Результаты исследований: Публикации в книгах, отчётах, сборниках, трудах конференций › статья в сборнике материалов конференции › научная › Рецензирование
On the Parallel Strategies in Mathematical Modeling. / Il’in, Valery.
Parallel Computational Technologies - 11th International Conference, PCT 2017, Revised Selected Papers. ред. / L Sokolinsky; M Zymbler. Том 753 Springer-Verlag GmbH and Co. KG, 2017. стр. 73-85 (Communications in Computer and Information Science; Том 753).Результаты исследований: Публикации в книгах, отчётах, сборниках, трудах конференций › статья в сборнике материалов конференции › научная › Рецензирование
}
TY - GEN
T1 - On the Parallel Strategies in Mathematical Modeling
AU - Il’in, Valery
PY - 2017/1/1
Y1 - 2017/1/1
N2 - The article considers parallel strategies and tactics at different stages of mathematical modeling. These technological steps include geometrical and functional modeling, discretization and approximation, algebraic solvers and optimization methods for inverse problems, postprocessing and visualization of numerical results, as well as decision-making systems. Scalable parallelism can be provided by combined application of MPI tools, multi-thread computing, vectorization, and the use of graphics accelerators. The general method to achieve high-performance computing consists in minimizing data communications, which are the most time and energy consuming. The construction of efficient parallel algorithms and code optimization is based on various approaches at different levels of computational schemes. The implementation of the biggest interdisciplinary direct and inverse problems in cloud computing technologies is considered. The corresponding applied software with a long life cycle is represented as integrated environment oriented to large groups of end users.
AB - The article considers parallel strategies and tactics at different stages of mathematical modeling. These technological steps include geometrical and functional modeling, discretization and approximation, algebraic solvers and optimization methods for inverse problems, postprocessing and visualization of numerical results, as well as decision-making systems. Scalable parallelism can be provided by combined application of MPI tools, multi-thread computing, vectorization, and the use of graphics accelerators. The general method to achieve high-performance computing consists in minimizing data communications, which are the most time and energy consuming. The construction of efficient parallel algorithms and code optimization is based on various approaches at different levels of computational schemes. The implementation of the biggest interdisciplinary direct and inverse problems in cloud computing technologies is considered. The corresponding applied software with a long life cycle is represented as integrated environment oriented to large groups of end users.
KW - Accelerators
KW - Communications
KW - Domain decomposition
KW - Exchange buffers
KW - Hierarchical memory
KW - Multi-thread computing
KW - Runtime
KW - Scalable parallelism
KW - Speedup
KW - Vectorization
UR - http://www.scopus.com/inward/record.url?scp=85032509823&partnerID=8YFLogxK
U2 - 10.1007/978-3-319-67035-5_6
DO - 10.1007/978-3-319-67035-5_6
M3 - Conference contribution
AN - SCOPUS:85032509823
SN - 9783319670348
VL - 753
T3 - Communications in Computer and Information Science
SP - 73
EP - 85
BT - Parallel Computational Technologies - 11th International Conference, PCT 2017, Revised Selected Papers
A2 - Sokolinsky, L
A2 - Zymbler, M
PB - Springer-Verlag GmbH and Co. KG
T2 - 11th International Conference on Parallel Computational Technologies, PCT 2017
Y2 - 3 April 2017 through 7 April 2017
ER -
ID: 9753964