository.
+
+## MIGraphX C++ API
+
+Follow these steps:
+
+1. The following is a minimalist example that shows the usage of MIGraphX C++ API to load ONNX file, compile it for the GPU, and run inference on it. To use MIGraphX C++ API, you only need to load the `migraphx.hpp` file. This example runs inference on the Inception v3 model.
+
+    ```c++
+    #include <vector>
+    #include <string>
+    #include <algorithm>
+    #include <ctime>
+    #include <random>
+    #include <migraphx/migraphx.hpp>
+
+    int main(int argc, char** argv)
+    {
+        migraphx::program prog;
+        migraphx::onnx_options onnx_opts;
+        // import and parse onnx file into migraphx::program
+        prog = parse_onnx("inceptioni1.onnx", onnx_opts);
+        // print imported model
+        prog.print();
+        migraphx::target targ = migraphx::target("gpu");
+        migraphx::compile_options comp_opts;
+        comp_opts.set_offload_copy();
+        // compile for the GPU
+        prog.compile(targ, comp_opts);
+        // print the compiled program
+        prog.print();
+        // randomly generate input image
+        // of shape (1, 3, 299, 299)
+        std::srand(unsigned(std::time(nullptr)));
+        std::vector<float> input_image(1*299*299*3);
+        std::generate(input_image.begin(), input_image.end(), std::rand);
+        // users need to provide data for the input
+        // parameters in order to run inference
+        // you can query into migraph program for the parameters
+        migraphx::program_parameters prog_params;
+        auto param_shapes = prog.get_parameter_shapes();
+        auto input        = param_shapes.names().front();
+        // create argument for the parameter
+        prog_params.add(input, migraphx::argument(param_shapes[input], input_image.data()));
+        // run inference
+        auto outputs = prog.eval(prog_params);
+        // read back the output
+        float* results = reinterpret_cast<float*>(outputs[0].data());
+        float* max     = std::max_element(results, results + 1000);
+        int answer = max - results;
+        std::cout << "answer: " << answer << std::endl;
+    }
+    ```
+
+2. To compile this program, you can use CMake and you only need to link the `migraphx::c` library to use MIGraphX's C++ API. The following is the `CMakeLists.txt` file that can build the earlier example:
+
+    ```cmake
+    cmake_minimum_required(VERSION 3.5)
+    project (CAI)
+
+    set (CMAKE_CXX_STANDARD 14)
+    set (EXAMPLE inception_inference)
+
+    list (APPEND CMAKE_PREFIX_PATH /opt/rocm/hip /opt/rocm)
+    find_package (migraphx)
+
+    message("source file: " ${EXAMPLE}.cpp " ---> bin: " ${EXAMPLE})
+    add_executable(${EXAMPLE} ${EXAMPLE}.cpp)
+
+    target_link_libraries(${EXAMPLE} migraphx::c)
+    ```
+
+3. To build the executable file, run the following from the directory containing the `inception_inference.cpp` file:
+
+    ```bash
+    mkdir build
+    cd build
+    cmake ..
+    make -j$(nproc)
+    ./inception_inference
+    ```
+
+:::{note}
+    Set `LD_LIBRARY_PATH` to `/opt/rocm/lib` if required during the build. Additional examples can be found in the MIGraphX repository under the `/examples/` directory.
+:::
+
+## Tuning MIGraphX
+
+MIGraphX uses MIOpen kernels to target AMD GPU. For the model compiled with MIGraphX, tune MIOpen to pick the best possible kernel implementation. The MIOpen tuning results in a significant performance boost. Tuning can be done by setting the environment variable `MIOPEN_FIND_ENFORCE=3`.
+
+:::{note}
+    The tuning process can take a long time to finish.
+:::
+
+**Example:** The average inference time of the inception model example shown previously over 100 iterations using untuned kernels is 0.01383ms. After tuning, it reduces to 0.00459ms, which is a 3x improvement. This result is from ROCm v4.5 on a MI100 GPU.
+
+:::{note}
+    The results may vary depending on the system configurations.
+:::
+
+For reference, the following code snippet shows inference runs for only the first 10 iterations for both tuned and untuned kernels:
+
+```console
+### UNTUNED ###
+iterator : 0
+Inference complete
+Inference time: 0.063ms
+iterator : 1
+Inference complete
+Inference time: 0.008ms
+iterator : 2
+Inference complete
+Inference time: 0.007ms
+iterator : 3
+Inference complete
+Inference time: 0.007ms
+iterator : 4
+Inference complete
+Inference time: 0.007ms
+iterator : 5
+Inference complete
+Inference time: 0.008ms
+iterator : 6
+Inference complete
+Inference time: 0.007ms
+iterator : 7
+Inference complete
+Inference time: 0.028ms
+iterator : 8
+Inference complete
+Inference time: 0.029ms
+iterator : 9
+Inference complete
+Inference time: 0.029ms
+
+### TUNED ###
+iterator : 0
+Inference complete
+Inference time: 0.063ms
+iterator : 1
+Inference complete
+Inference time: 0.004ms
+iterator : 2
+Inference complete
+Inference time: 0.004ms
+iterator : 3
+Inference complete
+Inference time: 0.004ms
+iterator : 4
+Inference complete
+Inference time: 0.004ms
+iterator : 5
+Inference complete
+Inference time: 0.004ms
+iterator : 6
+Inference complete
+Inference time: 0.004ms
+iterator : 7
+Inference complete
+Inference time: 0.004ms
+iterator : 8
+Inference complete
+Inference time: 0.004ms
+iterator : 9
+Inference complete
+Inference time: 0.004ms
+```
+
+### YModel
+
+The best inference performance through MIGraphX is conditioned upon having tuned kernel configurations stored in a `/home` local User Database (DB). If a user were to move their model to a different server or allow a different user to use it, they would have to run through the MIOpen tuning process again to populate the next User DB with the best kernel configurations and corresponding solvers.
+
+Tuning is time consuming, and if the users have not performed tuning, they would see discrepancies between expected or claimed inference performance and actual inference performance. This has led to repetitive and time-consuming tuning tasks for each user.
+
+MIGraphX introduces a feature, known as YModel, that stores the kernel config parameters found during tuning into a `.mxr` file. This ensures the same level of expected performance, even when a model is copied to a different user/system.
+
+The YModel feature is available starting from ROCm 5.4.1 and UIF 1.1.
+
+#### YModel example
+
+Through the `migraphx-driver` functionality, you can generate `.mxr` files with tuning information stored inside it by passing additional `--binary --output model.mxr` to `migraphx-driver` along with the rest of the necessary flags.
+
+For example, to generate `.mxr` file from the ONNX model, use the following:
+
+```bash
+./path/to/migraphx-driver compile --onnx resnet50.onnx --enable-offload-copy --binary --output resnet50.mxr
+```
+
+To run generated `.mxr` files through `migraphx-driver`, use the following:
+
+```bash
+./path/to/migraphx-driver run --migraphx resnet50.mxr --enable-offload-copy
+```
+
+Alternatively, you can use the MIGraphX C++ or Python API to generate `.mxr` files.
+
+![Generating an MXR file](../data/conceptual/image018.png "Generating an MXR file")

