AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Add 'projects/hip/' from commit 'e74b05a7bd9454b97dc04d7cc4b66d1fe6c534a7'

Browse Source 
git-subtree-dir: projects/hip
git-subtree-mainline: 64df0940b8
git-subtree-split: e74b05a7bd
This commit is contained in:
systems-assistant[bot]
2025-08-10 02:09:42 +00:00
parent 64df0940b8 e74b05a7bd
commit d76041b87b
231 changed files with 63214 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
projects/hip/docs/install/build.rst
+241
View File
@@ -0,0 +1,241 @@
.. meta::
:description: This page gives instructions on how to build HIP from source.
:keywords: AMD, ROCm, HIP, build, build instructions, source
*******************************************
Build HIP from source
*******************************************
Prerequisites
=================================================
HIP code can be developed either on AMD ROCm platform using HIP-Clang compiler, or a CUDA platform with ``nvcc`` installed.
Before building and running HIP, make sure drivers and prebuilt packages are installed properly on the platform.
You also need to install Python 3, which includes the ``CppHeaderParser`` package.
Install Python 3 using the following command:
.. code-block:: shell
apt-get install python3
Check and install ``CppHeaderParser`` package using the command:
.. code-block:: shell
pip3 install CppHeaderParser
Install ``ROCm LLVM`` package using the command:
.. code-block:: shell
apt-get install rocm-llvm-dev
.. _Building the HIP runtime:
Building the HIP runtime
==========================================================
Set the repository branch using the variable: ``ROCM_BRANCH``. For example, for ROCm 6.1, use:
.. code-block:: shell
export ROCM_BRANCH=rocm-6.1.x
.. tab-set::
.. tab-item:: AMD
:sync: amd
#. Get HIP source code.
.. note::
Starting in ROCM 5.6, CLR is a new repository that includes the former ROCclr, HIPAMD and
OpenCl repositories. OpenCL provides headers that ROCclr runtime depends on.
.. note::
Starting in ROCM 6.1, a new repository ``hipother`` is added to ROCm, which is branched out from HIP.
``hipother`` provides files required to support the HIP back-end implementation on some non-AMD platforms,
like NVIDIA.
.. code-block:: shell
git clone -b "$ROCM_BRANCH" https://github.com/ROCm/clr.git
git clone -b "$ROCM_BRANCH" https://github.com/ROCm/hip.git
CLR (Compute Language Runtime) repository includes ROCclr, HIPAMD and OpenCL.
ROCclr (ROCm Compute Language Runtime) is a virtual device interface which
is defined on the AMD platform. HIP runtime uses ROCclr to interact with different backends.
HIPAMD provides implementation specifically for HIP on the AMD platform.
OpenCL provides headers that ROCclr runtime currently depends on.
hipother provides headers and implementation specifically for non-AMD HIP platforms, like NVIDIA.
#. Set the environment variables.
.. code-block:: shell
export CLR_DIR="$(readlink -f clr)"
export HIP_DIR="$(readlink -f hip)"
#. Build HIP.
.. code-block:: shell
cd "$CLR_DIR"
mkdir -p build; cd build
cmake -DHIP_COMMON_DIR=$HIP_DIR -DHIP_PLATFORM=amd -DCMAKE_PREFIX_PATH="/opt/rocm/" -DCMAKE_INSTALL_PREFIX=$PWD/install -DHIP_CATCH_TEST=0 -DCLR_BUILD_HIP=ON -DCLR_BUILD_OCL=OFF ..
make -j$(nproc)
sudo make install
.. note::
Note, if you don't specify ``CMAKE_INSTALL_PREFIX``, the HIP runtime is installed at
``<ROCM_PATH>``.
By default, release version of HIP is built. If need debug version, you can put the option ``CMAKE_BUILD_TYPE=Debug`` in the command line.
Default paths and environment variables:
* HIP is installed into ``<ROCM_PATH>``. This can be overridden by setting the ``INSTALL_PREFIX`` as the command option.
environment variable.
* HSA is in ``<ROCM_PATH>``. This can be overridden by setting the ``HSA_PATH``
environment variable.
