Synchronization Minimization in a SPMD Execution Model

Synchronization Minimization in a SPMD Execution Model
复制标题

SPMD 执行模型中的同步最小化

DOI:
--
复制
发表时间:
1995
期刊:
J. Parallel Distributed Comput.
影响因子:
--
通讯作者:
F. Bodin
F. Bodin
中科院分区:
--
文献类型:
--
作者:
M. O’Boyle;L. Kervella;F. Bodin

文献摘要

被引文献

相似文献

本文提出了一种算法,同步放置时,使用SPMD执行模型,其中同步只在存在跨处理器的数据依赖。在本文中,我们研究了两种调度技术,基于循环和基于数据,这两种技术都使用SPMD模型。使用编译过程中先前阶段的调度信息,提出了一种确定潜在跨处理器数据依赖性的新技术。给定必须满足的跨处理器数据依赖的最小数量,使用新的优化,以便最小化满足它们所需的同步点的数量。该算法已成功地实现在一个实验编译器。初步的实验数据表明,这种技术是非常有效的,优于现有的方法。
This paper presents an algorithm for synchronization placement when using a SPMD execution model, where synchronizations are enforced only when there exists a cross-processor data dependence. In this paper, we investigate two scheduling techniques, loop-based and data-based, both of which use a SPMD model. Using scheduling information from previous stages in the compilation process, a new technique to determine potential cross-processor data dependences is presented. Given the minimum number of cross-processor data dependences that must be satisfied, a new optimization is used so as to minimize the number of synchronization points needed to satisfy them. This algorithm has been successfully implemented in an experimental compiler. Initial experimental data show this technique to be very effective, outperforming existing methods.