Approaches to Interpreter Composition: Repeatable Experiments

Edd Barrett, Carl Friedrich Bolz, and Laurence Tratt

Our paper, "Approaches to Interpreter Composition", contains some micro-benchmarks. We believe strongly in repeatability in computing experiments. We provide source code and a VirtualBox image of our experiment.

The source tarball contains the source code (including version control history) and scripts to build the VMs from the paper and run the experiment. The VirtualBox image contains the VMs pre-built and the benchmarks ready to run. The VirtualBox image also demonstrates the Connect4 demo, ready to run.

Download

These large files are kindly hosted by the Internet Archive under the identifier comlan15_interpcomp_artefacts.

To use these files, decompress and read the README file contained within.

Raw Results

Raw results from different benchmarking machines can be downloaded below. Each download is a xz tarball containing raw data and information about the system upon which it ran.

The results used in the paper are from interpcomp_results_bencher3_20150515.tar.xz.

Note that the composed VM names in the raw results are codenames that we used during development:

CodenameFormal Name
UnipycationUnipycation
SwithonCPython-SWI
JythogJython-tuProlog
SwiPyPyPy-SWI