Table of Contents
A Comparison of MPI and OpenMP Implementations of a Finite Element Analysis Code
Authors
Acknowledgements
Overview
Contents
Introduction
The Problems
Element-by-Element Approach
CERN Tunnels
The Codes
Fortran 90 for PCG
Element Level Approach
System Used
System Used
Reproducible Timings?
Implementation with MPI
Parallelism in Code
Data Distribution
MPI Results
MPI results
Implementation with OpenMP
OpenMP
Compiler Options of particular interest
Using BLAS Routines
Matrix Multiplication Section
Matrix Multiplication
Dgemm versus OpenMP
Data placement - note
Gather no problem
scatter
Parallelising scatter
Other parts of the code
OpenMP Results
OpenMP comments
Comparison between MPI and OpenMP Implementations
Performance
Unfair Comparison?
For MPI
For OpenMP
Conclusions
Conclusions |
Author: Michael Bane
Presenter: Terry Hewitt |