Dossier: SimRace 2015: Numerical Methods and High Performance Computing for Industrial Fluid Flows
Open Access

Table 1.

A short synopsis of programming paradigms

Programming paradigm Description
Sequential implementation Direct mapping of algorithm description to code, with sequential linear algebra primitives applied on large vectors (simple loops), and sparse-matrix times vector operations.
Multi-threaded/Multi-core implementation Linear algebra primitives implemented with Linux POSIX threads.
GPU implementation Memory for iteration vectors, sparse matrix was allocated in the GPU accelerator, and linear algebra primitives done in the GPU, with a program on CPU acting as master.
Cluster-programming Memory for iteration vectors is distributed in different processes spanning several machines, sparse matrix lines bands are also distributed, and communication is implemented between processes to consolidate needed data for each process to proceed.

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.