physnum_CUDA

C++ GPU/CUDA accelerated numerical algorithms.

This repository contains some of the scripts developed for the Physique-Numerique class held by Prof. Laurent Villard at EPFL.

The numerical algorithms are here GPU-accelerated (CPU version only has been developed in the framework of the Physique numerique class) with the fewest modifications compared to the CPU version.

Requirements

• The MonteCarlo algorithm in this project requires the boost library for random number generation
• CUDA libreries

Benchmark

To test the GPU acceleration there is the comparison with the CPU baseline. Possible by using the correct input parameter and run the algorithms on host -still relying on CUDA libreries for comparison but without the acceleration -

Format

The codes are Jupp. nootebok (.ipynb) that can be open and run on colab to exploit the acceleration.

Output

The notebooks compile and run. Finally some physical quantities are plot for comparison.