You can use the roofline_python tool from the command line or a python notebook.
Use the python notebook in either jupyter notebook (recommended) or google colab and create a roofline plot with or without application points from a json (right now the json must be in the format generated by the Berkeley ERT)
Dependencies:
- Anaconda (or alternatively, just matplotlib and pandas)
- Seaborn (you can conda install seaborn, or pip install seaborn)
To run the script version here are the parameters:
- input (-i): Path to the roofline JSON file
- graphconfig (-g): The json file that contains the desired traits for your output graph
Graph_config.json default fields can be updated, keep keys the same:
- "title": false (or string, the graph title)
- "xlabel": "FLOPs / Byte" (string, the x axis label)
- "ylabel": "GFLOPs / Second" (string, the y axis label)
- "table": false (boolean, whether you want a table or not)
- "outfile":"roofline_graph.svg" (string, the filename to store your graph)
- "appdata": "app_data.csv" (false or string, the filename containing application data)
Example:
python3 roofline_python.py -i input_files/roofline.json -g input_files/graph_config.json