docs/conceptual/cmake-packages.rst

@@ -0,0 +1,407 @@
+.. meta::
+   :description: Using CMake
+   :keywords: CMake, dependencies, HIP, C++, AMD, ROCm
+
+*********************************
+Using CMake
+*********************************
+
+Most components in ROCm support CMake. Projects depending on header-only or
+library components typically require CMake 3.5 or higher whereas those wanting
+to make use of the CMake HIP language support will require CMake 3.21 or higher.
+
+Finding dependencies
+====================
+
+.. note::
+
+  For a complete
+  reference on how to deal with dependencies in CMake, refer to the CMake docs
+  on `find_package
+  <https://cmake.org/cmake/help/latest/command/find_package.html>`_ and the
+  `Using Dependencies Guide
+  <https://cmake.org/cmake/help/latest/guide/using-dependencies/index.html>`_
+  to get an overview of CMake related facilities.
+
+In short, CMake supports finding dependencies in two ways:
+
+* In Module mode, it consults a file ``Find<PackageName>.cmake`` which tries to find the component
+  in typical install locations and layouts. CMake ships a few dozen such scripts, but users and projects
+  may ship them as well.
+
+* In Config mode, it locates a file named ``<packagename>-config.cmake`` or
+  ``<PackageName>Config.cmake`` which describes the installed component in all regards needed to
+  consume it.
+
+ROCm predominantly relies on Config mode, one notable exception being the Module
+driving the compilation of HIP programs on NVIDIA runtimes. As such, when
+dependencies are not found in standard system locations, one either has to
+instruct CMake to search for package config files in additional folders using
+the ``CMAKE_PREFIX_PATH`` variable (a semi-colon separated list of file system
+paths), or using ``<PackageName>_ROOT`` variable on a project-specific basis.
+
+There are nearly a dozen ways to set these variables. One may be more convenient
+over the other depending on your workflow. Conceptually the simplest is adding
+it to your CMake configuration command on the command line via
+``-D CMAKE_PREFIX_PATH=....`` . AMD packaged ROCm installs can typically be
+added to the config file search paths such as:
+
+*  Windows: ``-D CMAKE_PREFIX_PATH=${env:HIP_PATH}``
+
+*  Linux: ``-D CMAKE_PREFIX_PATH=/opt/rocm``
+
+ROCm provides the respective *config-file* packages, and this enables
+``find_package`` to be used directly. ROCm does not require any Find module as
+the *config-file* packages are shipped with the upstream projects, such as
+rocPRIM and other ROCm libraries.
+
+For a complete guide on where and how ROCm may be installed on a system, refer
+to the installation guides for
+`Linux <https://rocm.docs.amd.com/projects/install-on-linux/en/latest/tutorial/quick-start.html>`_
+and
+`Windows <https://rocm.docs.amd.com/projects/install-on-windows/en/latest/index.html>`_.
+
+Using HIP in CMake
+==================
+
+ROCm components providing a C/C++ interface support consumption via any
+C/C++ toolchain that CMake knows how to drive. ROCm also supports the CMake HIP
+language features, allowing users to program using the HIP single-source
+programming model. When a program (or translation-unit) uses the HIP API without
+compiling any GPU device code, HIP can be treated in CMake as a simple C/C++
+library.
+
+Using the HIP single-source programming model
+---------------------------------------------
+
+Source code written in the HIP dialect of C++ typically uses the `.hip`
+extension. When the HIP CMake language is enabled, it will automatically
+associate such source files with the HIP toolchain being used.
+
+.. code-block:: cmake
+
+  cmake_minimum_required(VERSION 3.21) # HIP language support requires 3.21
+  cmake_policy(VERSION 3.21.3...3.27)
+  project(MyProj LANGUAGES HIP)
+  add_executable(MyApp Main.hip)
+
+Should you have existing CUDA code that is from the source compatible subset of
+HIP, you can tell CMake that despite their `.cu` extension, they're HIP sources.
+Do note that this mostly facilitates compiling kernel code-only source files,
+as host-side CUDA API won't compile in this fashion.
+
+.. code-block:: cmake
+
+  add_library(MyLib MyLib.cu)
+  set_source_files_properties(MyLib.cu PROPERTIES LANGUAGE HIP)
+
+CMake itself only hosts part of the HIP language support, such as defining
+HIP-specific properties, etc. while the other half ships with the HIP
+implementation, such as ROCm. CMake will search for a file
+`hip-lang-config.cmake` describing how the the properties defined by CMake
+translate to toolchain invocations. If one installs ROCm using non-standard
+methods or layouts and CMake can't locate this file or detect parts of the SDK,
+there's a catch-all, last resort variable consulted locating this file,
+``-D CMAKE_HIP_COMPILER_ROCM_ROOT:PATH=`` which should be set the root of the
+ROCm installation.
+
+.. note::
+    Imported targets defined by `hip-lang-config.cmake` are for internal use
+    only.
+
+If the user doesn't provide a semi-colon delimited list of device architectures
+via ``CMAKE_HIP_ARCHITECTURES``, CMake will select some sensible default. It is
+advised though that if a user knows what devices they wish to target, then set
+this variable explicitly.
+
+Consuming ROCm C/C++ libraries
+------------------------------
+
+Libraries such as rocBLAS, rocFFT, MIOpen, etc. behave as C/C++ libraries.
+Illustrated in the example below is a C++ application using MIOpen from CMake.
+It calls ``find_package(miopen)``, which provides the ``MIOpen`` imported
+target. This can be linked with ``target_link_libraries``
+
+.. code-block:: cmake
+
+  cmake_minimum_required(VERSION 3.5) # find_package(miopen) requires 3.5
+  cmake_policy(VERSION 3.5...3.27)
+  project(MyProj LANGUAGES CXX)
+  find_package(miopen)
+  add_library(MyLib ...)
+  target_link_libraries(MyLib PUBLIC MIOpen)
+
+.. note::
+
+  Most libraries are designed as host-only API, so using a GPU device
+  compiler is not necessary for downstream projects unless they use GPU device
+  code.
+
+Consuming the HIP API in C++ code
+---------------------------------
+
+Consuming the HIP API without compiling single-source GPU device code can be
+done using any C++ compiler. The ``find_package(hip)`` provides the
+``hip::host`` imported target to use HIP in this scenario.
+
+.. code-block:: cmake
+
+  cmake_minimum_required(VERSION 3.5) # find_package(hip) requires 3.5
+  cmake_policy(VERSION 3.5...3.27)
+  project(MyProj LANGUAGES CXX)
+  find_package(hip REQUIRED)
+  add_executable(MyApp ...)
+  target_link_libraries(MyApp PRIVATE hip::host)
+
+When mixing such ``CXX`` sources with ``HIP`` sources holding device-code, link
+only to `hip::host`. If HIP sources don't have `.hip` as their extension, use
+`set_source_files_properties(<hip_sources>... PROPERTIES LANGUAGE HIP)` on them.
+Linking to `hip::host` will set all the necessary flags for the ``CXX`` sources
+while ``HIP`` sources inherit all flags from the built-in language support.
+Having HIP sources in a target will turn the |LINK_LANG|_ into ``HIP``.
+
+.. |LINK_LANG| replace:: ``LINKER_LANGUAGE``
+.. _LINK_LANG: https://cmake.org/cmake/help/latest/prop_tgt/LINKER_LANGUAGE.html
+
+Compiling device code in C++ language mode
+------------------------------------------
+
+.. attention::
+
+  The workflow detailed here is considered legacy and is shown for
+  understanding's sake. It pre-dates the existence of HIP language support in
+  CMake. If source code has HIP device code in it, it is a HIP source file
+  and should be compiled as such. Only resort to the method below if your
+  HIP-enabled CMake code path can't mandate CMake version 3.21.
+
+If code uses the HIP API and compiles GPU device code, it requires using a
+device compiler. The compiler for CMake can be set using either the
+``CMAKE_C_COMPILER`` and ``CMAKE_CXX_COMPILER`` variable or using the ``CC``
+and ``CXX`` environment variables. This can be set when configuring CMake or
+put into a CMake toolchain file. The device compiler must be set to a
+compiler that supports AMD GPU targets, which is usually Clang.
+
+The ``find_package(hip)`` provides the ``hip::device`` imported target to add
+all the flags necessary for device compilation.
+
+.. code-block:: cmake
+
+  cmake_minimum_required(VERSION 3.8) # cxx_std_11 requires 3.8
+  cmake_policy(VERSION 3.8...3.27)
+  project(MyProj LANGUAGES CXX)
+  find_package(hip REQUIRED)
+  add_library(MyLib ...)
+  target_link_libraries(MyLib PRIVATE hip::device)
+  target_compile_features(MyLib PRIVATE cxx_std_11)
+
+.. note::
+
+  Compiling for the GPU device requires at least C++11.
+
+This project can then be configured with the following CMake commands:
+
+*  Windows: ``cmake -D CMAKE_CXX_COMPILER:PATH=${env:HIP_PATH}\bin\clang++.exe``
+*  Linux: ``cmake -D CMAKE_CXX_COMPILER:PATH=/opt/rocm/bin/amdclang++``
+
+Which use the device compiler provided from the binary packages of
+`ROCm HIP SDK <https://www.amd.com/en/developer/resources/rocm-hub/hip-sdk.html>`_ and
+`repo.radeon.com <https://repo.radeon.com>`_ respectively.
+
+When using the ``CXX`` language support to compile HIP device code, selecting the
+target GPU architectures is done via setting the ``GPU_TARGETS`` variable.
+``CMAKE_HIP_ARCHITECTURES`` only exists when the HIP language is enabled. By
+default, this is set to some subset of the currently supported architectures of
+AMD ROCm. It can be set to the CMake option ``-D GPU_TARGETS="gfx1032;gfx1035"``.
+
+ROCm CMake packages
+-------------------
+
++-----------+----------+--------------------------------------------------------+
+| Component | Package  | Targets                                                |
++===========+==========+========================================================+
+| HIP       | hip      | ``hip::host``, ``hip::device``                         |
++-----------+----------+--------------------------------------------------------+
+| rocPRIM   | rocprim  | ``roc::rocprim``                                       |
++-----------+----------+--------------------------------------------------------+
+| rocThrust | rocthrust| ``roc::rocthrust``                                     |
++-----------+----------+--------------------------------------------------------+
+| hipCUB    | hipcub   | ``hip::hipcub``                                        |
++-----------+----------+--------------------------------------------------------+
+| rocRAND   | rocrand  | ``roc::rocrand``                                       |
++-----------+----------+--------------------------------------------------------+
+| rocBLAS   | rocblas  | ``roc::rocblas``                                       |
++-----------+----------+--------------------------------------------------------+
+| rocSOLVER | rocsolver| ``roc::rocsolver``                                     |
++-----------+----------+--------------------------------------------------------+
+| hipBLAS   | hipblas  | ``roc::hipblas``                                       |
++-----------+----------+--------------------------------------------------------+
+| rocFFT    | rocfft   | ``roc::rocfft``                                        |
++-----------+----------+--------------------------------------------------------+
+| hipFFT    | hipfft   | ``hip::hipfft``                                        |
++-----------+----------+--------------------------------------------------------+
+| rocSPARSE | rocsparse| ``roc::rocsparse``                                     |
++-----------+----------+--------------------------------------------------------+
+| hipSPARSE | hipsparse| ``roc::hipsparse``                                     |
++-----------+----------+--------------------------------------------------------+
+| rocALUTION|rocalution| ``roc::rocalution``                                    |
++-----------+----------+--------------------------------------------------------+
+| RCCL      | rccl     | ``rccl``                                               |
++-----------+----------+--------------------------------------------------------+
+| MIOpen    | miopen   | ``MIOpen``                                             |
++-----------+----------+--------------------------------------------------------+
+| MIGraphX  | migraphx | ``migraphx::migraphx``, ``migraphx::migraphx_c``,      |
+|           |          | ``migraphx::migraphx_cpu``, ``migraphx::migraphx_gpu``,|
+|           |          | ``migraphx::migraphx_onnx``, ``migraphx::migraphx_tf`` |
++-----------+----------+--------------------------------------------------------+
+
+Using CMake presets
+===================
+
+CMake command lines depending on how specific users like to be when compiling
+code can grow to unwieldy lengths. This is the primary reason why projects tend
+to bake script snippets into their build definitions controlling compiler
+warning levels, changing CMake defaults (``CMAKE_BUILD_TYPE`` or
+``BUILD_SHARED_LIBS`` just to name a few) and all sorts anti-patterns, all in
+the name of convenience.
+
+Load on the command-line interface (CLI) starts immediately by selecting a
+toolchain, the set of utilities used to compile programs. To ease some of the
+toolchain related pains, CMake does consult the ``CC`` and ``CXX`` environmental
+variables when setting a default ``CMAKE_C[XX]_COMPILER`` respectively, but that
+is just the tip of the iceberg. There's a fair number of variables related to
+just the toolchain itself (typically supplied using
+`toolchain files <https://cmake.org/cmake/help/latest/manual/cmake-toolchains.7.html>`_
+), and then we still haven't talked about user preference or project-specific
+options.
+
+IDEs supporting CMake (Visual Studio, Visual Studio Code, CLion, etc.) all came
+up with their own way to register command-line fragments of different purpose in
+a setup-and-forget fashion for quick assembly using graphical front-ends. This is
+all nice, but configurations aren't portable, nor can they be reused in
+Continuous Integration (CI) pipelines. CMake has condensed existing practice
+into a portable JSON format that works in all IDEs and can be invoked from any
+command line. This is
+`CMake Presets <https://cmake.org/cmake/help/latest/manual/cmake-presets.7.html>`_.
+
+There are two types of preset files: one supplied by the project, called
+``CMakePresets.json`` which is meant to be committed to version control,
+typically used to drive CI; and one meant for the user to provide, called
+``CMakeUserPresets.json``, typically used to house user preference and adapting
+the build to the user's environment. These JSON files are allowed to include
+other JSON files and the user presets always implicitly includes the non-user
+variant.
+
+Using HIP with presets
+----------------------
+
+Following is an example ``CMakeUserPresets.json`` file which actually compiles
+the `amd/rocm-examples <https://github.com/amd/rocm-examples>`_ suite of sample
+applications on a typical ROCm installation:
+
+.. code-block:: json
+
+  {
+    "version": 3,
+    "cmakeMinimumRequired": {
+      "major": 3,
+      "minor": 21,
+      "patch": 0
+    },
+    "configurePresets": [
+      {
+        "name": "layout",
+        "hidden": true,
+        "binaryDir": "${sourceDir}/build/${presetName}",
+        "installDir": "${sourceDir}/install/${presetName}"
+      },
+      {
+        "name": "generator-ninja-multi-config",
+        "hidden": true,
+        "generator": "Ninja Multi-Config"
+      },
+      {
+        "name": "toolchain-makefiles-c/c++-amdclang",
+        "hidden": true,
+        "cacheVariables": {
+          "CMAKE_C_COMPILER": "/opt/rocm/bin/amdclang",
+          "CMAKE_CXX_COMPILER": "/opt/rocm/bin/amdclang++",
+          "CMAKE_HIP_COMPILER": "/opt/rocm/bin/amdclang++"
+        }
+      },
+      {
+        "name": "clang-strict-iso-high-warn",
+        "hidden": true,
+        "cacheVariables": {
+          "CMAKE_C_FLAGS": "-Wall -Wextra -pedantic",
+          "CMAKE_CXX_FLAGS": "-Wall -Wextra -pedantic",
+          "CMAKE_HIP_FLAGS": "-Wall -Wextra -pedantic"
+        }
+      },
+      {
+        "name": "ninja-mc-rocm",
+        "displayName": "Ninja Multi-Config ROCm",
+        "inherits": [
+          "layout",
+          "generator-ninja-multi-config",
+          "toolchain-makefiles-c/c++-amdclang",
+          "clang-strict-iso-high-warn"
+        ]
+      }
+    ],
+    "buildPresets": [
+      {
+        "name": "ninja-mc-rocm-debug",
+        "displayName": "Debug",
+        "configuration": "Debug",
+        "configurePreset": "ninja-mc-rocm"
+      },
+      {
+        "name": "ninja-mc-rocm-release",
+        "displayName": "Release",
+        "configuration": "Release",
+        "configurePreset": "ninja-mc-rocm"
+      },
+      {
+        "name": "ninja-mc-rocm-debug-verbose",
+        "displayName": "Debug (verbose)",
+        "configuration": "Debug",
+        "configurePreset": "ninja-mc-rocm",
+        "verbose": true
+      },
+      {
+        "name": "ninja-mc-rocm-release-verbose",
+        "displayName": "Release (verbose)",
+        "configuration": "Release",
+        "configurePreset": "ninja-mc-rocm",
+        "verbose": true
+      }
+    ],
+    "testPresets": [
+      {
+        "name": "ninja-mc-rocm-debug",
+        "displayName": "Debug",
+        "configuration": "Debug",
+        "configurePreset": "ninja-mc-rocm",
+        "execution": {
+          "jobs": 0
+        }
+      },
+      {
+        "name": "ninja-mc-rocm-release",
+        "displayName": "Release",
+        "configuration": "Release",
+        "configurePreset": "ninja-mc-rocm",
+        "execution": {
+          "jobs": 0
+        }
+      }
+    ]
+  }
+
+.. note::
+
+  Getting presets to work reliably on Windows requires some CMake improvements
+  and/or support from compiler vendors. (Refer to
+  `Add support to the Visual Studio generators <https://gitlab.kitware.com/cmake/cmake/-/issues/24245>`_
+  and `Sourcing environment scripts <https://gitlab.kitware.com/cmake/cmake/-/issues/21619>`_
+  .)

