IMPLEMENTATION OF PARALLEL ALGORITHM USING MPI AND CONCURRENT.FUTURES FOR CALCULATING PIJ IN THE COLLISION PROBABILITY METHOD

• Main Author: Abrar, Ghulam
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/81484
• Institution: Institut Teknologi Bandung

There are two methods for solving the neutron transport equation: the stochastic method (Monte Carlo) and the deterministic method. The Monte Carlo method is more popular for use in cases of complex geometry because of its good accuracy, ease of use, and ease of parallelization. However, to achieve the desired accuracy, the Monte Carlo method requires very large samples, which slows down computing time. On the other hand, deterministic methods are difficult to work with for complex geometries, but their computation is faster. Deterministic methods require many assumptions and simplifications, so their accuracy is not as good as that of the Monte Carlo method. One of the deterministic methods that is good for handling complex geometries is the Collision Probability Method (CPM). This method is based on the integral transport equation. More research related to CPM has been carried out on the development of numerical methods. In fact, along with the development of supercomputers, in the last two decades code development has focused more on using parallel algorithms. In this research, a parallel algorithm was implemented for calculating the ???????????? in CPM using the Python programming language. Message Passing Interface (MPI) and concurrent.futures are two modules used for parallelization. Both modules give the same performance on a single CPU. The program can be run in parallel with several times the speedup over serial computation. The highest speedup achieved was 8.18 times in the case with 40 cells.