Computer File

Parallel Flood Simulations for Wet-Dry Problems Using Dynamic Load Balancing Concept

In this paper, a shared-memory parallel simulation of flood modeling is presented. The model used has second-order spatial and temporal accuracy, where the Monotonic Upwind Scheme for Conservation Laws (MUSCL) method is applied for spatial discretization and the Runge-Kutta second-order method is employed for temporal discretization. A cell-centered finite-volume model is used and solved in an edge-based data structure. The model is well-balanced and able to efficiently simulate flood cases on complex topography with wet–dry
problems. A cell–edge reordering strategy is designed to ease vectorization and parallelization of the code. To tackle load imbalances among threads due to wet–dry problems, a novel weighted-dynamic load balancing is proposed. The model shows accurate results, and the strategy proposed shows very good parallel efficiencies for problems of different sizes (up to 6.4 million cells=12.8 million edges) on varying numbers of cores (up to 64 cores). As such, this load balancing technique could become a promising strategy for efficient parallel simulations of real flood cases.

Ketersediaan

Informasi Detail

Judul Seri

JOURNAL OF COMPUTING IN CIVIL ENGINEERING; Vol.033 No.3 2019

No. Panggil

-

Penerbit

Reston : American Society of Civil Engineers., 2019

Deskripsi Fisik

p. 04019013-1-04019013-18

Bahasa

English

ISBN/ISSN

-

Klasifikasi

-

Tipe Isi

-

Tipe Media

-

Tipe Pembawa

-

Edisi

-

Subjek

-

Info Detail Spesifik

-

Pernyataan Tanggungjawab

Bobby Minola Ginting, Ralf-Peter Mundani

Versi lain/terkait

Lampiran Berkas

Komentar