Postgraduate Course: Design and Analysis of Parallel Algorithms (INFR11028)
Course Outline
| School | School of Informatics |
College | College of Science and Engineering |
| Credit level (Normal year taken) | SCQF Level 11 (Postgraduate) |
Availability | Available to all students |
| SCQF Credits | 10 |
ECTS Credits | 5 |
| Summary | This module introduces the design principles and analysis techniques which enable the creation of efficient, scalable and portable algorithms for parallel computers. Concrete examples will span a range of application areas and architectural models seeking wherever possible to exploit commonality through appropriate abstraction. |
| Course description |
* Introduction: Conceptual frameworks for parallelism, message passing, shared address space, PRAM. Cost models for parallel algorithms. Cost efficiency and scalability. Inter-model emulation. Simple examples.
* Problem solving strategies: Embarrassing parallelism, divide & conquer, pipelining, step-by-step parallelisation. Amdahl's Law.
* Useful primitives: Collective communications, reduction, prefix.
* Algorithms in selected problem areas, for example: Sorting (bitonic mergesort, hyperquicksort). Matrix oriented algorithms (multiplication, solving linear systems). Graph algorithms (spanning trees, single source & all-to-all shortest paths).
Relevant QAA Computing Curriculum Sections: Concurrency and Parallelism, Data Structures and Algorithms
|
Entry Requirements (not applicable to Visiting Students)
| Pre-requisites |
|
Co-requisites | |
| Prohibited Combinations | |
Other requirements | This course is open to all Informatics students including those on joint degrees. For external students where this course is not listed in your DPT, please seek special permission from the course organiser.
The following mathematics skills are also assumed:
- familiarity with binary numbers, conversion to/from decimal
- simple facts about and manipulation of logs and exponentials
- O notation, proper definition and intuitive feel
- summation of simple arithmetic and geometric series
- matrix multiplication and Gaussian Elimination
- very simple, neat recurrences, cf. easiest ones in MforInf2
The specific algorithms are not important - what matters is experience of working at this level of abstraction. |
Information for Visiting Students
| Pre-requisites | None |
Course Delivery Information
| Not being delivered |
Learning Outcomes
1 - define the structure of, and cost models associated with, the PRAM, mesh and hypercube models of parallel computation.
2 - define the metrics of cost, speed-up and efficiency and use these as conceptual tools with which to analyse and discriminate between alternative candidate parallel algorithms for given problems. They will be able to demonstrate, by the use of appropriately chosen examples, the importance of scalability in parallel algorithm design.
3 - explain and, with appropriate use of diagrams, sketch the structure and operation of well known parallel algorithms in a range of application areas, including sorting, matrix and graph based problems.
4 - apply a range of parallel algorithm design techniques (including divide-and-conquer and pipelining) to previously unseen problems, in order to create new parallel algorithms, which they will be able to describe using an informal mix of pseudocode, textual explanation and diagrams.
|
Reading List
| V. Kumar, A. Grama, A. Gupta & G. Karypis, 'Introduction to Parallel Computing: Design and Analysis of Algorithms', (2nd Ed), 2003. |
Contacts
| Course organiser | Dr Iain Murray
Tel: (0131 6)51 9078
Email: |
Course secretary | Ms Katey Lee
Tel: (0131 6)50 2701
Email: |
|
|
University Homepage