olekscode / lapack-experiment

pharo-ai Lapack experiment

Requirements

• Have installed python3, scikit-learn, pandas and Pharo11 on your computer.
• Have LAPACK installed on your computer
• Have around 7GB of free memory space in the hard drive for the datasets.

Generating the datasets

To generate the datasets, you need to run the file DatasetGenerator.py using python3.

python3 DatasetGenerator.py

This step can take until 15 minutes because the datasets can be big.

The script will generate the three datasets (small, medium and large) into the data directory. Note that the datasets are generated using a random seed. So you always have exactly the same datasets.

Running the Python benchmarks

To run the Python benchmarks, you need to execute the file PythonExperiment.py using python3.

The script will run the benchmarks and create a csv file with the results in the directory ./experiment-results/python-results.csv

Running the Pharo benchmarks

First, you need to download the Pharo launcher from the Pharo website https://pharo.org/download. Then, create a Pharo 11 image.

For importing the code into the Pharo image you need to use iceberg:

• Open iceberg
• Click on the right corner button "Add +"
• Select "Import from existing clone" option
• Then select the directory of where did you cloned the repo. Note that you must have already the datasets generated to be able to run the code
• Finnally, in the initial window of Iceberg, select the repository in the list "lapack-experiment", then right click - Metacello - Install baseline of AILapackExperiment (Default)

Finally, you need execute in a Playground

PharoExperiment new runExperiment

The script runs the benchmarks and opens a text presenter with the results.