docs/conceptual/compiler-disambiguation.md

@@ -0,0 +1,21 @@
+<head>
+  <meta charset="UTF-8">
+  <meta name="description" content="ROCm compilers disambiguation">
+  <meta name="keywords" content="compilers, compiler naming, AMD, ROCm">
+</head>
+
+# ROCm compilers disambiguation
+
+ROCm ships multiple compilers of varying origins and purposes. This article
+disambiguates compiler naming used throughout the documentation.
+
+## Compiler terms
+
+| Term | Description |
+| - | - |
+| `amdclang++` | Clang/LLVM-based compiler that is part of `rocm-llvm` package. The source code is available at <a href="https://github.com/ROCm/llvm-project" target="_blank">https://github.com/ROCm/llvm-project</a>. |
+| AOCC | Closed-source clang-based compiler that includes additional CPU optimizations. Offered as part of ROCm via the `rocm-llvm-alt` package. See for details, <a href="https://developer.amd.com/amd-aocc/" target="_blank">https://developer.amd.com/amd-aocc/</a>. |
+| HIP-Clang | Informal term for the `amdclang++` compiler |
+| HIPIFY | Tools including `hipify-clang` and `hipify-perl`, used to automatically translate CUDA source code into portable HIP C++. The source code is available at <a href="https://github.com/ROCm/HIPIFY" target="_blank">https://github.com/ROCm/HIPIFY</a> |
+| `hipcc` | HIP compiler driver. A utility that invokes `clang` or `nvcc` depending on the target and passes the appropriate include and library options for the target compiler and HIP infrastructure. The source code is available at <a href="https://github.com/ROCm/HIPCC" target="_blank">https://github.com/ROCm/HIPCC</a>. |
+| ROCmCC | Clang/LLVM-based compiler. ROCmCC in itself is not a binary but refers to the overall compiler. |

docs/conceptual/compiler-topics.md

@@ -0,0 +1,14 @@
+<head>
+  <meta charset="UTF-8">
+  <meta name="description" content="AMD ROCm documentation">
+  <meta name="keywords" content="documentation, guides, installation, compatibility, support,
+  reference, ROCm, AMD">
+</head>
+
+# Using compiler features
+
+The following topics describe using specific features of the compilation tools:
+
+* [Using AddressSanitizer](./using-gpu-sanitizer.md)
+* [Compiler disambiguation](./compiler-disambiguation.md)
+* [OpenMP support in ROCm](../about/compatibility/openmp.md)

docs/conceptual/file-reorg.md

@@ -0,0 +1,172 @@
+<head>
+  <meta charset="UTF-8">
+  <meta name="description" content="ROCm Linux Filesystem Hierarchy Standard reorganization">
+  <meta name="keywords" content="FHS, Linux Filesystem Hierarchy Standard, directory structure,
+  AMD, ROCm">
+</head>
+
+# ROCm Linux Filesystem Hierarchy Standard reorganization
+
+## Introduction
+
+The ROCm Software has adopted the Linux Filesystem Hierarchy Standard (FHS) [https://refspecs.linuxfoundation.org/FHS_3.0/fhs/index.html](https://refspecs.linuxfoundation.org/FHS_3.0/fhs/index.html) in order to to ensure ROCm is consistent with standard open source conventions. The following sections specify how current and future releases of ROCm adhere to FHS, how the previous ROCm file system is supported, and how improved versioning specifications are applied to ROCm.
+
+## Adopting the FHS
+
+In order to standardize ROCm directory structure and directory content layout ROCm has adopted the [FHS](https://refspecs.linuxfoundation.org/FHS_3.0/fhs/index.html), adhering to open source conventions for Linux-based distribution. FHS ensures internal consistency within the ROCm stack, as well as external consistency with other systems and distributions. The ROCm proposed file structure is outlined below:
+
+```none
+/opt/rocm-<ver>
+    | -- bin
+         | -- all public binaries
+    | -- lib
+         | -- lib<soname>.so->lib<soname>.so.major->lib<soname>.so.major.minor.patch
+              (public libaries to link with applications)
+         | -- <component>
+              | -- architecture dependent libraries and binaries used internally by components
+         | -- cmake
+              | -- <component>
+                   | --<component>-config.cmake
+    | -- libexec
+         | -- <component>
+              | -- non ISA/architecture independent executables used internally by components
+    | -- include
+         | -- <component>
+              | -- public header files
+    | -- share
+         | -- html
+              | -- <component>
+                   | -- html documentation
+         | -- info
+              | -- <component>
+                   | -- info files
+         | -- man
+              | -- <component>
+                   | -- man pages
+         | -- doc
+              | -- <component>
+                   | -- license files
+         | -- <component>
+              | -- samples
+              | -- architecture independent misc files
+```
+
+## Changes from earlier ROCm versions
+
+The following table provides a brief overview of the new ROCm FHS layout, compared to the layout of earlier ROCm versions. Note that /opt/ is used to denote the default rocm-installation-path and should be replaced in case of a non-standard installation location of the ROCm distribution.
+
+```none
+ ______________________________________________________
+|  New ROCm Layout            |  Previous ROCm Layout  |
+|_____________________________|________________________|
+| /opt/rocm-<ver>             | /opt/rocm-<ver>        |
+|     | -- bin                |     | -- bin           |
+|     | -- lib                |     | -- lib           |
+|          | -- cmake         |     | -- include       |
+|     | -- libexec            |     | -- <component_1> |
+|     | -- include            |          | -- bin      |
+|          | -- <component_1> |          | -- cmake    |
+|     | -- share              |          | -- doc      |
+|          | -- html          |          | -- lib      |
+|          | -- info          |          | -- include  |
+|          | -- man           |          | -- samples  |
+|          | -- doc           |     | -- <component_n> |
+|          | -- <component_1> |          | -- bin      |
+|               | -- samples  |          | -- cmake    |
+|               | -- ..       |          | -- doc      |
+|          | -- <component_n> |          | -- lib      |
+|               | -- samples  |          | -- include  |
+|               | -- ..       |          | -- samples  |
+|______________________________________________________|
+```
+
+## ROCm FHS reorganization: backward compatibility
+
+The FHS file organization for ROCm was first introduced in the release of ROCm 5.2 . Backward compatibility was implemented to make sure users could still run their ROCm applications while transitioning to the new FHS. ROCm has moved header files and libraries to their new locations as indicated in the above structure, and included symbolic-links and wrapper header files in their old location for backward compatibility. The following sections detail ROCm backward compatibility implementation for wrapper header files, executable files, library files and CMake config files.
+
+### Wrapper header files
+
+Wrapper header files are placed in the old location (
+`/opt/rocm-<ver>/<component>/include`) with a warning message to include files
+from the new location (`/opt/rocm-<ver>/include`) as shown in the example below.
+
+```cpp
+#pragma message "This file is deprecated. Use file from include path /opt/rocm-ver/include/ and prefix with hip."
+#include <hip/hip_runtime.h>
+```
+
+* Starting at ROCm 5.2 release, the deprecation for backward compatibility wrapper header files is: `#pragma` message announcing `#warning`.
+* Starting from ROCm 6.0 (tentatively) backward compatibility for wrapper header files will be removed, and the `#pragma` message will be announcing `#error`.
+
+### Executable files
+
+Executable files are available in the `/opt/rocm-<ver>/bin` folder. For backward
+compatibility, the old library location (`/opt/rocm-<ver>/<component>/bin`) has a
+soft link to the library at the new location. Soft links will be removed in a
+future release, tentatively ROCm v6.0.
+
+```bash
+$ ls -l /opt/rocm/hip/bin/
+lrwxrwxrwx 1 root root   24 Jan 1 23:32 hipcc -> ../../bin/hipcc
+```
+
+### Library files
+
+Library files are available in the `/opt/rocm-<ver>/lib` folder. For backward
+compatibility, the old library location (`/opt/rocm-<ver>/<component>/lib`) has a
+soft link to the library at the new location. Soft links will be removed in a
+future release, tentatively ROCm v6.0.
+
+```shell
+$ ls -l /opt/rocm/hip/lib/
+drwxr-xr-x 4 root root 4096 Jan 1 10:45 cmake
+lrwxrwxrwx 1 root root   24 Jan 1 23:32 libamdhip64.so -> ../../lib/libamdhip64.so
+```
+
+### CMake config files
+
+All CMake configuration files are available in the
+`/opt/rocm-<ver>/lib/cmake/<component>` folder. For backward compatibility, the
+old CMake locations (`/opt/rocm-<ver>/<component>/lib/cmake`) consist of a soft
+link to the new CMake config. Soft links will be removed in a future release,
+tentatively ROCm v6.0.
+
+```shell
+$ ls -l /opt/rocm/hip/lib/cmake/hip/
+lrwxrwxrwx 1 root root 42 Jan 1 23:32 hip-config.cmake -> ../../../../lib/cmake/hip/hip-config.cmake
+```
+
+## Changes required in applications using ROCm
+
+Applications using ROCm are advised to use the new file paths. As the old files
+will be deprecated in a future release. Applications have to make sure to include
+correct header file and use correct search paths.
+
+1. `#include<header_file.h>` needs to be changed to
+   `#include <component/header_file.h>`
+
+   For example: `#include <hip.h>` needs to change
+   to `#include <hip/hip.h>`
+
+2. Any variable in CMake or Makefiles pointing to component folder needs to
+   changed.
+
+   For example: `VAR1=/opt/rocm/hip` needs to be changed to `VAR1=/opt/rocm`
+   `VAR2=/opt/rocm/hsa` needs to be changed to `VAR2=/opt/rocm`
+
+3. Any reference to `/opt/rocm/<component>/bin` or `/opt/rocm/<component>/lib`
+   needs to be changed to `/opt/rocm/bin` and `/opt/rocm/lib/`, respectively.
+
+## Changes in versioning specifications
+
+In order to better manage ROCm dependencies specification and allow smoother releases of ROCm while avoiding dependency conflicts, ROCm software shall adhere to the following scheme when numbering and incrementing ROCm files versions:
+
+rocm-\<ver\>, where \<ver\> = \<x.y.z\>
+
+x.y.z denote: MAJOR.MINOR.PATCH
+
+z: PATCH - increment z when implementing backward compatible bug fixes.
+
+y: MINOR - increment y when implementing minor changes that add functionality but are still backward compatible.
+
+x: MAJOR - increment x when implementing major changes that are not backward compatible.

docs/conceptual/gpu-arch.md

@@ -0,0 +1,72 @@
+<head>
+  <meta charset="UTF-8">
+  <meta name="description" content="GPU architecture">
+  <meta name="keywords" content="GPU architecture, architecture support, MI200, MI250, RDNA,
+  MI100, AMD Instinct">
+</head>
+
+(gpu-arch-documentation)=
+
+# GPU architecture documentation
+
+:::::{grid} 1 1 2 2
+:gutter: 1
+
+:::{grid-item-card}
+**AMD Instinct MI300 series**
+
+Review hardware aspects of the AMD Instinct™ MI300 series of GPU accelerators and the CDNA™ 3
+architecture.
+
+* [AMD Instinct™ MI300 microarchitecture](./gpu-arch/mi300.md)
+* [AMD Instinct MI300/CDNA3 ISA](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)
+* [Performance counters](./gpu-arch/mi300-mi200-performance-counters.rst)
+:::
+
+:::{grid-item-card}
+**AMD Instinct MI200 series**
+
+Review hardware aspects of the AMD Instinct™ MI200 series of GPU accelerators and the CDNA™ 2
+architecture.
+
+* [AMD Instinct™ MI250 microarchitecture](./gpu-arch/mi250.md)
+* [AMD Instinct MI200/CDNA2 ISA](https://www.amd.com/system/files/TechDocs/instinct-mi200-cdna2-instruction-set-architecture.pdf)
+* [White paper](https://www.amd.com/system/files/documents/amd-cdna2-white-paper.pdf)
+* [Performance counters](./gpu-arch/mi300-mi200-performance-counters.rst)
+
+:::
+
+:::{grid-item-card}
+**AMD Instinct MI100**
+
+Review hardware aspects of the AMD Instinct™ MI100 series of GPU accelerators and the CDNA™ 1
+architecture.
+
+* [AMD Instinct™ MI100 microarchitecture](./gpu-arch/mi100.md)
+* [AMD Instinct MI100/CDNA1 ISA](https://www.amd.com/system/files/TechDocs/instinct-mi100-cdna1-shader-instruction-set-architecture%C2%A0.pdf)
+* [White paper](https://www.amd.com/system/files/documents/amd-cdna-whitepaper.pdf)
+
+:::
+
+:::{grid-item-card}
+**RDNA**
+
+* [AMD RDNA3 ISA](https://www.amd.com/system/files/TechDocs/rdna3-shader-instruction-set-architecture-feb-2023_0.pdf)
+* [AMD RDNA2 ISA](https://www.amd.com/system/files/TechDocs/rdna2-shader-instruction-set-architecture.pdf)
+* [AMD RDNA ISA](https://www.amd.com/system/files/TechDocs/rdna-shader-instruction-set-architecture.pdf)
+* [AMD RDNA Architecture White Paper](https://www.amd.com/system/files/documents/rdna-whitepaper.pdf)
+
+:::
+
+:::{grid-item-card}
+**Older architectures**
+
+* [AMD Instinct MI50/Vega 7nm ISA](https://www.amd.com/system/files/TechDocs/vega-7nm-shader-instruction-set-architecture.pdf)
+* [AMD Instinct MI25/Vega ISA](https://www.amd.com/system/files/TechDocs/vega-shader-instruction-set-architecture.pdf)
+* [AMD GCN3 ISA](https://www.amd.com/system/files/TechDocs/gcn3-instruction-set-architecture.pdf)
+* [AMD Vega Architecture White Paper](https://en.wikichip.org/w/images/a/a1/vega-whitepaper.pdf)
+
+:::
+
+:::::

