Parallelization of Heirholzer's Algorithm using OpenMP, MPI and CUDA

Graphs are used practically in many different areas, such as transportation networks, road planning, and DNA sequencing. They are effective models for representing intricate linkages and streamlining numerous operations in these fields. Breaking down jobs into smaller, concurrent processes is known...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:2025 International Conference on Next Generation Computing Systems (ICNGCS) S. 1 - 7
Hauptverfasser: M, Devadarshini, Manimegalai, R, S, Vaishnavi, S K, Noor Mahammad
Format: Tagungsbericht
Sprache:Englisch
Veröffentlicht: IEEE 21.08.2025
Schlagworte:
Circuits
Computational efficiency
CUDA
Eulerian circuit
Graph Algorithms
Graphics processing units
Heirholzer's Algorithm
Instruction sets
Memory management
MPI
OpenMP
Parallel processing
Parallelization
Scalability
System performance
Transportation
Workstations
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Graphs are used practically in many different areas, such as transportation networks, road planning, and DNA sequencing. They are effective models for representing intricate linkages and streamlining numerous operations in these fields. Breaking down jobs into smaller, concurrent processes is known as parallelization, and it is a crucial strategy for increasing the effectiveness of graph algorithms. Breaking the big-problem into smaller sub-problems drastically cuts down on calculation time and improves system performance. Several methods have been proposed to parallelize graph algorithms. Two prominent APIs, OpenMP and MPI, are prominently used for parallization. OpenMP is suitable for shared memory systems, because it enables effective parallelization by taking advantage of several CPU cores. On the other hand, MPI works well with distributed memory systems and enables the distribution of computations across numerous connected workstations. In addition, NVIDIA GPUs benefit greatly from the parallelization of graph algorithms by CUDA. A powerful option for computationally heavy jobs, GPUs excel at parallel processing because of their large number of cores. The Heirholzer algorithm, which is the focus of this work, is parallized to address problems with Eulerian circuits and pathways in graphs. This work compares the performance of OpenMP, MPI, and CUDA and it is observed that the performance of Heirholzer's algorithm is enhanced in practical applications when parallelized.
DOI:10.1109/ICNGCS64900.2025.11183273