Special Seminar: Prof. X. Sharon Hu, University of Notre Dame: "Temperature-Aware Scheduling"

Wednesday, October 7, 2009
2:00 p.m.
AV Williams 2328
Peter Petrov
ppetrov@umd.edu

Speaker: Professor X. Sharon Hu, University of Notre Dame

Talk title: "Temperature-Aware Scheduling and Assignment for Real-Time Applications on MPSoCs"

Location: AVW 2328

Time: 2pm, Oct. 7

Abstract:

As CMOS transistors continue their scaling downward, multiprocessor-system-on-chips (MPSoCs) will likely be used in many applications, as they often permit higher performance and better power efficiency than uniprocessor architectures. However, MPSoCs can have high power density and temperature, which degrade reliability and increase packaging and cooling cost. Techniques such as processor throttling are being considered to prevent the chip from reaching an unsafe temperature at runtime. However, runtime throttling can have undesirable performance consequences. The problem is particularly acute for real-time applications, which demand predictable timing behavior.

This talk discusses two solutions to the problem of assignment and scheduling of real-time tasks being executed on a MPSoC in order to minimize the chip's peak temperature. The first solution is an exact solution based on a mixed-integer linear programming (MILP) formulation. The formulation considers both spatial and temporal variations in power density, and relies on a phased steady state thermal analysis technique that is directly integrated within the MILP formulation. The second solution, targeting at larger problem instances, is a heuristic in which the actual method of computing the temperature profile can be adjusted for different physical parameters. Simulation results show that the solutions outperform existing approaches.

Faculty Host:

Dr. Peter Petrov

Audience: Clark School  Graduate  Undergraduate  Faculty 

remind we with google calendar

 

December 2025

SU MO TU WE TH FR SA
30 1 2 3 4 5 6
7 8 9 10 11 12 13
14 15 16 17 18 19 20
21 22 23 24 25 26 27
28 29 30 31 1 2 3
Submit an Event