docs/conceptual/gpu-arch/mi100.md

@@ -0,0 +1,94 @@
+<head>
+  <meta charset="UTF-8">
+  <meta name="description" content="AMD Instinct MI100 microarchitecture">
+  <meta name="keywords" content="Instinct, MI100, microarchitecture, AMD, ROCm">
+</head>
+
+# 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™ accelerators.
+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 accelerator](../../data/conceptual/gpu-arch/image004.png "Node-level system architecture with two AMD EPYC™ processors and eight AMD Instinct™ accelerators.")
+
+In a typical node configuration, each processor can host up to four AMD
+Instinct™ accelerators 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 accelerators can participate in a fully connected, coherent AMD
+Instinct™ fabric that connects the four accelerators 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™ accelerators.
+
+## Microarchitecture
+
+The microarchitecture of the AMD Instinct accelerators 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 accelerator (MI100 generation)](../../data/conceptual/gpu-arch/image005.png "Structure of the AMD Instinct accelerator (MI100 generation)")
+
+The above image shows the AMD Instinct accelerator 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 accelerator 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](../../data/conceptual/gpu-arch/image006.png "An MI100 compute unit with detailed SIMD view of the AMD CDNA architecture")
+
+The preceding image shows the block diagram of a single CU of an AMD Instinct™
+MI100 accelerator 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.
+
+:::{table} Peak-performance capabilities of MI100 for different data types.
+:name: mi100-perf
+| 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        |
+
+:::

docs/conceptual/gpu-arch/mi250.md

@@ -0,0 +1,133 @@
+<head>
+  <meta charset="UTF-8">
+  <meta name="description" content="AMD Instinct MI250 microarchitecture">
+  <meta name="keywords" content="Instinct, MI250, microarchitecture, AMD, ROCm">
+</head>
+
+# AMD Instinct™ MI250 microarchitecture
+
+The microarchitecture of the AMD Instinct MI250 accelerators 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 accelerator.](../../data/conceptual/gpu-arch/image001.png "Structure of a single GCD in the AMD Instinct MI250 accelerator.")
+
+```{list-table} Peak-performance capabilities of the MI250 OAM for different data types.
+:header-rows: 1
+:name: mi250-perf-table
+
+*
+  - 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 OCP Open Accelerator Modules (OAM, OCP is short for Open Compute
+Platform) 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 accelerators](../../data/conceptual/gpu-arch/image002.png "Dual-GCD architecture of the AMD Instinct MI250 accelerators")
+
+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 accelerator. 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 Accelerators with 3rd Generation AMD EPYC processor](../../data/conceptual/gpu-arch/image003.png "Block diagram of AMD Instinct MI250 Accelerators with 3rd Generation AMD EPYC processor")
+
+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
+accelerators. 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.

docs/conceptual/gpu-arch/mi300-mi200-performance-counters.rst

@@ -0,0 +1,758 @@
+.. meta::
+  :description: MI300 and MI200 series performance counters and metrics
+  :keywords: MI300, MI200, performance counters, command processor counters
+
+***************************************************************************************************
+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
+:doc:`ROCProfiler tool <rocprofiler:rocprofv1>`.
+
+MI300 and MI200 series performance counters
+===============================================================
+
+Series performance counters include the following categories:
+
+* :ref:`command-processor-counters`
+* :ref:`graphics-register-bus-manager-counters`
+* :ref:`spi-counters`
+* :ref:`compute-unit-counters`
+* :ref:`l1i-and-sl1d-cache-counters`
+* :ref:`vector-l1-cache-subsystem-counters`
+* :ref:`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
+---------------------------------------------------------------------------------------------------------------
+
+Command processor counters are further classified into command processor-fetcher and command
+processor-compute.
+
+Command processor-fetcher counters
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+.. csv-table::
+  :header: "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
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+.. csv-table::
+  :header: "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:
+
+Graphics register bus manager counters
+---------------------------------------------------------------------------------------------------------------
+
+.. csv-table::
+  :header: "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
+
+.. _spi-counters:
+
+Shader processor input counters
+---------------------------------------------------------------------------------------------------------------
+
+.. csv-table::
+  :header: "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 (n\ :sup:`th` pipe) requests but no compute shader (n\ :sup:`th` pipe) allocation"
+  "``SPI_RA_RES_STALL_CSN``", "Cycles", "Number of arbiter stall cycles due to shortage of compute shader (n\ :sup:`th` 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 (n\ :sup:`th` 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 (n\ :sup:`th` 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 (n\ :sup:`th` 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 (n\ :sup:`th` pipe) wave dispatch"
+  "``SPI_RA_BAR_CU_FULL_CSN``", "CU", "Number of compute units with compute shader (n\ :sup:`th` pipe) waves waiting at a BARRIER"
+  "``SPI_RA_BULKY_CU_FULL_CSN``", "CU", "Number of compute units with compute shader (n\ :sup:`th` pipe) waves waiting for BULKY resource"
+  "``SPI_RA_TGLIM_CU_FULL_CSN``", "Cycles", "Number of compute shader (n\ :sup:`th` 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 (n\ :sup:`th` 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:
+
+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
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+.. csv-table::
+  :header: "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
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+.. csv-table::
+  :header: "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.
+
+.. csv-table::
+  :header: "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
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+.. csv-table::
+  :header: "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
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+.. csv-table::
+  :header: "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
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+.. csv-table::
+  :header: "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
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+.. csv-table::
+  :header: "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"
+
+.. _l1i-and-sl1d-cache-counters:
+
+L1 instruction cache (L1i) and scalar L1 data cache (L1d) counters
+---------------------------------------------------------------------------------------------------------------
+
+.. csv-table::
+  :header: "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:
+
+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
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+.. csv-table::
+  :header: "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
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+.. csv-table::
+  :header: "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
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+.. csv-table::
+  :header: "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
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+.. csv-table::
+  :header: "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 access counters
+---------------------------------------------------------------------------------------------------------------
+
+L2 cache is also known as texture cache per channel.
+
+.. tab-set::
+
+    .. tab-item:: MI300 hardware counter
+
+      .. csv-table::
+        :header: "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"
+
+    .. tab-item:: MI200 hardware counter
+
+      .. csv-table::
+        :header: "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
+==============================================================
+
+.. csv-table::
+  :header: "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.
+
+.. csv-table::
+  :header: "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_READ_WAVEFRONTS_sum``", "Total number of flat opcode read 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.
+
+.. csv-table::
+  :header: "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
+---------------------------------------------------------------------------------------------------------------
+
+.. csv-table::
+  :header: "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.
+
+.. csv-table::
+  :header: "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.
+
+.. csv-table::
+  :header: "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.
+
+.. csv-table::
+  :header: "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
+--------------------------------------------------------
+
+.. csv-table::
+  :header: "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"

docs/conceptual/gpu-arch/mi300.md

@@ -0,0 +1,122 @@
+# AMD Instinct™ MI300 series microarchitecture
+
+The AMD Instinct MI300 series accelerators 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 accelerators 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 accelerator series.
+
+```{figure} ../../data/conceptual/gpu-arch/image007.png
+---
+name: mi300-xcd
+align: center
+---
+XCD-level system architecture showing 40 Compute Units, each with 32 KB L1 cache, a Unified Compute System with 4 ACE Compute Accelerators, shared 4MB of L2 cache and an HWS Hardware Scheduler.
+```
+
+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
+accelerators, 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.
+
+```{list-table} Peak-performance capabilities of the MI300X for different data types.
+:header-rows: 1
+:name: mi300x-perf-table
+
+*
+  - 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.
+
+```{figure} ../../data/conceptual/gpu-arch/image008.png
+---
+name: mi300-arch
+alt:
+align: center
+---
+MI300 series system architecture showing MI300A (left) with 6 XCDs and 3 CCDs, while the MI300X (right) has 8 XCDs.
+```
+
+## Node-level architecture
+
+```{figure} ../../data/conceptual/gpu-arch/image009.png
+---
+name: mi300-node
+
+align: center
+---
+MI300 series node-level architecture showing 8 fully interconnected MI300X OAM modules connected to (optional) PCIEe switches via retimers and HGX connectors.
+```
+
+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 accelerators. 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.
+
+<!---
+We need performance data about the P2P communication here.
+-->

docs/conceptual/gpu-isolation.md

@@ -0,0 +1,116 @@
+<head>
+  <meta charset="UTF-8">
+  <meta name="description" content="GPU isolation techniques">
+  <meta name="keywords" content="GPU isolation techniques, UUID, universally unique identifier,
+  environment variables, virtual machines, AMD, ROCm">
+</head>
+
+# 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.
+
+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 {abbr}`UUID (universally unique identifier)`s
+that will be exposed to applications.
+
+Runtime
+: ROCm Software Runtime. Applies to all applications using the user mode ROCm
+  software stack.
+
+```{code-block} shell
+:caption: Example to expose the 1. device and a device based on UUID.
+export ROCR_VISIBLE_DEVICES="0,GPU-DEADBEEFDEADBEEF"
+```
+
+### `GPU_DEVICE_ORDINAL`
+
+Devices indices exposed to OpenCL and HIP applications.
+
+Runtime
+: ROCm Common Language Runtime (`ROCclr`). Applies to applications and runtimes
+  using the `ROCclr` abstraction layer including HIP and OpenCL applications.
+
+```{code-block} shell
+:caption: Example to expose the 1. and 3. device in the system.
+export GPU_DEVICE_ORDINAL="0,2"
+```
+
+(hip_visible_devices)=
+
+### `HIP_VISIBLE_DEVICES`
+
+Device indices exposed to HIP applications.
+
+Runtime: HIP runtime. Applies only to applications using HIP on the AMD platform.
+
+```{code-block} shell
+:caption: Example to expose the 1. and 3. devices in the system.
+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.
+
+```{code-block} shell
+:caption: Example on setting the default device to the third device.
+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
+{ref}`docker-access-gpus-in-container` for details.
+Specifically refer to {ref}`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 officially
+supports [VMware ESXi](https://www.vmware.com/products/esxi-and-esx.html)
+for select GPUs.
+
+<!--
+TODO: This should link to a page about virtualization that explains
+      pass-through and SR-IOV and how-tos for maybe `libvirt` and `VMWare`
+-->

