Installation

Docker

• Build the docker image with the Dockerfile contained in the docker directory
• Run the image

Local installation

• (tested on Fedora 22)
• Download and install Racket and optionally Larceny
• Clone the racr repo
• Install racr using install-libraries.bash
• Let scheme IDE know about racr
  • Racket: Create a symlink in $RACKET_INSTALL_DIR/collects pointing to $RACR_REPO/racr/racket-bin/racr
• Install Fabric
  • best installed using pip install fabric

Using racr-mquat

• to build the project for Racket, use make (prefered if you changed the source files and want to recompile just the affected files)
  • alternatively use fab install.racket to unconditionally build everything for Racket
• everything else is guided by fabric
• to list all possible commands, use fab -l
• to get a description of a task, use fab -d <TASKNAME>

Generell workflow for executing measurements

• Optional: check ilp-measurements.scm to contain the correct experiments to run (usually the case, if the file was not changed)
• Run fab properties.check to see the current experiment settings
• Optional: run fab properties.setup to change all properties or fab properties.setup:<NAME> to change a specific property
• Start the measurements with fab measure.racket (*) or fab measure.larceny (*) using Racket or Larceny, respectively
• Optional: run fab measure.sol (*) to run the solver
  • this uses glpsol and will fail, if it's not installed
  • this requires to run a measurement beforehand with the properties measure.lp.write = 1, i.e., some ILP files to operate on
• Run fab measure.conflate-results to update the conflated result files
• Run fab measure.distinction-of-changes to create result files split by strategy found in directory profiling/splitted
• Optional: Rerun notebook to update diagrams using runipy ilp-measurement.ipynb

(*) For these tasks, there are variants with "-n" append taking as first parameter the number of repetitions