Optimizations for Power

Participants: Maria Garzaran, David Padua, Bruno Virlet, Xing Zhou

Executive Summary

The objective of this project is to study the effect on execution time and power of scheduling parts of computations on different devices such as CPUs, GPUs, DSPs, etc. In this study, we will use analytical models that contain program trace information as parameters. Once effect of scheduling is better understood, autotuning and/or compiler techniques will be developed to optimize for power. The focus will be video post processing mainly in the context of augmented reality applications.

Goals - Extended Description

Deliverables: By the end of the year, we will deliver a study of the notations needed for maximum flexility in scheduling codes as well as measurement of the effect of scheduling on power consumption for a video post processing application.

Collaborators: Jean Pierre Giacalone and Robert Kuhn (Intel)

Autotuning

Gluon: Interface for
Trusted Programming

Interactive Porting

Libraries

Optimizations for Power

Refactoring

Safe Parallel Prog. Languages

Scheduling

Verification & Validation