Energy-Efficient Parallel Computing2013HT
Presentation session with opposition and written summary.
Research-oriented individual case studies (including programming and
experimental evaluation), short presentation and term paper.
Graduate (CUGS, CIS, ...) students interested in the areas of computing systems, parallel computing, software engineering, optimization, scheduling theory, compiler construction, or algorithms and complexity.
The course was last given
This is a new course.
The course gives an introduction to energy modeling and optimization
in modern computing systems, with an emphasis on scheduling, speed
scaling and parallelism. The participants will actively contribute by
literature study/presentation and own experimental work.
A course in parallel programming (e.g., TDDD56 or TDDC78).
Some background in computer architecture.
We also recommend a course in Complexity theory and analysis of algorithms such as Computation II.
Programming in C and some familiarity with Linux (or similar OS) is necessary for the case study work.
The course introduces energy modeling concepts of modern computing systems with
an emphasis on parallel computing and frequency scaling.
A main focus is on recent results in scheduling theory that involves frequency scaling, temporary shutdown of resources, and varying the degree of parallelism for parallel tasks, but also on the energy cost of memory access and communication.
Scheduling algorithms and other optimization techniques for energy efficiency from the literature will be discussed and applied in own case studies on concrete systems and example programs.
TBA (late ht1 or ht2).
The course will follow on DF21500 Multicore Computing.
To be announced on the course homepage.
UPG1: Paper presentation, opposition, and written summary, 1.5hp.
UPG2: Energy modeling and optimization case study, short presentation and term paper, 1.5hp.
3hp if all examination moments are fulfilled.
Page responsible: Director of Graduate Studies