* Clang is in ``<ROCM_PATH>/llvm/bin``. This can be overridden by setting the
``HIP_CLANG_PATH`` environment variable.
* The device library is in ``<ROCM_PATH>/lib``. This can be overridden by setting the
``DEVICE_LIB_PATH`` environment variable.
* Optionally, you can add ``<ROCM_PATH>/bin`` to your ``PATH``, which can make it easier to
use the tools.
* Optionally, you can set ``HIPCC_VERBOSE=7`` to output the command line for compilation.
After you run the ``make install`` command, HIP is installed to ``<ROCM_PATH>`` by default, or ``$PWD/install/hip`` while ``INSTALL_PREFIX`` is defined.
#. Generate a profiling header after adding/changing a HIP API.
When you add or change a HIP API, you may need to generate a new ``hip_prof_str.h`` header.
This header is used by ROCm tools to track HIP APIs, such as ``rocprofiler`` and ``roctracer``.
To generate the header after your change, use the ``hip_prof_gen.py`` tool located in
``hipamd/src``.
Usage:
.. code-block:: shell
`hip_prof_gen.py [-v] <input HIP API .h file> <patched srcs path> <previous output> [<output>]`
Flags:
* ``-v``: Verbose messages
* ``-r``: Process source directory recursively
* ``-t``: API types matching check
* ``--priv``: Private API check
* ``-e``: On error exit mode
* ``-p``: ``HIP_INIT_API`` macro patching mode
Example usage:
.. code-block:: shell
hip_prof_gen.py -v -p -t --priv <hip>/include/hip/hip_runtime_api.h \
<hipamd>/src <hipamd>/include/hip/amd_detail/hip_prof_str.h \
<hipamd>/include/hip/amd_detail/hip_prof_str.h.new
.. tab-item:: NVIDIA
:sync: nvidia
#. Get the HIP source code.
.. code-block:: shell
git clone -b "$ROCM_BRANCH" https://github.com/ROCm/clr.git
git clone -b "$ROCM_BRANCH" https://github.com/ROCm/hip.git
git clone -b "$ROCM_BRANCH" https://github.com/ROCm/hipother.git
#. Set the environment variables.
.. code-block:: shell
export CLR_DIR="$(readlink -f clr)"
export HIP_DIR="$(readlink -f hip)"
export HIP_OTHER="$(readlink -f hipother)"
#. Build HIP.
.. code-block:: shell
cd "$CLR_DIR"
mkdir -p build; cd build
cmake -DHIP_COMMON_DIR=$HIP_DIR -DHIP_PLATFORM=nvidia -DCMAKE_INSTALL_PREFIX=$PWD/install -DHIP_CATCH_TEST=0 -DCLR_BUILD_HIP=ON -DCLR_BUILD_OCL=OFF -DHIPNV_DIR=$HIP_OTHER/hipnv ..
make -j$(nproc)
sudo make install
Build HIP tests
=================================================
.. tab-set::
.. tab-item:: AMD
:sync: amd
* Build HIP catch tests.
HIP catch tests are separate from the HIP project and use Catch2.
* Get HIP tests source code.
.. code-block:: shell
git clone -b "$ROCM_BRANCH" https://github.com/ROCm/hip-tests.git
* Build HIP tests from source.
.. code-block:: shell
export HIPTESTS_DIR="$(readlink -f hip-tests)"
cd "$HIPTESTS_DIR"
mkdir -p build; cd build
cmake ../catch -DHIP_PLATFORM=amd -DHIP_PATH=$CLR_DIR/build/install # or any path where HIP is installed; for example: ``/opt/rocm``
make build_tests
ctest # run tests
HIP catch tests are built in ``$HIPTESTS_DIR/build``.
To run any single catch test, use this example:
.. code-block:: shell
cd $HIPTESTS_DIR/build/catch_tests/unit/texture
./TextureTest
* Build a HIP Catch2 standalone test.
