Skip to content

Latest commit

 

History

History
87 lines (60 loc) · 2.38 KB

README.md

File metadata and controls

87 lines (60 loc) · 2.38 KB
Raw

Sandmark Notebooks

These notebooks are used to visualize the results from the benchmarking runs.

Requirements

You'll need to install python3 with jupyter, and the following dependency packages:

  • seaborn
  • pandas
  • nested_dict
  • papermill

If you are using JupyterHub then you can install the packages using pip inside the JupyterHub terminal:

$ sudo -E pip install pandas seaborn nested_dict papermill

Run

You can start the notebook using the following command:

$ jupyter notebook

Sequential

The sequential-interactive.ipynb notebook can be invoked interactively in the GUI or from the command-line.

GUI

If you are running the notebook from the GUI, ensure that you set the artifacts_dir to point to the correct location of the artifacts/ folder.

As you execute the cells in the notebook, you will be presented with an option to select the number of bench comparisons, the individual bench files and a baseline variant for comparison as shown below:

Select comparisons and benches

Select baseline variant

Command-line

The notebook can be executed through the command-line using papermill as shown below:

$ papermill sequential-interactive.ipynb sequential-interactive-output.ipynb -f parameters.yaml

A sample parameters.yaml file is provided below for reference:

enable_gui: False
artifacts_dir: ./test/artifacts/
selected_benches:
  - [ turing.ocamllabs.io, ocaml_5.00__5.00, b73cbbea4bc40ffd26a459d594a39b99cec4273d vanilla 20220228_010330 5.00+stable+sequential_1.orun.summary.bench ]
  - [ turing.ocamllabs.io, ocaml_5.00__5.00, f3ea33a4c9b0d2872b0bcccbcce5428b2e468373 vanilla 20220223_010015 5.00+trunk+sequential_1.orun.summary.bench ]
selected_baseline: 5.0.0

Since, we are running the notebook through the CLI, the enable_gui variable should be set to False.

The artifacts_dir should be set correctly and must include the artifacts folder.

The list of bench files for comparison are to be provided in an array that uses the following syntax and semantics:

- [HOST, REPOSITORY, COMMIT VARIANT TIMESTAMP BENCH]

You can specify a baseline variant in the selected_baseline variable.

A sample test/artifacts/ folder with bench files is available that you you can try to execute using papermill.