This github repository contains the necessary docker script to set the environment for seven general-purpose polyhedral compilers and build those compilers in an Unix/Linux environment. Those seven polyhedral compilers are as follows: Pluto, PoCC, Polly, Graphite, PolyOpt, PPCG, and Polygeist.

It also contains the PolyBench/C benchmarks to test them and measure the execution time of the resulting codes. Currently, we only support the execution for CPU architectures.

This Docker installation will consume more than 50GB of disk space. Clone the above git repository. The folder docker contains all the necessary files to build the docker environment. The commands are as follows:

git clone https://github.com/vincentloechner/PolyhedralCompilers.git
cd PolyhedralCompilers/docker
docker build -f Docker-polyhedral-src . -t polyhedral
docker run --volume $(pwd):/results -it --privileged polyhedral bash

Once the docker environment is set with docker run the docker machine with built polyhedral compilers can be invoked. The polyhedral compiler's source files and their respective build binaries are found inside the docker.

We rely on PolyBench/C - a set of 30 numerical benchmarks that has polyhedral nature of nested loops. We have two scripts execute_polyhedral.sh and execute_polyhedral_syn.sh to test the parallel and non-parallel execution of polyhedral compilers. Some examples:

To execute all 30 benchmarks against gcc, clang, icc, rose, and seven polyhedral compilers in parallel, run:

cd polybench
./execute_polyhedral.sh

To execute the 2mm benchmark against gcc, clang, icc, rose, and seven polyhedral compilers in sequential, run:

cd polybench
./execute_polyhedral_syn.sh -bench 2mm

More optins are available, check the first lines of the script for more information.

The results can be found in the /results/output_data repository in the docker machine, or in the docker/output_data repository of the host.