colramos-amd
2.4 KiB
Standalone GUI Analyzer
.. toctree::
:glob:
:maxdepth: 4
Features
Omniperf's standalone GUI analyzer is a lightweight web page that can be generated straight from the command-line. This option is great for users who want immediate access to graphical results without the server-side overhead of the Omniperf's detailed Grafana interface
The standalone GUI analyzer is a simple Flask web application that uses port forwarding (DEFAULT: 8050) to allow users to view results from their web browser.
Because the standalone GUI analyzer uses port forwarding we reccomend users who profile on shared clusters scp their results to their personal workstation for analysis.
Useage
To launch the standalone GUI, users will include the --gui flag in their analysis command.
To launch the web application on a port other than 8050 (default)
use --gui <desired port>
$ omniperf analyze -p workloads/vcopy/mi200/ --gui
--------
Analyze
--------
Dash is running on http://0.0.0.0:8050/
* Serving Flask app 'omniperf_cli.omniperf_cli' (lazy loading)
* Environment: production
WARNING: This is a development server. Do not use it in a production deployment.
Use a production WSGI server instead.
* Debug mode: off
* Running on all addresses (0.0.0.0)
WARNING: This is a development server. Do not use it in a production deployment.
* Running on http://127.0.0.1:8050
* Running on http://10.228.32.139:8050 (Press CTRL+C to quit)
Users can then lauch their web browser of choice and go to http://localhost:8050/ (substituting port if overridden).
When no filters are applied users will see 5 basic section derived from their application's profiling data:
- Memory Chart Analysis
- Empirical Roofline Analysis
- Top Stats (Top Kernal Statistics)
- System Info
- System Speed-of-Light
To dive deeper, use the top drop down menus to isolate a particular kernel(s) or dispatch(s). You'll then see the webpage updates with metrics specific to the filter you've applied.
Once you've applied a filter you'll also see several additional sections become availible with detailed metrics specific to that area of AMD hardware. These detailed sections mirror the data we display in Omniperf's Grafana interface