Parallele Algorithmen in der Numerik
| Vorlesung: |
Dr. Sven Beuchler, Raum KG 502, Tel. 9164 Donnerstag, 13.45 – 15.15, HS 11 |
| Übung: |
Dipl.-Ing. Clemens Pechstein, Raum KG 503, Tel. 9169 Donnerstag, 15.30 – 17.00, HF 136 (ca. alle 2 Wochen) |
| Zielgruppe: | Studierende der Mathematik ab 5. Semester, interessierte Studierende |
| Art der LV: | wahlobligatorisch |
| Semesterwochenstunden: | 2 V, 1 Ü |
| Benötigtes Vorwissen: | Linear Algebra, Programmierung (C/C++ oder Fortran), Numerische Methoden für partielle Differentialgleichungen |
| Vorlesung sowie Übung werden falls notwendig auf Englisch abgehalten |
Lecture
- Introduction
- Why parallel computers?
- Classification of parallel computers due to Flynn
- Classification of parallel computers due to memory access
- Network topologies
- Performance measurement of parallel algorithms: Slide for Speed-Up
- Message passing
- Introduction
- Initialization programs in Fortran, programs in C
- Data exchange, routine for the computation of ln 2 in Fortran und C
- Parallelization of methods from Linear Algebra
- BLAS-routines, e.g., Adding two vectors, Multiplication by a scalar, scalar product, Storage of a matrix,...
- LU-Decomposition/Gauss-algorithm
- Iterative methods for the solution of systems of linear equations
- Parallel methods for the solution of systems of nonlinear equations
- Parallel methods for the solution of ordinary differential equations
- Domain decomposition for finite element methods
- Sequential algorithm
- Parallel FEM
- The parallele conjugate gradient method with Jacobi preconditioner
- Primal domain decomposition preconditioners
- The additive Schwarz method with applications
- FETI-preconditioners (if there is time)
Additional lecture: Thursday, April 10, 15.30 – 17.00, HS 11
Tutorials
| Nr. | Time | Room | Topic | Assignment |
|---|---|---|---|---|
| 1 | Th, April 03, 13.45–15.15 | HF 136 | Introduction to MPI | |
| 2 | Th, April 17, 15.30–17.00 | KG 519 | Parallelization of simple problems | |
| 3 | Mo, April 21, 12.45–14.15 | HF 136 | Graph partitioning (presentation by the PhD students) | |
| 4 | Th, May 15, 15.30–17.00 | HF 136 | Parallelization of FEM | |
| 5 | Th, June 5, 15.30–17.00 | HF 136 | Preconditioning I | |
| 6 | Th, June 12, 15.30–17.00 | HF 136 | Preconditioning II | |
| 7 | Th, June 26, 15.30–17.00 | HF 136 | Domain Decomposition |
Generation of a parallel cluster in NuMa pool in room KG 519:
- login with ssh to pool2-01.numa.uni-linz.ac.at
- login to pool2-0x with ssh
- lamboot -b -v hostfile
- The file hostfile contains the names of the computers
Further information
- To get an account at the NuMa pool contact Clemens Pechstein or directly Oliver Koch
- Official homepage of the MPI standard
- MPI-Tutorial of P. Jimack
letzte Änderung:
05.12.2013
