Towards automatic compilation for energy efficient iterative stencil computations
Altmetrics
Abstract
Today, energy has become a critical concern in all aspects of computing. In this thesis, we address the energy efficiency of an important class of programs called "Stencil Computations", which occur frequently in a wide variety of scientific applications. We target the compute intensive stencil computations, and seek to automatically produce codes that minimize energy consumption. Two main energy consumption contributors are addressed in our work -- dynamic memory energy and static energy -- which are proportional to the number of off-chip memory accesses and execution time separately. We first ...
(For more, see "View full record.")