Reproducing Experiments with Docker

For easier reproducibility, we provide a docker image to run the experiments. However, for the measurements in the papers, we run the code directly and with more data points.

Building the Docker Image

Run the following command to build the Docker image. Building the image takes about 10 minutes, as some packages (including LaTeX for the plots) have to be installed. Note that your machine needs to support the AVX2 instruction set extension to successfully run the experiments.

docker build -t mphf_experiments --no-cache .

Some compiler warnings (red) are expected when building competitors and will not prevent building the image or running the experiments. Please ignore them!

Running the Experiments

Due to the long total running time of all experiments in our paper, we provide run scripts for a slightly simplified version of the experiments. They run fewer iterations and output fewer data points.

You can modify the benchmarks scripts in scripts/dockerVolume if you want to change the number of runs or data points. This does not require the Docker image to recompile. Different experiments can be started by using the following command:

docker run --interactive --tty -v "$(pwd)/scripts/dockerVolume:/opt/dockerVolume" mphf_experiments /opt/dockerVolume/<script>.sh

<script> depends on the experiment you want to run.

Figure	Launch command	Estimated runtime
SicHash, Figure 6	/opt/dockerVolume/comparison-N.sh	20 minutes
SicHash, Figure 8	/opt/dockerVolume/pareto.sh	45 minutes
GpuRecSplit, Figure 6	/opt/dockerVolume/pareto-threads.sh	45 minutes
ShockHash, Figure 5	/opt/dockerVolume/pareto-zoomed.sh	60 minutes
Dominance Maps, Figure 8	/opt/dockerVolume/dominance-map.sh	4 hours

The resulting plots can be found in scripts/dockerVolume and have the file extension .pdf. Note again that the scripts in scripts/dockerVolume are simplified versions of the experiments. We give the full scripts in the scripts folder.