Loading…
The Open Computing Abstraction Layer for Parallel Complex Systems Modeling on Many-Core Systems
This article introduces OpenCAL, a new open source computing abstraction layer for multi- and many-core computing based on the Extended Cellular Automata general formalism. OpenCAL greatly simplifies the implementation of structured grid applications, contextually making parallelism transparent to t...
Saved in:
Published in: | Journal of parallel and distributed computing 2018-11, Vol.121, p.53-70 |
---|---|
Main Authors: | , , , , , , , |
Format: | Article |
Language: | English |
Subjects: | |
Citations: | Items that this one cites Items that cite this one |
Online Access: | Get full text |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Summary: | This article introduces OpenCAL, a new open source computing abstraction layer for multi- and many-core computing based on the Extended Cellular Automata general formalism. OpenCAL greatly simplifies the implementation of structured grid applications, contextually making parallelism transparent to the user. Different OpenMP- and OpenCL-based implementations have been developed, together with a preliminary MPI-based distributed memory version, which is currently under development. The system software architecture is presented and underlying data structures and algorithms described. Numerical correctness and efficiency have been assessed by considering the SciddicaT Computational Fluid Dynamics landslide simulation model as reference example. Eventually, a comprehensive study has been performed to devise the best platform for execution as a function of numerical complexity and computational domain extent. Results obtained have highlighted the OpenCAL’s potential for numerical models development and their execution on the most suitable high-performance parallel computational device.
•Domain Specific Language for structured grids modeling.•MPI/OpenMP/OpenCL implementations for execution on heterogeneous systems.•Efficient built-in data structures and parallel algorithms.•High computational performances achieved.•Possibility to devise the best parallel hardware platform for execution. |
---|---|
ISSN: | 0743-7315 1096-0848 |
DOI: | 10.1016/j.jpdc.2018.07.005 |