docs/conceptual/gpu-memory.md

@@ -0,0 +1,241 @@
+<head>
+  <meta charset="UTF-8">
+  <meta name="description" content="GPU memory">
+  <meta name="keywords" content="GPU memory, VRAM, video random access memory, pageable
+  memory, pinned memory, managed memory, AMD, ROCm">
+</head>
+
+# GPU memory
+
+For the HIP reference documentation, see:
+
+* {doc}`hip:doxygen/html/group___memory`
+* {doc}`hip:doxygen/html/group___memory_m`
+
+Host memory exists on the host (e.g. CPU) of the machine in random access memory (RAM).
+
+Device memory exists on the device (e.g. GPU) of the machine in video random access memory (VRAM).
+Recent architectures use graphics double data rate (GDDR) synchronous dynamic random-access memory (SDRAM)such as GDDR6, or high-bandwidth memory (HBM) such as HBM2e.
+
+## Memory allocation
+
+Memory can be allocated in two ways: pageable memory, and pinned memory.
+The following API calls with result in these allocations:
+
+| API                | Data location | Allocation |
+|--------------------|---------------|------------|
+| System allocated   | Host          | Pageable   |
+| `hipMallocManaged` | Host          | Managed    |
+| `hipHostMalloc`    | Host          | Pinned     |
+| `hipMalloc`        | Device        | Pinned     |
+
+:::{tip}
+`hipMalloc` and `hipFree` are blocking calls, however, HIP recently added non-blocking versions `hipMallocAsync` and `hipFreeAsync` which take in a stream as an additional argument.
+:::
+
+### Pageable memory
+
+Pageable memory is usually gotten when calling `malloc` or `new` in a C++ application.
+It is unique in that it exists on "pages" (blocks of memory), which can be migrated to other memory storage.
+For example, migrating memory between CPU sockets on a motherboard, or a system that runs out of space in RAM and starts dumping pages of RAM into the swap partition of your hard drive.
+
+### Pinned memory
+
+Pinned memory (or page-locked memory, or non-pageable memory) is host memory that is mapped into the address space of all GPUs, meaning that the pointer can be used on both host and device.
+Accessing host-resident pinned memory in device kernels is generally not recommended for performance, as it can force the data to traverse the host-device interconnect (e.g. PCIe), which is much slower than the on-device bandwidth (>40x on MI200).
+
+Pinned host memory can be allocated with one of two types of coherence support:
+
+:::{note}
+In HIP, pinned memory allocations are coherent by default (`hipHostMallocDefault`).
+There are additional pinned memory flags (e.g. `hipHostMallocMapped` and `hipHostMallocPortable`).
+On MI200 these options do not impact performance.
+<!-- TODO: link to programming_manual#memory-allocation-flags -->
+For more information, see the section *memory allocation flags* in the HIP Programming Guide: {doc}`hip:user_guide/programming_manual`.
+:::
+
+Much like how a process can be locked to a CPU core by setting affinity, a pinned memory allocator does this with the memory storage system.
+On multi-socket systems it is important to ensure that pinned memory is located on the same socket as the owning process, or else each cache line will be moved through the CPU-CPU interconnect, thereby increasing latency and potentially decreasing bandwidth.
+
+In practice, pinned memory is used to improve transfer times between host and device.
+For transfer operations, such as `hipMemcpy` or `hipMemcpyAsync`, using pinned memory instead of pageable memory on host can lead to a ~3x improvement in bandwidth.
+
+:::{tip}
+If the application needs to move data back and forth between device and host (separate allocations), use pinned memory on the host side.
+:::
+
+### Managed memory
+
+Managed memory refers to universally addressable, or unified memory available on the MI200 series of GPUs.
+Much like pinned memory, managed memory shares a pointer between host and device and (by default) supports fine-grained coherence, however, managed memory can also automatically migrate pages between host and device.
+The allocation will be managed by AMD GPU driver using the Linux HMM (Heterogeneous Memory Management) mechanism.
+
+If heterogenous memory management (HMM) is not available, then `hipMallocManaged` will default back to using system memory and will act like pinned host memory.
+Other managed memory API calls will have undefined behavior.
+It is therefore recommended to check for managed memory capability with: `hipDeviceGetAttribute` and `hipDeviceAttributeManagedMemory`.
+
+HIP supports additional calls that work with page migration:
+
+* `hipMemAdvise`
+* `hipMemPrefetchAsync`
+
+:::{tip}
+If the application needs to use data on both host and device regularly, does not want to deal with separate allocations, and is not worried about maxing out the VRAM on MI200 GPUs (64 GB per GCD), use managed memory.
+:::
+
+:::{tip}
+If managed memory performance is poor, check to see if managed memory is supported on your system and if page migration (XNACK) is enabled.
+:::
+
+## Access behavior
+
+Memory allocations for GPUs behave as follow:
+
+| API                | Data location | Host access  | Device access        |
+|--------------------|---------------|--------------|----------------------|
+| System allocated   | Host          | Local access | Unhandled page fault |
+| `hipMallocManaged` | Host          | Local access | Zero-copy            |
+| `hipHostMalloc`    | Host          | Local access | Zero-copy*           |
+| `hipMalloc`        | Device        | Zero-copy    | Local access         |
+
+Zero-copy accesses happen over the Infinity Fabric interconnect or PCI-E lanes on discrete GPUs.
+
+:::{note}
+While `hipHostMalloc` allocated memory is accessible by a device, the host pointer must be converted to a device pointer with `hipHostGetDevicePointer`.
+
+Memory allocated through standard system allocators such as `malloc`, can be accessed a device by registering the memory via `hipHostRegister`.
+The device pointer to be used in kernels can be retrieved with `hipHostGetDevicePointer`.
+Registered memory is treated like `hipHostMalloc` and will have similar performance.
+
+On devices that support and have [](#xnack) enabled, such as the MI250X, `hipHostRegister` is not required as memory accesses are handled via automatic page migration.
+:::
+
+### XNACK
+
+Normally, host and device memory are separate and data has to be transferred manually via `hipMemcpy`.
+
+On a subset of GPUs, such as the MI200, there is an option to automatically migrate pages of memory between host and device.
+This is important for managed memory, where the locality of the data is important for performance.
+Depending on the system, page migration may be disabled by default in which case managed memory will act like pinned host memory and suffer degraded performance.
+
+*XNACK* describes the GPUs ability to retry memory accesses that failed due a page fault (which normally would lead to a memory access error), and instead retrieve the missing page.
+
+This also affects memory allocated by the system as indicated by the following table:
+
+| API                | Data location | Host after device access | Device after host access |
+|--------------------|---------------|--------------------------|--------------------------|
+| System allocated   | Host          | Migrate page to host     | Migrate page to device   |
+| `hipMallocManaged` | Host          | Migrate page to host     | Migrate page to device   |
+| `hipHostMalloc`    | Host          | Local access             | Zero-copy                |
+| `hipMalloc`        | Device        | Zero-copy                | Local access             |
+
+To check if page migration is available on a platform, use `rocminfo`:
+
+```sh
+$ rocminfo | grep xnack
+      Name:                    amdgcn-amd-amdhsa--gfx90a:sramecc+:xnack-
+```
+
+Here, `xnack-` means that XNACK is available but is disabled by default.
+Turning on XNACK by setting the environment variable `HSA_XNACK=1` and gives the expected result, `xnack+`:
+
+```sh
+$ HSA_XNACK=1 rocminfo | grep xnack
+Name:                    amdgcn-amd-amdhsa--gfx90a:sramecc+:xnack+
+```
+
+`hipcc`by default will generate code that runs correctly with both XNACK enabled or disabled.
+Setting the `--offload-arch=`-option with `xnack+` or `xnack-` forces code to be only run with XNACK enabled or disabled respectively.
+
+```sh
+# Compiled kernels will run regardless if XNACK is enabled or is disabled. 
+hipcc --offload-arch=gfx90a
+
+# Compiled kernels will only be run if XNACK is enabled with XNACK=1.
+hipcc --offload-arch=gfx90a:xnack+
+
+# Compiled kernels will only be run if XNACK is disabled with XNACK=0.
+hipcc --offload-arch=gfx90a:xnack-
+```
+
+:::{tip}
+If you want to make use of page migration, use managed memory. While pageable memory will migrate correctly, it is not a portable solution and can have performance issues if the accessed data isn't page aligned.
+:::
+
+### Coherence
+
+* *Coarse-grained coherence* means that memory is only considered up to date at kernel boundaries, which can be enforced through `hipDeviceSynchronize`, `hipStreamSynchronize`, or any blocking operation that acts on the null stream (e.g. `hipMemcpy`).
+For example, cacheable memory is a type of coarse-grained memory where an up-to-date copy of the data can be stored elsewhere (e.g. in an L2 cache).
+* *Fine-grained coherence* means the coherence is supported while a CPU/GPU kernel is running.
+This can be useful if both host and device are operating on the same dataspace using system-scope atomic operations (e.g. updating an error code or flag to a buffer).
+Fine-grained memory implies that up-to-date data may be made visible to others regardless of kernel boundaries as discussed above.
+
+| API                     | Flag                         | Coherence      |
+|-------------------------|------------------------------|----------------|
+| `hipHostMalloc`         | `hipHostMallocDefault`       | Fine-grained   |
+| `hipHostMalloc`         | `hipHostMallocNonCoherent`   | Coarse-grained |
+
+| API                     | Flag                         | Coherence      |
+|-------------------------|------------------------------|----------------|
+| `hipExtMallocWithFlags` | `hipDeviceMallocDefault`     | Coarse-grained |
+| `hipExtMallocWithFlags` | `hipDeviceMallocFinegrained` | Fine-grained   |
+
+| API                     | `hipMemAdvise` argument      | Coherence      |
+|-------------------------|------------------------------|----------------|
+| `hipMallocManaged`      |                              | Fine-grained   |
+| `hipMallocManaged`      | `hipMemAdviseSetCoarseGrain` | Coarse-grained |
+| `malloc`                |                              | Fine-grained   |
+| `malloc`                | `hipMemAdviseSetCoarseGrain` | Coarse-grained |
+
+:::{tip}
+Try to design your algorithms to avoid host-device memory coherence (e.g. system scope atomics). While it can be a useful feature in very specific cases, it is not supported on all systems, and can negatively impact performance by introducing the host-device interconnect bottleneck.
+:::
+
+The availability of fine- and coarse-grained memory pools can be checked with `rocminfo`:
+
+```sh
+$ rocminfo
+...
+*******
+Agent 1
+*******
+Name:                    AMD EPYC 7742 64-Core Processor
+...
+Pool Info:
+Pool 1
+Segment:                 GLOBAL; FLAGS: FINE GRAINED
+...
+Pool 3
+Segment:                 GLOBAL; FLAGS: COARSE GRAINED
+...
+*******
+Agent 9
+*******
+Name:                    gfx90a
+...
+Pool Info:
+Pool 1
+Segment:                 GLOBAL; FLAGS: COARSE GRAINED
+...
+```
+
+## System direct memory access
+
+In most cases, the default behavior for HIP in transferring data from a pinned host allocation to device will run at the limit of the interconnect.
+However, there are certain cases where the interconnect is not the bottleneck.
+
+The primary way to transfer data onto and off of a GPU, such as the MI200, is to use the onboard System Direct Memory Access engine, which is used to feed blocks of memory to the off-device interconnect (either GPU-CPU or GPU-GPU).
+Each GCD has a separate SDMA engine for host-to-device and device-to-host memory transfers.
+Importantly, SDMA engines are separate from the computing infrastructure, meaning that memory transfers to and from a device will not impact kernel compute performance, though they do impact memory bandwidth to a limited extent.
+The SDMA engines are mainly tuned for PCIe-4.0 x16, which means they are designed to operate at bandwidths up to 32 GB/s.
+
+:::{note}
+An important feature of the MI250X platform is the Infinity Fabric™ interconnect between host and device.
+The Infinity Fabric interconnect supports improved performance over standard PCIe-4.0 (usually ~50% more bandwidth); however, since the SDMA engine does not run at this speed, it will not max out the bandwidth of the faster interconnect.
+:::
+
+The bandwidth limitation can be countered by bypassing the SDMA engine and replacing it with a type of copy kernel known as a "blit" kernel.
+Blit kernels will use the compute units on the GPU, thereby consuming compute resources, which may not always be beneficial.
+The easiest way to enable blit kernels is to set an environment variable `HSA_ENABLE_SDMA=0`, which will disable the SDMA engine.
+On systems where the GPU uses a PCIe interconnect instead of an Infinity Fabric interconnect, blit kernels will not impact bandwidth, but will still consume compute resources.
+The use of SDMA vs blit kernels also applies to MPI data transfers and GPU-GPU transfers.

