TGA Project: Parallel Merge Sort (OpenCL vs CUDA)

Introduction

This project explores different implementations of the Merge Sort algorithm, combined with Insertion Sort (Also known was MultiSort), to enhance performance.

The main objectives were:

• Implement and compare: Develop a parallel version of the MultiSort algorithm in CUDA and compare it with the sequential version.
• Explore PathMerge: Implement and evaluate a lesser-known parallel variant of the Merge Sort algorithm called PathMerge.
• Evaluate OpenCL vs. CUDA: Test the OpenCL API and conduct a comparative analysis against CUDA to understand their respective strengths and limitations.

Execution

For detailed instructions on running the experiments, see experiments/README.md.

Team

• Alex Herrero Bravo
• Walter J. Troiani Vargas

Acknowledgements

Many thanks to Agustín, an outstanding wise folk which taught us many life-changing lessons and that NVIDIA is investing a lot of cash... Finally, thanks to Daniel Jimenez for never giving up on us (PRINGAOOO).