Python 3 environment for benchmarking PRIISM

This repository manages dependent Python packages by Poetry. Please install it to your environment if you do not have it.

curl -sSL https://raw.githubusercontent.com/python-poetry/poetry/master/get-poetry.py | python3 -

This will create a virtual environment and install the dependencies on it.

git clone https://github.com/astropenguin/priism-benchmark.git
cd priism-benchmark
poetry install

This will download the HL Tau SV data and run SpM imaging with multiple PRIISM versions and numbers of threads:

poetry run bin/hltau/benchmark priism-0.7.15,priism-0.7.16 32,64

where the first argument should be comma-separated PRIISM versions (i.e. tag or commit ref.) and the second argument should be comma-separated OMP_NUM_THERADSs.

Output FITS images and a log file will be put in /path/to/repository/log/hltau. The format of the log file (benchmark.csv) is like:

PRIISM version,Number of threads,Elapsed time (s)
priism-0.7.15,32,113.62
priism-0.7.15,64,144.67
priism-0.7.16,32,97.42
priism-0.7.16,64,106.51

where the first, second, and third columns correspond to the PRIISM version, the number of threads used for imaging, and the elapsed time (in seconds) of imaging (i.e. imager.solve(...)), respectively.

PRIISM build options can be specified by an environment variable PRIISM_BUILD_OPTIONS:

PRIISM_BUILD_OPTIONS="-X yes" poetry run bin/hltau/benchmark priism-0.7.15,priism-0.7.16 8,32,64

You may also use this repository as a simple way of installing PRIISM (and dependencies):

poetry install
PRIISM_BUILD_OPTIONS="-X yes" poetry run bin/install-priism priism-0.7.16

Then you can run your own script that uses PRIISM:

poetry run python /path/to/script.py