docs/conceptual/using-gpu-sanitizer.md

@@ -0,0 +1,433 @@
+<head>
+  <meta charset="UTF-8">
+  <meta name="description" content="Using the LLVM ASan on a GPU">
+  <meta name="keywords" content="LLVM, ASan, address sanitizer, AddressSanitizer, instrumented
+  libraries, instrumented applications, AMD, ROCm">
+</head>
+
+# Using the AddressSanitizer on a GPU (beta release)
+
+The LLVM AddressSanitizer (ASan) provides a process that allows developers to detect runtime addressing errors in applications and libraries. The detection is achieved using a combination of compiler-added instrumentation and runtime techniques, including function interception and replacement.
+
+Until now, the LLVM ASan process was only available for traditional purely CPU applications. However, ROCm has extended this mechanism to additionally allow the detection of some addressing errors on the GPU in heterogeneous applications. Ideally, developers should treat heterogeneous HIP and OpenMP applications exactly like pure CPU applications. However, this simplicity has not been achieved yet.
+This document provides documentation on using ROCm ASan.
+
+For information about LLVM ASan, see the [LLVM documentation](https://clang.llvm.org/docs/AddressSanitizer.html).
+
+:::{note}
+The beta release of LLVM ASan for ROCm is currently tested and validated on Ubuntu 20.04.
+:::
+
+## Compiling for ASan
+
+The ASan process begins by compiling the application of interest with the ASan instrumentation.
+
+Recommendations for doing this are:
+
+* Compile as many application and dependent library sources as possible using an AMD-built clang-based compiler such as `amdclang++`.
+* Add the following options to the existing compiler and linker options:
+  
+  * `-fsanitize=address` - enables instrumentation
+
+  * `-shared-libsan` - use shared version of runtime
+
+  * `-g` - add debug info for improved reporting
+
+* Explicitly use `xnack+` in the offload architecture option. For example, `--offload-arch=gfx90a:xnack+`
+
+Other architectures are allowed, but their device code will not be instrumented, and a warning will be issued.
+
+:::{tip}
+It is not an error to compile some files without ASan instrumentation, but doing so reduces the ability of the process to detect addressing errors. However, if the main program "`a.out`" does not directly depend on the ASan runtime (`libclang_rt.asan-x86_64.so`) after the build completes (check by running `ldd` (List Dynamic Dependencies) or `readelf`), the application will immediately report an error at runtime as described in the next section.
+:::
+
+:::{note}
+When compiling OpenMP programs with ASan instrumentation, it is currently necessary to set the environment variable `LIBRARY_PATH` to `/opt/rocm-<version>/lib/llvm/lib/asan:/opt/rocm-<version>/lib/asan`. At runtime, it may be necessary to add `/opt/rocm-<version>/lib/llvm/lib/asan` to `LD_LIBRARY_PATH`.
+:::
+
+### About compilation time
+
+When `-fsanitize=address` is used, the LLVM compiler adds instrumentation code around every memory operation. This added code must be handled by all downstream components of the compiler toolchain, and results in increased overall compilation time. This increase is especially evident in the AMDGPU device compiler and has in a few instances increased compile time to an unacceptable level.
+
+There are a few options if the compile time becomes unacceptable:
+
+* Avoid instrumentation of the files which have the worst compile times. This will reduce the effectiveness of the ASan process.
+* Add the option `-fsanitize-recover=address` to the compiles with the worst compile times. This option simplifies the added instrumentation resulting in faster compilation. See below for more information.
+* Disable instrumentation on a per-function basis by adding `__attribute__`((no_sanitize("address"))) to functions found to be responsible for the large compile time. Again, this will reduce the effectiveness of the process.
+
+## Installing ROCm GPU ASan packages
+
+For a complete ROCm GPU Sanitizer installation, including packages, instrumented HSA and HIP runtimes, tools, and math libraries, use the following instruction,
+
+```bash
+    sudo apt-get install rocm-ml-sdk-asan
+
+```
+
+## Using AMD-supplied ASan instrumented libraries
+
+ROCm releases have optional packages that contain additional ASan instrumented builds of the ROCm libraries (usually found in `/opt/rocm-<version>/lib`). The instrumented libraries have identical names to the regular uninstrumented libraries, and are located in `/opt/rocm-<version>/lib/asan`.
+These additional libraries are built using the `amdclang++` and `hipcc` compilers, while some uninstrumented libraries are built with `g++`. The preexisting build options are used but, as described above, additional options are used: `-fsanitize=address`, `-shared-libsan` and `-g`.
+
+These instrumented libraries avoid additional developer effort to locate repositories, identify the correct branch, check out the correct tags, and other efforts needed to build the libraries from the source. And they extend the ability of the process to detect addressing errors into the ROCm libraries themselves.
+
+When adjusting an application build to add instrumentation, linking against these instrumented libraries is unnecessary. For example, any `-L` `/opt/rocm-<version>/lib` compiler options need not be changed. However, the instrumented libraries should be used when the application is run. It is particularly important that the instrumented language runtimes, like `libamdhip64.so` and `librocm-core.so`, are used; otherwise, device invalid access detections may not be reported.
+
+## Running ASan instrumented applications
+
+### Preparing to run an instrumented application
+
+Here are a few recommendations to consider before running an ASan instrumented heterogeneous application.
+
+* Ensure the Linux kernel running on the system has Heterogeneous Memory Management (HMM) support. A kernel version of 5.6 or higher should be sufficient.
+* Ensure XNACK is enabled
+  * For `gfx90a` (MI-2X0) or `gfx940` (MI-3X0) use environment `HSA_XNACK = 1`.
+  * For `gfx906` (MI-50) or `gfx908` (MI-100) use environment `HSA_XNACK = 1` but also ensure the amdgpu kernel module is loaded with module argument `noretry=0`.
+This requirement is due to the fact that the XNACK setting for these GPUs is system-wide.
+
+* Ensure that the application will use the instrumented libraries when it runs. The output from the shell command `ldd <application name>` can be used to see which libraries will be used.
+If the instrumented libraries are not listed by `ldd`, the environment variable `LD_LIBRARY_PATH` may need to be adjusted, or in some cases an `RPATH` compiled into the application may need to be changed and the application recompiled.
+
+* Ensure that the application depends on the ASan runtime. This can be checked by running the command `readelf -d <application name> | grep NEEDED` and verifying that shared library: `libclang_rt.asan-x86_64.so` appears in the output.
+If it does not appear, when executed the application will quickly output an ASan error that looks like:
+
+```bash
+==3210==ASan runtime does not come first in initial library list; you should either link runtime to your application or manually preload it with LD_PRELOAD.
+```
+
+* Ensure that the application `llvm-symbolizer` can be executed, and that it is located in `/opt/rocm-<version>/llvm/bin`. This executable is not strictly required, but if found is used to translate ("symbolize") a host-side instruction address into a more useful function name, file name, and line number (assuming the application has been built to include debug information).
+
+There is an environment variable, `ASAN_OPTIONS`, that can be used to adjust the runtime behavior of the ASan runtime itself. There are more than a hundred "flags" that can be adjusted (see an old list at [flags](https://github.com/google/sanitizers/wiki/AddressSanitizerFlags)) but the default settings are correct and should be used in most cases. It must be noted that these options only affect the host ASan runtime. The device runtime only currently supports the default settings for the few relevant options.
+
+There are three `ASAN_OPTION` flags of note.
+
+* `halt_on_error=0/1 default 1`.
+
+  This tells the ASan runtime to halt the application immediately after detecting and reporting an addressing error. The default makes sense because the application has entered the realm of undefined behavior. If the developer wishes to have the application continue anyway, this option can be set to zero. However, the application and libraries should then be compiled with the additional option `-fsanitize-recover=address`. Note that the ROCm optional ASan instrumented libraries are not compiled with this option and if an error is detected within one of them, but halt_on_error is set to 0, more undefined behavior will occur.
+
+* `detect_leaks=0/1 default 1`.
+
+  This option directs the ASan runtime to enable the [Leak Sanitizer](https://clang.llvm.org/docs/LeakSanitizer.html) (LSan). For heterogeneous applications, this default results in significant output from the leak sanitizer when the application exits due to allocations made by the language runtime which are not considered to be leaks. This output can be avoided by adding `detect_leaks=0` to the `ASAN_OPTIONS`, or alternatively by producing an LSan suppression file (syntax described [here](https://github.com/google/sanitizers/wiki/AddressSanitizerLeakSanitizer)) and activating it with environment variable `LSAN_OPTIONS=suppressions=/path/to/suppression/file`. When using a suppression file, a suppression report is printed by default. The suppression report can be disabled by using the `LSAN_OPTIONS` flag `print_suppressions=0`.
+
+* `quarantine_size_mb=N default 256`
+
+  This option defines the number of megabytes (MB) `N` of memory that the ASan runtime will hold after it is `freed` to detect use-after-free situations. This memory is unavailable for other purposes. The default of 256 MB may be too small to detect some use-after-free situations, especially given that the large size of many GPU memory allocations may push `freed` allocations out of quarantine before the attempted use.
+
+  :::{note}
+  Setting the value of `quarantine_size_mb` larger may enable more problematic uses to be detected, but at the cost of reducing memory available for other purposes.
+  :::
+
+## Runtime overhead
+
+Running an ASan instrumented application incurs
+overheads which may result in unacceptably long runtimes
+or failure to run at all.
+
+### Higher execution time
+
+ASan detection works by checking each address at runtime
+before the address is actually accessed by a load, store, or atomic
+instruction.
+This checking involves an additional load to "shadow" memory which
+records whether the address is "poisoned" or not, and additional logic
+that decides whether to produce a detection report or not.
+
+This extra runtime work can cause the application to slow down by
+a factor of three or more, depending on how many memory accesses are
+executed.
+
+For heterogeneous applications, the shadow memory must be accessible by all devices
+and this can mean that shadow accesses from some devices may be more costly
+than non-shadow accesses.
+
+### Higher memory use
+
+The address checking described above relies on the compiler to surround
+each program variable with a red zone and on ASan
+runtime to surround each runtime memory allocation with a red zone and
+fill the shadow corresponding to each red zone with poison.
+The added memory for the red zones is additional overhead on top
+of the 13% overhead for the shadow memory itself.
+
+Applications which consume most one or more available memory pools when
+run normally are likely to encounter allocation failures when run with
+instrumentation.
+
+## Runtime reporting
+
+It is not the intention of this document to provide a detailed explanation of all types of reports that can be output by the ASan runtime. Instead, the focus is on the differences between the standard reports for CPU issues, and reports for GPU issues.
+
+An invalid address detection report for the CPU always starts with
+
+```bash
+==<PID>==ERROR: AddressSanitizer: <problem type> on address <memory address> at pc <pc> bp <bp> sp <sp> <access> of size <N> at <memory address> thread T0
+```
+
+and continues with a stack trace for the access, a stack trace for the allocation and deallocation, if relevant, and a dump of the shadow near the <memory address>.
+
+In contrast, an invalid address detection report for the GPU always starts with
+
+```bash
+==<PID>==ERROR: AddressSanitizer: <problem type> on amdgpu device <device> at pc <pc> <access> of size <n> in workgroup id (<X>,<Y>,<Z>)
+```
+
+Above, `<device>` is the integer device ID, and `(<X>, <Y>, <Z>)` is the ID of the workgroup or block where the invalid address was detected.
+
+While the CPU report include a call stack for the thread attempting the invalid access, the GPU is currently to a call stack of size one, i.e. the (symbolized) of the invalid access, e.g.
+
+```bash
+#0 <pc> in <fuction signature> at /path/to/file.hip:<line>:<column>
+```
+
+This short call stack is followed by a GPU unique section that looks like
+
+```bash
+Thread ids and accessed addresses:
+<lid0> <maddr 0> : <lid1> <maddr1> : ...
+```
+
+where each `<lid j> <maddr j>` indicates the lane ID and the invalid memory address held by lane `j` of the wavefront attempting the invalid access.
+
+Additionally, reports for invalid GPU accesses to memory allocated by GPU code via `malloc` or new starting with, for example,
+
+```bash
+==1234==ERROR: AddressSanitizer: heap-buffer-overflow on amdgpu device 0 at pc 0x7fa9f5c92dcc
+```
+
+or
+
+```bash
+==5678==ERROR: AddressSanitizer: heap-use-after-free on amdgpu device 3 at pc 0x7f4c10062d74
+```
+
+currently may include one or two surprising CPU side tracebacks mentioning :`hostcall`". This is due to how `malloc` and `free` are implemented for GPU code and these call stacks can be ignored.
+
+## Running ASan with `rocgdb`
+
+`rocgdb` can be used to further investigate ASan detected errors, with some preparation.
+
+Currently, the ASan runtime complains when starting `rocgdb` without preparation.
+
+```bash
+$ rocgdb my_app
+==1122==ASan` runtime does not come first in initial library list; you should either link runtime to your application or manually preload it with LD_PRELOAD.
+```
+
+This is solved by setting environment variable `LD_PRELOAD` to the path to the ASan runtime, whose path can be obtained using the command
+
+```bash
+amdclang++ -print-file-name=libclang_rt.asan-x86_64.so
+```
+
+You should also set the environment variable `HIP_ENABLE_DEFERRED_LOADING=0` before debugging HIP applications.
+
+After starting `rocgdb` breakpoints can be set on the ASan runtime error reporting entry points of interest. For example, if an ASan error report includes
+
+```bash
+WRITE of size 4 in workgroup id (10,0,0)
+```
+
+the `rocgdb` command needed to stop the program before the report is printed is
+
+```bash
+(gdb) break __asan_report_store4
+```
+
+Similarly, the appropriate command for a report including
+
+```bash
+READ of size <N> in workgroup ID (1,2,3)
+```
+
+is
+
+```bash
+(gdb) break __asan_report_load<N>
+```
+
+It is possible to set breakpoints on all ASan report functions using these commands:
+
+```bash
+$ rocgdb <path to application>
+(gdb) start <commmand line arguments>
+(gdb) rbreak ^__asan_report
+(gdb) c
+```
+
+## Using ASan with a short HIP application
+
+Consider the following simple and short demo of using the Address Sanitizer with a HIP application:
+
+```C++
+
+#include <cstdlib>
+#include <hip/hip_runtime.h>
+
+__global__ void
+set1(int *p)
+{
+    int i = blockDim.x*blockIdx.x + threadIdx.x;
+    p[i] = 1;
+}
+
+int
+main(int argc, char **argv)
+{
+    int m = std::atoi(argv[1]);
+    int n1 = std::atoi(argv[2]);
+    int n2 = std::atoi(argv[3]);
+    int c = std::atoi(argv[4]);
+    int *dp;
+    hipMalloc(&dp, m*sizeof(int));
+    hipLaunchKernelGGL(set1, dim3(n1), dim3(n2), 0, 0, dp);
+    int *hp = (int*)malloc(c * sizeof(int));
+    hipMemcpy(hp, dp, m*sizeof(int), hipMemcpyDeviceToHost);
+    hipDeviceSynchronize();
+    hipFree(dp);
+    free(hp);
+    std::puts("Done.");
+    return 0;
+}
+```
+
+This application will attempt to access invalid addresses for certain command line arguments. In particular, if `m < n1 * n2` some device threads will attempt to access
+unallocated device memory.
+
+Or, if `c < m`, the `hipMemcpy` function will copy past the end of the `malloc` allocated memory.
+
+**Note**: The `hipcc` compiler is used here for simplicity.
+
+Compiling without XNACK results in a warning.
+
+```bash
+$ hipcc -g --offload-arch=gfx90a:xnack- -fsanitize=address -shared-libsan mini.hip -o mini
+clang++: warning: ignoring` `-fsanitize=address' option for offload arch 'gfx90a:xnack-`, as it is not currently supported there. Use it with an offload arch containing 'xnack+' instead [-Woption-ignored]`.
+```
+
+The binary compiled above will run, but the GPU code will not be instrumented and the `m < n1 * n2` error will not be detected. Switching to `--offload-arch=gfx90a:xnack+` in the command above results in a warning-free compilation and an instrumented application. After setting `PATH`, `LD_LIBRARY_PATH` and `HSA_XNACK` as described earlier, a check of the binary with `ldd` yields the following,
+
+```bash
+$ ldd mini
+        linux-vdso.so.1 (0x00007ffd1a5ae000)
+        libclang_rt.asan-x86_64.so => /opt/rocm-6.1.0-99999/llvm/lib/clang/17.0.0/lib/linux/libclang_rt.asan-x86_64.so (0x00007fb9c14b6000)
+        libamdhip64.so.5 => /opt/rocm-6.1.0-99999/lib/asan/libamdhip64.so.5 (0x00007fb9bedd3000)
+        libstdc++.so.6 => /lib/x86_64-linux-gnu/libstdc++.so.6 (0x00007fb9beba8000)
+        libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6 (0x00007fb9bea59000)
+        libgcc_s.so.1 => /lib/x86_64-linux-gnu/libgcc_s.so.1 (0x00007fb9bea3e000)
+        libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007fb9be84a000)
+        libdl.so.2 => /lib/x86_64-linux-gnu/libdl.so.2 (0x00007fb9be844000)
+        libpthread.so.0 => /lib/x86_64-linux-gnu/libpthread.so.0 (0x00007fb9be821000)
+        librt.so.1 => /lib/x86_64-linux-gnu/librt.so.1 (0x00007fb9be817000)
+        libamd_comgr.so.2 => /opt/rocm-6.1.0-99999/lib/asan/libamd_comgr.so.2 (0x00007fb9b4382000)
+        libhsa-runtime64.so.1 => /opt/rocm-6.1.0-99999/lib/asan/libhsa-runtime64.so.1 (0x00007fb9b3b00000)
+        libnuma.so.1 => /lib/x86_64-linux-gnu/libnuma.so.1 (0x00007fb9b3af3000)
+        /lib64/ld-linux-x86-64.so.2 (0x00007fb9c2027000)
+        libz.so.1 => /lib/x86_64-linux-gnu/libz.so.1 (0x00007fb9b3ad7000)
+        libtinfo.so.6 => /lib/x86_64-linux-gnu/libtinfo.so.6 (0x00007fb9b3aa7000)
+        libelf.so.1 => /lib/x86_64-linux-gnu/libelf.so.1 (0x00007fb9b3a89000)
+        libdrm.so.2 => /opt/amdgpu/lib/x86_64-linux-gnu/libdrm.so.2 (0x00007fb9b3a70000)
+        libdrm_amdgpu.so.1 => /opt/amdgpu/lib/x86_64-linux-gnu/libdrm_amdgpu.so.1 (0x00007fb9b3a62000)
+
+```
+
+This confirms that the address sanitizer runtime is linked in, and the ASan instrumented version of the runtime libraries are used.
+Checking the `PATH` yields
+
+```bash
+$ which llvm-symbolizer
+/opt/rocm-6.1.0-99999/llvm/bin/llvm-symbolizer
+```
+
+Lastly, a check of the OS kernel version yields
+
+```bash
+$ uname -rv
+5.15.0-73-generic #80~20.04.1-Ubuntu SMP Wed May 17 14:58:14 UTC 2023
+```
+
+which indicates that the required HMM support (kernel version > 5.6) is available. This completes the necessary setup. Running with `m = 100`, `n1 = 11`, `n2 = 10` and `c = 100` should produce
+a report for an invalid access by the last 10 threads.
+
+```bash
+=================================================================
+==3141==ERROR: AddressSanitizer: heap-buffer-overflow on amdgpu device 0 at pc 0x7fb1410d2cc4
+WRITE of size 4 in workgroup id (10,0,0)
+  #0 0x7fb1410d2cc4 in set1(int*) at /home/dave/mini/mini.cpp:0:10
+
+Thread ids and accessed addresses:
+00 : 0x7fb14371d190 01 : 0x7fb14371d194 02 : 0x7fb14371d198 03 : 0x7fb14371d19c 04 : 0x7fb14371d1a0 05 : 0x7fb14371d1a4 06 : 0x7fb14371d1a8 07 : 0x7fb14371d1ac
+08 : 0x7fb14371d1b0 09 : 0x7fb14371d1b4
+
+0x7fb14371d190 is located 0 bytes after 400-byte region [0x7fb14371d000,0x7fb14371d190)
+allocated by thread T0 here:
+    #0 0x7fb151c76828 in hsa_amd_memory_pool_allocate /work/dave/git/compute/external/llvm-project/compiler-rt/lib/asan/asan_interceptors.cpp:692:3
+    #1 ...
+
+    #12 0x7fb14fb99ec4 in hipMalloc /work/dave/git/compute/external/clr/hipamd/src/hip_memory.cpp:568:3
+    #13 0x226630 in hipError_t hipMalloc<int>(int**, unsigned long) /opt/rocm-6.1.0-99999/include/hip/hip_runtime_api.h:8367:12
+    #14 0x226630 in main /home/dave/mini/mini.cpp:19:5
+    #15 0x7fb14ef02082 in __libc_start_main /build/glibc-SzIz7B/glibc-2.31/csu/../csu/libc-start.c:308:16
+
+Shadow bytes around the buggy address:
+  0x7fb14371cf00: ...
+
+=>0x7fb14371d180: 00 00[fa]fa fa fa fa fa fa fa fa fa fa fa fa fa
+  0x7fb14371d200: ...
+
+Shadow byte legend (one shadow byte represents 8 application bytes):
+  Addressable:           00
+  Partially addressable: 01 02 03 04 05 06 07
+  Heap left redzone:       fa
+  ...
+==3141==ABORTING
+```
+
+Running with `m = 100`, `n1 = 10`, `n2 = 10` and `c = 99` should produce a report for an invalid copy.
+
+```shell
+=================================================================
+==2817==ERROR: AddressSanitizer: heap-buffer-overflow on address 0x514000150dcc at pc 0x7f5509551aca bp 0x7ffc90a7ae50 sp 0x7ffc90a7a610
+WRITE of size 400 at 0x514000150dcc thread T0
+    #0 0x7f5509551ac9 in __asan_memcpy /work/dave/git/compute/external/llvm-project/compiler-rt/lib/asan/asan_interceptors_memintrinsics.cpp:61:3
+    #1 ...
+
+    #9 0x7f5507462a28 in hipMemcpy_common(void*, void const*, unsigned long, hipMemcpyKind, ihipStream_t*) /work/dave/git/compute/external/clr/hipamd/src/hip_memory.cpp:637:10
+    #10 0x7f5507464205 in hipMemcpy /work/dave/git/compute/external/clr/hipamd/src/hip_memory.cpp:642:3
+    #11 0x226844 in main /home/dave/mini/mini.cpp:22:5
+    #12 0x7f55067c3082 in __libc_start_main /build/glibc-SzIz7B/glibc-2.31/csu/../csu/libc-start.c:308:16
+    #13 0x22605d in _start (/home/dave/mini/mini+0x22605d)
+
+0x514000150dcc is located 0 bytes after 396-byte region [0x514000150c40,0x514000150dcc)
+allocated by thread T0 here:
+    #0 0x7f5509553dcf in malloc /work/dave/git/compute/external/llvm-project/compiler-rt/lib/asan/asan_malloc_linux.cpp:69:3
+    #1 0x226817 in main /home/dave/mini/mini.cpp:21:21
+    #2 0x7f55067c3082 in __libc_start_main /build/glibc-SzIz7B/glibc-2.31/csu/../csu/libc-start.c:308:16
+
+SUMMARY: AddressSanitizer: heap-buffer-overflow /work/dave/git/compute/external/llvm-project/compiler-rt/lib/asan/asan_interceptors_memintrinsics.cpp:61:3 in __asan_memcpy
+Shadow bytes around the buggy address:
+  0x514000150b00: ...
+
+=>0x514000150d80: 00 00 00 00 00 00 00 00 00[04]fa fa fa fa fa fa
+  0x514000150e00: ...
+
+Shadow byte legend (one shadow byte represents 8 application bytes):
+  Addressable:           00
+  Partially addressable: 01 02 03 04 05 06 07
+  Heap left redzone:       fa
+  ...
+==2817==ABORTING
+```
+
+## Known issues with using GPU sanitizer
+
+* Red zones must have limited size. It is possible for an invalid access to completely miss a red zone and not be detected.
+
+* Lack of detection or false reports can be caused by the runtime not properly maintaining red zone shadows.
+
+* Lack of detection on the GPU might also be due to the implementation not instrumenting accesses to all GPU specific address spaces. For example, in the current implementation accesses to "private" or "stack" variables on the GPU are not instrumented, and accesses to HIP shared variables (also known as "local data store" or "LDS") are also not instrumented.
+
+* It can also be the case that a memory fault is reported for an invalid address even with the instrumentation. This is usually caused by the invalid address being so wild that its shadow address is outside any memory region, and the fault actually occurs on the access to the shadow address. It is also possible to hit a memory fault for the `NULL` pointer. While address 0 does have a shadow location, it is not poisoned by the runtime.
+
+* There is currently a bug which can result in memory faults being reported when running instrumented device code which makes use of `malloc`, `free`, `new`, or `delete`.
+
+* There is currently a bug which can result in undefined symbols being reported at compile time when instrumented device code makes use of `new` and `delete`.

