This project is part of the Parallel Programming course (2023-24) and focuses on implementing the K-means clustering algorithm for pixel clustering in bitmap images using parallel programming techniques.

• Source Code: Implementation of the K-means algorithm in C.
• Documentation: Detailed explanations and instructions for compiling and running the code.
• Performance Analysis: Tools and scripts for analyzing the performance of the parallel implementations.

1. Clone the repository:

   git clone https://github.com/luciarevaliente/Kmeans_Parallel_Programming_using_OpenMP_OpenACC_CUDA_MPI.git
   cd Kmeans_Parallel_Programming_using_OpenMP_OpenACC_CUDA_MPI

2. Compile the code:

   make

3. Run the K-means algorithm

Final project for the Parallel Programming course (2023-24). The objective is to design and implement parallel versions of the K-means clustering algorithm to reduce the number of colors in a bitmap image. The project involves using different parallel programming approaches, including shared memory (OpenMP), accelerators (OpenACC/CUDA), and distributed memory (MPI).

• K-means Algorithm: An unsupervised machine learning algorithm used for clustering data points.
• Pixel Clustering: Application of K-means to group pixels in an image based on their RGB values.
• Parallel Implementations: Development of parallel versions using OpenMP, OpenACC, CUDA, and MPI to improve performance.
• Performance Analysis: Evaluation of the speedup and efficiency of the parallel implementations.

• OpenMP: Utilizes shared memory parallelism to distribute the workload across multiple CPU cores. This involves parallelizing loops and tasks to minimize execution time.
• OpenACC: Provides a high-level approach to parallelization, allowing code to be offloaded to GPUs with minimal changes.
• CUDA: Offers fine-grained control over GPU execution, enabling significant acceleration of compute-intensive parts of the algorithm.
• MPI: Uses distributed memory parallelism to enable communication between multiple processors across different nodes, allowing the algorithm to scale efficiently on large clusters.

This project is part of an academic course and does not accept external contributions.

This project does not have a specific license and is for educational purposes only.