Signed-off-by: Karl W. Schulz <karl.schulz@amd.com>
5.2 KiB
Omniperf Deployment
.. toctree::
:glob:
:maxdepth: 4
Omniperf is broken into two installation components:
- Omniperf Client-side (Required)
- Provides core application profiling capability
- Allows collection of performance counters, filtering by IP block, dispatch, kernel, etc
- CLI based analysis mode
- Stand alone web interface for importing analysis metrics
- Omniperf Server-side (Optional)
- Mongo DB backend + Grafana instance
Client-side Installation
Omniperf requires the following basic software dependencies prior to usage:
- Python (>=3.7)
- CMake (>= 3.19)
- ROCm (>= 5.1)
In addition, Omniperf leverages a number of Python packages that are
documented in the top-level requirements.txt file. These must be
installed prior to Omniperf configuration.
The recommended procedure for Omniperf usage is to install into a shared file system so that multiple users can access the final installation. The following steps illustrate how to install the necessary python dependencies using pip and Omniperf into a shared location controlled by the INSTALL_DIR environment variable.
The following installation example leverages several
[CMake](https://cmake.org/cmake/help/latest/) project variables
defined as follows:
| Variable | Description |
| -------------------- | -------------------------------------------------------------------- |
| CMAKE_INSTALL_PREFIX | controls install path for Omniperf files |
| PYTHON_DEPS | provides optional path to resolve Python package dependencies |
| MOD_INSTALL_PATH | provides optional path for separate Omniperf modulefile installation |
A typical install will begin by downloading the latest release tarball available from the Releases section of the Omniperf development site. From there, untar and descend into the top-level directory as follows:
$ tar xfz omniperf-v{__VERSION__}.tar.gz
$ cd omniperf-v{__VERSION__}
Next, install Python dependencies and complete the Omniperf configuration/install process as follows:
# define top-level install path
$ export INSTALL_DIR=<your-top-level-desired-install-path>
# install python deps
$ python3 -m pip install -t ${INSTALL_DIR}/python-libs -r requirements.txt
# configure Omniperf for shared install
$ mkdir build
$ cd build
$ cmake -DCMAKE_INSTALL_PREFIX=${INSTALL_DIR}/{__VERSION__} \
-DPYTHON_DEPS=${INSTALL_DIR}/python-libs \
-DMOD_INSTALL_PATH=${INSTALL_DIR}/modulefiles ..
# install
$ make install
You may require `sudo` during the final install step if you
do not have write access to the chose install path.
After completing these steps, a successful top-level installation directory looks as follows:
$ ls $INSTALL_DIR
modulefiles {__VERSION__} python-libs
Execution using modulefiles
The installation process includes creation of an environment modulefile for use with Lmod. On systems that support Lmod, a user can register the Omniperf modulefile directory and setup their environment for execution of Omniperf as follows:
$ module use $INSTALL_DIR/modulefiles
$ module load omniperf
$ which omniperf
/opt/apps/omniperf/{__VERSION__}/bin/omniperf
$ omniperf --version
ROC Profiler: /opt/rocm-5.1.0/bin/rocprof
omniperf (v{__VERSION__})
customize the resulting Omniperf modulefile post-installation to
include additional module dependencies.
Execution without modulefiles
To use Omniperf without the companion modulefile, update your PATH
settings to enable access to the command-line binary. If you installed Python
dependencies in a shared location, update your PYTHONPATH config as well:
export PATH=$INSTALL_DIR/{__VERSION__}/bin:$PATH
export PYTHONPATH=$INSTALL_DIR/python-libs
rocProf
Omniperf relies on a rocprof binary during the profiling
process. Normally the path to this binary will be detected
automatically, but it can also be overridden via the use of an
optional ROCPROF environment variable.
%%% ### Generate Packaging
%%% console %%% cd build %%% cpack -G STGZ %%% cpack -G DEB -D CPACK_PACKAGING_INSTALL_PREFIX=/opt/omniperf %%% cpack -G RPM -D CPACK_PACKAGING_INSTALL_PREFIX=/opt/omniperf %%%
Omniperf Server Setup
Note: Server-side setup is not required to profile or analyze performance data from the CLI. It is provided as an additional mechanism to import performance data for examination within a detailed Grafana GUI.
The recommended process for enabling the server-side of Omniperf is to use the provided Docker file to build the Grafana and MongoDB instance.
Persist Storage
$ sudo mkdir -p /usr/local/persist && cd /usr/local/persist/
$ sudo mkdir -p grafana-storage mongodb
$ sudo mkdir -p grafana-storage mongodb
Build and Launch
$ sudo docker-compose build
$ sudo docker-compose up