docs/conf.py

@@ -0,0 +1,130 @@
+# Configuration file for the Sphinx documentation builder.
+#
+# This file only contains a selection of the most common options. For a full
+# list see the documentation:
+# https://www.sphinx-doc.org/en/master/usage/configuration.html
+
+import shutil
+import jinja2
+import os
+
+# Environment to process Jinja templates.
+jinja_env = jinja2.Environment(loader=jinja2.FileSystemLoader("."))
+
+# Jinja templates to render out.
+templates = []
+
+# Render templates and output files without the last extension.
+# For example: 'install.md.jinja' becomes 'install.md'.
+for template in templates:
+    rendered = jinja_env.get_template(template).render()
+    with open(os.path.splitext(template)[0], 'w') as file:
+        file.write(rendered)
+
+shutil.copy2('../RELEASE.md','./about/release-notes.md')
+# Keep capitalization due to similar linking on GitHub's markdown preview.
+shutil.copy2('../CHANGELOG.md','./about/CHANGELOG.md')
+
+latex_engine = "xelatex"
+latex_elements = {
+    "fontpkg": r"""
+\usepackage{tgtermes}
+\usepackage{tgheros}
+\renewcommand\ttdefault{txtt}
+"""
+}
+
+# configurations for PDF output by Read the Docs
+project = "ROCm Documentation"
+author = "Advanced Micro Devices, Inc."
+copyright = "Copyright (c) 2024 Advanced Micro Devices, Inc. All rights reserved."
+version = "6.1.0"
+release = "6.1.0"
+setting_all_article_info = True
+all_article_info_os = ["linux", "windows"]
+all_article_info_author = ""
+
+# pages with specific settings
+article_pages = [
+    {
+        "file":"about/release-notes",
+        "os":["linux", "windows"],
+        "date":"2024-04-16"
+    },
+    {
+        "file":"about/CHANGELOG",
+        "os":["linux", "windows"],
+        "date":"2024-04-16"
+    },
+
+    {"file":"install/windows/install-quick", "os":["windows"]},
+    {"file":"install/linux/install-quick", "os":["linux"]},
+
+    {"file":"install/linux/install", "os":["linux"]},
+    {"file":"install/linux/install-options", "os":["linux"]},
+    {"file":"install/linux/prerequisites", "os":["linux"]},
+
+    {"file":"install/docker", "os":["linux"]},
+    {"file":"install/magma-install", "os":["linux"]},
+    {"file":"install/pytorch-install", "os":["linux"]},
+    {"file":"install/tensorflow-install", "os":["linux"]},
+
+    {"file":"install/windows/install", "os":["windows"]},
+    {"file":"install/windows/prerequisites", "os":["windows"]},
+    {"file":"install/windows/cli/index", "os":["windows"]},
+    {"file":"install/windows/gui/index", "os":["windows"]},
+
+    {"file":"about/compatibility/docker-image-support-matrix", "os":["linux"]},
+    {"file":"about/compatibility/user-kernel-space-compat-matrix", "os":["linux"]},
+
+    {"file":"reference/library-index", "os":["linux"]},
+
+    {"file":"how-to/deep-learning-rocm", "os":["linux"]},
+    {"file":"how-to/gpu-enabled-mpi", "os":["linux"]},
+    {"file":"how-to/system-debugging", "os":["linux"]},
+    {"file":"how-to/tuning-guides", "os":["linux", "windows"]},
+
+    {"file":"how-to/rocm-for-ai/index", "os":["linux"]},
+    {"file":"how-to/rocm-for-ai/install", "os":["linux"]},
+    {"file":"how-to/rocm-for-ai/train-a-model", "os":["linux"]},
+    {"file":"how-to/rocm-for-ai/deploy-your-model", "os":["linux"]},
+    {"file":"how-to/rocm-for-ai/hugging-face-models", "os":["linux"]},
+
+    {"file":"how-to/rocm-for-hpc/index", "os":["linux"]},
+
+    {"file":"how-to/llm-fine-tuning-optimization/index", "os":["linux"]},
+    {"file":"how-to/llm-fine-tuning-optimization/overview", "os":["linux"]},
+    {"file":"how-to/llm-fine-tuning-optimization/fine-tuning-and-inference", "os":["linux"]},
+    {"file":"how-to/llm-fine-tuning-optimization/single-gpu-fine-tuning-and-inference", "os":["linux"]},
+    {"file":"how-to/llm-fine-tuning-optimization/multi-gpu-fine-tuning-and-inference", "os":["linux"]},
+    {"file":"how-to/llm-fine-tuning-optimization/llm-inference-frameworks", "os":["linux"]},
+    {"file":"how-to/llm-fine-tuning-optimization/model-acceleration-libraries", "os":["linux"]},
+    {"file":"how-to/llm-fine-tuning-optimization/model-quantization", "os":["linux"]},
+    {"file":"how-to/llm-fine-tuning-optimization/optimizing-with-composable-kernel", "os":["linux"]},
+    {"file":"how-to/llm-fine-tuning-optimization/optimizing-triton-kernel", "os":["linux"]},
+    {"file":"how-to/llm-fine-tuning-optimization/profiling-and-debugging", "os":["linux"]},
+]
+
+exclude_patterns = ['temp']
+
+external_toc_path = "./sphinx/_toc.yml"
+
+extensions = ["rocm_docs", "sphinx_reredirects"]
+
+external_projects_current_project = "rocm"
+
+html_theme = "rocm_docs_theme"
+html_theme_options = {"flavor": "rocm-docs-home"}
+
+html_static_path = ["sphinx/static/css"]
+html_css_files = ["rocm_custom.css"]
+
+html_title = "ROCm Documentation"
+
+html_theme_options = {
+    "link_main_doc": False
+}
+
+redirects = {
+     "reference/openmp/openmp": "../../about/compatibility/openmp.html"
+}

docs/contribute/building.md

@@ -0,0 +1,150 @@
+<head>
+  <meta charset="UTF-8">
+  <meta name="description" content="Building ROCm documentation">
+  <meta name="keywords" content="documentation, Visual Studio Code, GitHub, command line,
+  AMD, ROCm">
+</head>
+
+# Building documentation
+
+You can build our documentation via GitHub (in a pull request) or locally (using the command line or
+Visual Studio (VS) Code.
+
+## GitHub
+
+If you open a pull request on the `develop` branch of a ROCm repository and scroll to the bottom of
+the page, there is a summary panel. 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.
+
+![Screenshot of the GitHub documentation build link](../data/contribute/github-docs-build.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. We use Python 3.8; other
+versions may not support the build.
+
+Use the Python Virtual Environment (`venv`) and run the following commands from the project root:
+
+```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
+```
+
+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.

