SC25 Academy Evaluation

Benchmarks and evaluation code for the SC25 submission on Academy.

Installation

The bench/ package contains all of the microbenchmarks for the paper.

python -m venv venv
. venv/bin/activate
pip install -e .

This installs the requirements for notebooks/ as well, but the analysis specific requirements are also defined in notebooks/requirements.txt.

Running Benchmarks

Each benchmark is an executable module in the bench package. E.g.,

python -m bench.action_latency --help

Performing Analysis

All results used in the paper are provided in data/, and all figures have an associated Jupyter Notebook in notebooks/.