| Recommended semester: 7th - 9th semester |
| Scope and form: Lectures and project work. |
Evaluation: Approval of coursework/reports
3 reports on assignments solved by groups of 2 participants. |
| Examination: 13-scale |
| Previous course: 04310 |
| Prerequisites: Experience with programming in C (or Fortran) and numerical computations. |
| Preferred prerequisites: Computer science aspects of parallel programming. |
| Participant limitation: Max. 50 |
| Aim: To enable the students to work with large-scale problems in mathematical modelling and their implementation on high-performance computers. |
| Contents: Introduction to high-performance computers emphasizing those aspects that are important in connection with numerical computations on large data sets. Parallel computers: architecture and network. Parallel programming using MPI. Paradigms for parallel programming: decomposition, divide-and-conquer, recursive doubling, synchronization, local balancing, stopping criteria. Parallel algorithms: sorting, matrix computations and block algorithms, optimization, branch-and-bound. |
Remarks: Programming and implementation is an essential part of the course. Contact P. C. Hansen about the prerequisites.
Textbook: B. Wilkinson & M. Allen, "Parallel Programming - Techniques and Applications Using Networked Workstations and Parallel Computers", Prentice-Hall, 1999. |
| Contact: Per Christian Hansen, building 305, (+45) 4525 3097, pch@imm.dtu.dk |
| Department: 002 Informatics and Mathematical Modelling |
| Course URL: http://www.imm.dtu.dk/courses/02613 |
| Keywords: high-performance computers, large-scale problems, parallel computers, parallel algorithms, MPI |
| Updated: 10-08-2001 |