docs/contribute/contributing.md

@@ -0,0 +1,112 @@
+<head>
+  <meta charset="UTF-8">
+  <meta name="description" content="Contributing to ROCm">
+  <meta name="keywords" content="ROCm, contributing, contribute, maintainer, contributor">
+</head>
+
+# Contribute to ROCm documentation
+
+All ROCm projects are GitHub-based, so if you want to contribute, you can do so by:
+
+* [Submitting a pull request in the appropriate GitHub repository](#submit-a-pull-request)
+* [Creating an issue in the appropriate GitHub repository](#create-an-issue)
+* [Suggesting a new feature](#suggest-a-new-feature)
+
+```{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 may use different licenses.
+```
+
+## Submit a pull request
+
+To make edits to our documentation via PR, follow these steps:
+
+1. Identify the repository and the file you want to update. For example, to update this page, you would
+  need to modify content located in this file:
+  `https://github.com/ROCm/ROCm/blob/develop/docs/contribute/contributing.md`
+
+2. (optional, but recommended) Fork the repository.
+
+3. Clone the repository locally and (optionally) add your fork. Select the green 'Code' button and copy
+   the URL (e.g., `git@github.com:ROCm/ROCm.git`).
+
+   * From your terminal, run:
+
+      ```bash
+      git clone git@github.com:ROCm/ROCm.git
+      ```
+
+   * Optionally add your fork to this local copy of the repository by running:
+
+      ```bash
+      git add remote <name-of-my-fork> <git@github.com:my-username/ROCm.git>
+      ```
+
+      To get the URL of your fork, go to your GitHub profile, select the fork and click the green 'Code'
+      button (the same process you followed to get the main GitHub repository URL).
+
+4. Change directory into your local copy of the repository, and run ``git pull`` (or ``git pull origin develop``) to ensure your local copy has the most recent content.
+
+5. Create and checkout a new branch using the following command:
+
+    ```bash
+    git checkout -b <branch_name>
+    ```
+
+6. Change directory into the `./docs` folder and make any documentation changes locally using your preferred code editor. Follow the guidelines listed on the
+   [documentation structure](./doc-structure.md) page.
+
+7. Optionally run a local test build of the documentation to ensure the content builds and looks as expected. In your terminal, run the following commands from within the `./docs` folder of your cloned repository:
+
+     ```bash
+     pip3 install -r sphinx/requirements.txt  # You only need to run this command once
+     python3 -m sphinx -T -E -b html -d _build/doctrees -D language=en . _build/html
+     ```
+
+    The build output files are located in the `docs/_build` folder. To preview your build, open the index file
+    (`docs/_build/html/index.html`) file. For more information, see [Building documentation](building.md). To learn
+    more about our build tools, see [Documentation toolchain](toolchain.md).
+
+8. Commit your changes and push them to GitHub by running:
+
+    ```bash
+    git add <path-to-my-modified-file> # To add all modified files, you can use: git add .
+    git commit -m "my-updates"
+    git push <name-of-my-fork>
+    ```
+
+    After pushing, you will get a GitHub link in the terminal output. Copy this link and paste it into a
+    browser to create your PR.
+
+## Create an issue
+
+1. To create a new GitHub issue, select the 'Issues' tab in the appropriate repository
+  (e.g., https://github.com/ROCm/ROCm/issues).
+2. Use the search bar to make sure the issue doesn't already exist.
+3. If your issue is not already listed, select the green 'New issue' button to the right of the page. Select
+  the type of issue and fill in the resulting template.
+
+### General issue guidelines
+
+* Use your best judgement for issue creation. If your issue is already listed, upvote the issue and
+  comment or post to provide additional details, such as how you reproduced this issue.
+* If you're not sure if your issue is the same, err on the side of caution and file your issue.
+  You can add a comment to include the issue number (and link) for the similar issue. If we evaluate
+  your issue as being the same as the existing issue, we'll close the duplicate.
+* If your issue doesn't exist, use the issue template to file a new issue.
+  * When filing an issue, be sure to provide as much information as possible, including script output so
+    we can collect information about your configuration. This helps reduce the time required to
+    reproduce your issue.
+  * Check your issue regularly, as we may require additional information to successfully reproduce the
+    issue.
+
+## Suggest a new feature
+
+Use the [GitHub Discussion forum](https://github.com/ROCm/ROCm/discussions)
+(Ideas category) to propose new features. Our maintainers are happy to provide direction and
+feedback on feature development.
+
+## Future development workflow
+
+The current ROCm development workflow is GitHub-based. If, in the future, we change this platform,
+the tools and links may change. In this instance, we will update contribution guidelines accordingly.

docs/contribute/doc-structure.md

@@ -0,0 +1,219 @@
+# Documentation structure
+
+Our documentation follows the Pitchfork folder structure. Most documentation files are stored in the
+`/docs` folder. Some special files (such as release, contributing, and changelog) are stored in the root
+(`/`) folder.
+
+All images are stored in the `/docs/data` folder. An image's file path mirrors that of the documentation
+file where it is used.
+
+Our naming structure uses kebab case; for example, `my-file-name.rst`.
+
+## Supported formats and syntax
+
+Our documentation includes both Markdown and RST files. We are gradually transitioning existing
+Markdown to RST in order to more effectively meet our documentation needs. When contributing,
+RST is preferred; if you must use Markdown, use GitHub-flavored Markdown.
+
+We use [Sphinx Design](https://sphinx-design.readthedocs.io/en/latest/index.html) syntax and compile
+our API references using [Doxygen](https://www.doxygen.nl/).
+
+The following table shows some common documentation components and the syntax convention we
+use for each:
+
+<table>
+<tr>
+<th>Component</th>
+<th>RST syntax</th>
+</tr>
+<tr>
+<td>Code blocks</td>
+<td>
+
+```rst
+
+.. code-block:: language-name
+
+  My code block.
+
+
+```
+
+</td>
+</tr>
+<tr>
+<td>Cross-referencing internal files</td>
+<td>
+
+```rst
+
+:doc:`Title <../path/to/file/filename>`
+
+```
+
+</td>
+</tr>
+<tr>
+<td>External links</td>
+<td>
+
+```rst
+
+`link name  <URL>`_
+
+```
+
+</td>
+</tr>
+<tr>
+<tr>
+<td>Headings</td>
+<td>
+
+```rst
+
+******************
+Chapter title (H1)
+******************
+
+Section title (H2)
+===============
+
+Subsection title (H3)
+---------------------
+
+Sub-subsection title (H4)
+^^^^^^^^^^^^^^^^^^^^
+
+
+```
+
+</td>
+</tr>
+<tr>
+<td>Images</td>
+<td>
+
+```rst
+
+.. image:: image1.png
+
+```
+
+</td>
+</tr>
+<tr>
+<td>Internal links</td>
+<td>
+
+```rst
+
+1. Add a tag to the section you want to reference:
+
+.. _my-section-tag: section-1
+
+Section 1
+==========
+
+2. Link to your tag:
+
+As shown in :ref:`section-1`.
+
+```
+
+</td>
+</tr>
+<tr>
+<tr>
+<td>Lists</td>
+<td>
+
+```rst
+
+# Ordered (numbered) list item
+
+* Unordered (bulleted) list item
+
+```
+
+</td>
+</tr>
+<tr>
+<tr>
+<td>Math (block)</td>
+<td>
+
+```rst
+
+.. math::
+
+  A = \begin{pmatrix}
+          0.0 & 1.0 & 1.0 & 3.0 \\
+          4.0 & 5.0 & 6.0 & 7.0 \\
+        \end{pmatrix}
+
+```
+
+</td>
+</tr>
+<tr>
+<td>Math (inline)</td>
+<td>
+
+```rst
+
+:math:`2 \times 2 `
+
+```
+
+</td>
+</tr>
+<tr>
+<td>Notes</td>
+<td>
+
+```rst
+
+.. note::
+
+  My note here.
+
+```
+
+</td>
+</tr>
+<tr>
+<td>Tables</td>
+<td>
+
+```rst
+
+.. csv-table::  Optional title here
+  :widths: 30, 70  #optional column widths
+  :header: "entry1 header", "entry2 header"
+
+   "entry1", "entry2"
+
+```
+
+</td>
+</tr>
+</table>
+
+## Language and style
+
+We use the
+[Google developer documentation style guide](https://developers.google.com/style/highlights) to
+guide our content.
+
+Font size and type, page layout, white space control, and other formatting
+details are controlled via
+[rocm-docs-core](https://github.com/ROCm/rocm-docs-core). If you want to notify us
+of any formatting issues, create a pull request in our
+[rocm-docs-core](https://github.com/ROCm/rocm-docs-core) GitHub repository.
+
+## Building our documentation
+
+<!--  % TODO: Fix the link to be able to work at every files  -->
+To learn how to build our documentation, refer to
+[Building documentation](./building.md).

docs/contribute/feedback.md

@@ -0,0 +1,31 @@
+<head>
+  <meta charset="UTF-8">
+  <meta name="description" content="Providing feedback for ROCm documentation">
+  <meta name="keywords" content="documentation, pull request, GitHub, AMD, ROCm">
+</head>
+
+# Providing feedback
+
+There are four standard ways to provide feedback on this repository.
+
+## Pull request
+
+All contributions to ROCm documentation should arrive via the
+[GitHub Flow](https://docs.github.com/en/get-started/quickstart/github-flow)
+targeting the develop branch of the repository.
+
+For more in-depth information on creating a pull request (PR), see
+[Contributing](./contributing.md).
+
+## GitHub discussions
+
+To ask questions or view answers to frequently asked questions, refer to
+[GitHub Discussions](https://github.com/ROCm/ROCm/discussions).
+On GitHub Discussions, in addition to asking and answering questions,
+members can share updates, have open-ended conversations,
+and follow along on via public announcements.
+
+## GitHub issue
+
+Issues on existing or absent documentation can be filed in
+[GitHub Issues](https://github.com/ROCm/ROCm/issues).

docs/contribute/toolchain.md

@@ -0,0 +1,65 @@
+<head>
+  <meta charset="UTF-8">
+  <meta name="description" content="ROCm documentation toolchain">
+  <meta name="keywords" content="documentation, toolchain, Sphinx, Doxygen, MyST, AMD, ROCm">
+</head>
+
+# ROCm documentation toolchain
+
+Our 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 customization for our documentation. This project is the tool most ROCm
+repositories use as part of the documentation build. It is also 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
+{doc}`rocm-docs-core documentation<rocm-docs-core:index>`.
+
+## 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 inline
+code. ROCm projects typically use Doxygen for public API documentation (unless the upstream project
+uses a different tool).
+
+## Breathe
+
+[Breathe](https://www.breathe-doc.org/) is a Sphinx plugin to integrate Doxygen content.
+
+## MyST
+
+[Markedly Structured Text (MyST)](https://myst-tools.org/docs/spec) is an extended flavor of
+Markdown ([CommonMark](https://commonmark.org/)) influenced by reStructuredText (RST) and
+Sphinx. It's integrated into ROCm documentation by the Sphinx extension
+[`myst-parser`](https://myst-parser.readthedocs.io/en/latest/).
+A MyST syntax cheat sheet is available on the [Jupyter reference](https://jupyterbook.org/en/stable/reference/cheatsheet.html) site.
+
+## Read the Docs
+
+[Read the Docs](https://docs.readthedocs.io/en/stable/) is the service that builds and hosts the HTML
+documentation generated using Sphinx to our end users.