Efficient Acceleration in Solving the 2D Neutron Diffusion Equation with CUDA: Exploring the Collaborative Practicality of Colab
This paper explores an approach to accelerate the finite difference method applied to solving the two-dimensional (2D) neutron diffusion equation for two energy groups (2G) independent of time. The main innovation lies in the implementation of a performance optimization method, emphasizing the prac...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Brazilian Radiation Protection Society (Sociedade Brasileira de Proteção Radiológica, SBPR)
2025-06-01
|
| Series: | Brazilian Journal of Radiation Sciences |
| Subjects: | |
| Online Access: | https://bjrs.org.br/revista/index.php/REVISTA/article/view/2498 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | This paper explores an approach to accelerate the finite difference method applied to solving the two-dimensional (2D) neutron diffusion equation for two energy groups (2G) independent of time. The main innovation lies in the implementation of a performance optimization method, emphasizing the practicality of development in Python using direct browser collaboration through Google Colaboratory (Colab). Utilizing CUDA (Compute Unified Device Architecture) for GPU acceleration, we achieve significant computational performance improvements. The study compares Python implementations using CuPy and NumPy libraries with traditional FORTRAN implementations utilizing the LAPACK library, highlighting the efficiency and precision of GPU-accelerated calculations. Results show that Python with CuPy significantly outperforms NumPy, both in a Colab environment and on a personal desktop computer. This demonstrates the practicality of cloud-based solutions for intensive computations, as the ability to run code directly in the browser through Colab eliminates the need for extensive local hardware resources. The results emphasize the convenience of executing complex simulations without relying on physical computers, promoting greater flexibility and accessibility in computational research. All computational codes are available on GitHub for transparency and reproducibility.
|
|---|---|
| ISSN: | 2319-0612 |