Iteration Space Tiling for Distributed Memory Machines

Document Type

Article

Publication Date

1-1-1992

Abstract

This paper addresses the problem of compiling nested loops for distributed memory machines. The relatively high communication start-up costs in these machines renders frequent communication very expensive. Motivated by this, we present a method for aggregating a number of loop iterations into tiles where the tiles execute atomically. Since synchronization is not allowed during the execution of a tile, partitioning the iteration space into tiles must not result in deadlock. We present an approach to determine the shape, size, allocation and scheduling of tiles for efficient parallel execution of 2-level tightly nested loops on distributed memory machines with given execution and communication costs. © 1992, Elsevier B.V.

Publication Source (Journal or Book title)

Advances in Parallel Computing

First Page

255

Last Page

270

This document is currently not available here.

Share

COinS