skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: A communication scheme for the distrubted execution of loop nests with while loops

Abstract

The mathematical model for the parallelization, or {open_quotes}space-time mapping,{close_quotes} of loop nests is the polyhedron model. The presence of while loops in the nest complicates matters for two reasons: (1) the parallelized loop nest does not correspond to a polyhedron but instead to a subset that resembles a (multi-dimensional) comb and (2) it is not clear when the entire loop nest has terminated. We describe a communication scheme which can deal with both problems and which can be added to the parallel target loop nest by a compiler.

Authors:
;  [1]
  1. Universitaet Passau (Germany)
Publication Date:
Sponsoring Org.:
USDOE
OSTI Identifier:
191987
Resource Type:
Journal Article
Resource Relation:
Journal Name: International Journal of Parallel Programming; Journal Volume: 23; Journal Issue: 5; Other Information: PBD: Oct 1995
Country of Publication:
United States
Language:
English
Subject:
99 MATHEMATICS, COMPUTERS, INFORMATION SCIENCE, MANAGEMENT, LAW, MISCELLANEOUS; PARALLEL PROCESSING; MATHEMATICAL MODELS; COMMUNICATIONS; SPACE-TIME

Citation Formats

Griebl, M., and Lengauer, C.. A communication scheme for the distrubted execution of loop nests with while loops. United States: N. p., 1995. Web. doi:10.1007/BF02577774.
Griebl, M., & Lengauer, C.. A communication scheme for the distrubted execution of loop nests with while loops. United States. doi:10.1007/BF02577774.
Griebl, M., and Lengauer, C.. Sun . "A communication scheme for the distrubted execution of loop nests with while loops". United States. doi:10.1007/BF02577774.
@article{osti_191987,
title = {A communication scheme for the distrubted execution of loop nests with while loops},
author = {Griebl, M. and Lengauer, C.},
abstractNote = {The mathematical model for the parallelization, or {open_quotes}space-time mapping,{close_quotes} of loop nests is the polyhedron model. The presence of while loops in the nest complicates matters for two reasons: (1) the parallelized loop nest does not correspond to a polyhedron but instead to a subset that resembles a (multi-dimensional) comb and (2) it is not clear when the entire loop nest has terminated. We describe a communication scheme which can deal with both problems and which can be added to the parallel target loop nest by a compiler.},
doi = {10.1007/BF02577774},
journal = {International Journal of Parallel Programming},
number = 5,
volume = 23,
place = {United States},
year = {Sun Oct 01 00:00:00 EDT 1995},
month = {Sun Oct 01 00:00:00 EDT 1995}
}
  • Automatic parallelization of imperative sequential programs has focused on nests of for-loops. The most recent of them consist in finding an affine mapping with respect to the loop indices to simultaneously capture the temporal and spatial properties of the parallelized program. Such a mapping is usually called a {open_quotes}space-time transformation.{close_quotes} This work describes an extension of these techniques to while-loops using speculative execution. We show that space-time transformations are a good framework for summing up previous restructuration techniques of while-loop, such as pipelining. Moreover, we show that these transformations can be derived and applied automatically.
  • The authors examine the effectiveness of optimizations aimed a allowing a distributed machine to efficiently compute inner loops over globally defined data structures. Their optimizations are specifically targeted toward loops in which some array references are made through a level of indirection. Unstructured mesh codes and sparse matrix solvers are examples of programs with kernels of this sort. Experimental data that quantify the performance obtainable using the methods discussed are included.
  • This article proposes a method of isolating independent program branches based on finding the resulting iterations and then solving the problem of minimizing the execution time for a given number of processors or minimizing the number of processors for a fixed execution time. The distribution of iterations can be organized in a more regular pattern given the interconnection through variables with linear indices of a certain simple form. Each family of iterations connected informationally can be seen as a pyramid in the space of iterations. Paralleling in programs involves repeating areas of calculation by using loops. An attempt is mademore » to solve the problem of paralleling with complete or at least maximum possible independence in the execution of parallel branches, even if it involves the duplication of certain computation fragments in different branches.« less
  • The problem of parallel execution of DO loops is addressed. Various levels of parallelism are mentioned and a number of architectures for these different levels of parallelism are reviewed. It is then pointed out that due to the deficiencies in the existing architectures only restricted types of DO loops can be run with these known systems. Therefore, additional requirements for running a variety of DO loops are studied. Finally, a feasible parallel machine of MSIMD type is presented. In this machine a modified multistage interconnection network is employed which, together with other system components, makes the machine suitable for runningmore » a variety of DO loops. Some special features of the machine when executing DO loops are stated and the process of running DO loops on this machine is described.« less
  • We present in analytic form the matching conditions for the strong coupling constant {alpha}{sup (n{sub f})}{sub s }({mu}) at the flavor thresholds to three loops in the modified minimal-subtraction scheme. Taking into account the recently calculated coefficient {beta}{sub 3} of the Callan-Symanzik beta function of quantum chromodynamics, we thus derive a four-loop formula for {alpha}{sup (n{sub f})}{sub s }({mu}) together with appropriate relationships between the asymptotic scale parameters {Lambda}{sup (n{sub f})} for different numbers of flavors n{sub f} . {copyright} {ital 1997} {ital The American Physical Society}