.. code-block:: shell
cd "$HIPTESTS_DIR"
hipcc $HIPTESTS_DIR/catch/unit/memory/hipPointerGetAttributes.cc \
-I ./catch/include ./catch/hipTestMain/standalone_main.cc \
-I ./catch/external/Catch2 -o hipPointerGetAttributes
./hipPointerGetAttributes
...
All tests passed
.. tab-item:: NVIDIA
:sync: nvidia
The commands to build HIP tests on an NVIDIA platform are the same as on an AMD platform.
However, you must first set ``-DHIP_PLATFORM=nvidia``.
Run HIP
=================================================
After installation and building HIP, you can compile your application and run.
Simple examples can be found in the `ROCm-examples repository <https://github.com/ROCm/rocm-examples>`_.
projects/hip/docs/install/install.rst
+111
View File
@@ -0,0 +1,111 @@
.. meta::
:description: This page explains how to install HIP
:keywords: AMD, ROCm, HIP, install, installation
*******************************************
Install HIP
*******************************************
HIP can be installed on AMD (ROCm with HIP-Clang) and NVIDIA (CUDA with NVCC) platforms.
.. note::
The version definition for the HIP runtime is different from CUDA. On AMD
platforms, the :cpp:func:`hipRuntimeGetVersion` function returns the HIP
runtime version. On NVIDIA platforms, this function returns the CUDA runtime
version.
.. _install_prerequisites:
Prerequisites
=======================================
.. tab-set::
.. tab-item:: AMD
:sync: amd
Refer to the Prerequisites section in the ROCm install guides:
* :doc:`rocm-install-on-linux:reference/system-requirements`
* :doc:`rocm-install-on-windows:reference/system-requirements`
.. tab-item:: NVIDIA
:sync: nvidia
With NVIDIA GPUs, HIP requires unified memory. All CUDA-enabled NVIDIA
GPUs with compute capability 5.0 or later should be supported. For more
information, see `NVIDIA's list of CUDA enabled GPUs <https://developer.nvidia.com/cuda-gpus>`_.
Installation
=======================================
.. tab-set::
.. tab-item:: AMD
:sync: amd
HIP is automatically installed during the ROCm installation. If you haven't yet installed ROCm, you
can find installation instructions here:
* :doc:`rocm-install-on-linux:index`
* :doc:`rocm-install-on-windows:index`
By default, HIP is installed into ``/opt/rocm``.
.. note::
There is no autodetection for the HIP installation. If you choose to install it somewhere other than the default location, you must set the ``HIP_PATH`` environment variable as explained in `Build HIP from source <./build.html>`_.
.. tab-item:: NVIDIA
:sync: nvidia
#. Install the NVIDIA toolkit.
The latest release can be found here:
`CUDA Toolkit <https://developer.nvidia.com/cuda-downloads>`_.
#. Setup the radeon repo.
.. code-block::shell
# Replace url with appropriate link in the table below
wget https://repo.radeon.com/amdgpu-install/6.2/distro/version_name/amdgpu-install_6.2.60200-1_all.deb
sudo apt install ./amdgpu-install_6.2.60200-1_all.deb
sudo apt update
.. list-table:: amdgpu-install links
:widths: 25 100
:header-rows: 1
* - Ubuntu version
- URL
* - 24.04
- https://repo.radeon.com/amdgpu-install/6.2.4/ubuntu/noble/amdgpu-install_6.2.60204-1_all.deb
* - 22.04
- https://repo.radeon.com/amdgpu-install/6.2.4/ubuntu/jammy/amdgpu-install_6.2.60204-1_all.deb
#. Install the ``hip-runtime-nvidia`` and ``hip-dev`` packages. This installs the CUDA SDK and HIP
porting layer.
.. code-block:: shell
apt-get install hip-runtime-nvidia hip-dev
The default paths are:
* CUDA SDK: ``/usr/local/cuda``
* HIP: ``/opt/rocm``
#. Set the HIP_PLATFORM to nvidia.
.. code-block:: shell
export HIP_PLATFORM="nvidia"
Verify your installation
==========================================================
Run ``hipconfig`` in your installation path.
.. code-block:: shell
/opt/rocm/bin/hipconfig --full