Szczegóły publikacji
Opis bibliograficzny
Concurrency of the exponential integrators for the glioblastoma brain tumor simulations / Magdalena Pabisz, Askold VILKHA, Dominika Ciupek, Maciej WOŹNIAK, Maciej PASZYŃSKI // Computers and Mathematics with Applications ; ISSN 0898-1221. — 2025 — vol. 196, s. 13–29. — Bibliogr. s. 28–29, Abstr. — Publikacja dostępna online od: 2025-07-14
Autorzy (5)
Słowa kluczowe
Dane bibliometryczne
| ID BaDAP | 161388 |
|---|---|
| Data dodania do BaDAP | 2025-07-30 |
| Tekst źródłowy | URL |
| DOI | 10.1016/j.camwa.2025.07.003 |
| Rok publikacji | 2025 |
| Typ publikacji | artykuł w czasopiśmie |
| Otwarty dostęp |  |
| Creative Commons | |
| Czasopismo/seria | Computers & Mathematics with Applications |
Abstract
We employ the trace theory as a tool to investigate the concurrency of the novel implementation of the exponential integrators method using the multinomial theorem. The exponential integrators code is applied for the glioblastoma brain tumor simulations. The paper introduces a solid theoretical background for trace theory-based concurrency analysis and uses an important example of brain tumor simulations to validate its applicability. We perform our simulations on available public MRI scan databases. The spatial discretization uses the finite difference method. The temporal discretization uses the exponential integrators method. We employ the trace theory to identify basic undividable tasks and their dependency relation. Based on the coloring of the task dependency graph, we localize sets of tasks that can be executed fully in parallel. From the results of the trace theory-based concurrency analysis, we designed an efficient MATLAB code parallelized for the GPGPU. We test the implementation by running 1000 time steps of the glioblastoma brain tumor simulation over the three-dimensional brain data with 194 x 248 x 256 pixels. We measure the scalability of the code, reaching 32 times speedup, using 128 x 128 x 128 computational mesh that can be computed on RTX 2080 Super GPU card equipped with 3072 processing cores and 8 GB of memory.
