K-Means in Parallel

Description

This is an implementation of the well known clustering algorithm K-Means. The implementation was done in C++ 17 using OpenMP 5.1. The IDE used for the project was Visual Studio 2019 and the compiler was Intel Compiler 19.1. The Effective Logical Core Utiliation percentage was 96.1 %.

Installation

To test the code in Linux, execute the commands:

1. git clone https://github.com/andreasceid/k-means-parallel.git
2. cd k-means-parallel
3. chmod u+x disp_system_thread_count-ubu20.sh
4. ./disp_system_thread_count-ubu20.sh

Write down the number on the output. We will use it later

5. cd k-means-parallel

Edit Common.h, at line 27, and set variable NUM_THR equal to the number we noted in step 5

6. make
7. ./kmeans

Core Utilization

The Intel VTune Profiler Performance Snapshot results were:

Scaling

The program was also tested with different number of available threads, and the scaling is pictured below.

Additional Content

There are 2 scritps that detect system logical core number:

• For Linux (Tested on Ubuntu 20.04 64 bit): Bash Script disp_system_thread_count-ubu20.sh
• For Windows (Tested on Windows 10 Version 20H2 64 bit): Powershell Script disp_system_thread_count-win64.ps1

There is a PDF project report in Greek.