K-means parallel computing in CUDA

This project was an assignment for the High Performance Computing class at University of Siena. You can look at the final report for more details.

How to run this code

Before trying to execute both sequential and parallel version of k-means algorithm it is important to extract datapoints.zip in the same folder of the codes.

The parameters will be initialized according to user choices after the execution start.

If you want to visualize the results you need to create in the same folder of the code a directory named outdir, execute your simulation and then take the file clusters.csv from outdir, which must be used in the python script cluster_displayer.py.

1. sequential compiler line on linux:

   g++ -o kmeans_sequential kmeans_sequential.cpp
   

2. parallel compiler line on linux:

   nvcc kmeans_parallel.cu -o kmeans_parallel