Implementation of Private Variables Contraction Transformation of Sequential Fortran Programs for their Effective Parallelization into Computing Clusters in the SAPFOR
Abstract:
The process of automated parallelization of programs can be significantly complicated due to their structure and optimization for sequential execution. Because of this, the resulting parallel version may be ineffective, and in some cases parallelization turns out to be completely impossible. Transformations of the source code of sequential programs help to solve these problems. This article discusses the development of an algorithm for transformation of sequential Fortran programs called “removing of private variables” and its implementation in the SAPFOR automated parallelization system (System FOR Automated Parallelization). The application of the implemented transformations in the SAPFOR system is demonstrated on four application programs included in the NAS Parallel Benchmarks package.
Keywords:
SAPFOR (System FOR Automated Parallelization); parallelization automation for clusters; transformation automation; parallel computing; DVM (Distributed Virtual Memory); GPU clusters.
Received: 01.10.2024 Revised: 14.01.2025
Document Type:
Article
UDC:
04.272
Language: Russian
Citation:
A. S. Kolganov, G. D. Gusev, “Implementation of Private Variables Contraction Transformation of Sequential Fortran Programs for their Effective Parallelization into Computing Clusters in the SAPFOR”, Num. Meth. Prog., 26:1 (2025), 58–84