The paper presents a source-to-source compiler, TRACO, for automatic extraction of both coarse- and fine-grained parallelism available in C/C++ loops. Parallelization techniques implemented in TRACO are based on the transitive closure of a relation describing all the dependences in a loop. Coarse- and fine-grained parallelism is represented with synchronization-free slices (space partitions) and a legal loop statement instance schedule (time partitions), respectively. TRACO enables also applying scalar and array variable privatization as well as parallel reduction. On its output, TRACO produces compilable parallel OpenMP C/C++ and/or OpenACC C/C++ code. The effectiveness of TRACO, efficiency of parallel code produced by TRACO, and the time of parallel code production are evaluated by means of the NAS Parallel Benchmark and Polyhedral Benchmark suites. These features of TRACO are compared with closely related compilers such as ICC, Pluto, Par4All, and Cetus. Feature work is outlined.
Publié le : 2017-02-13
Classification:
Parallel and Distributed Computing,
Source-to-source parallelizing compiler, loop parallelization, iteration space slicing, fine- and coarse-grained parallelism, free scheduling, transitive closure,
68N20, 65Y05, 52Bxx, 97E60, 05-XX
@article{cai2315,
author = {Marek Palkowski; West Pomeranian University of Technology, Szczecin and Wlodzimierz Bielecki; West Pomeranian University of Technology, Szczecin},
title = {TRACO: Source-to-Source Parallelizing Compiler},
journal = {Computing and Informatics},
volume = {35},
number = {4},
year = {2017},
language = {en},
url = {http://dml.mathdoc.fr/item/cai2315}
}
Marek Palkowski; West Pomeranian University of Technology, Szczecin; Wlodzimierz Bielecki; West Pomeranian University of Technology, Szczecin. TRACO: Source-to-Source Parallelizing Compiler. Computing and Informatics, Tome 35 (2017) no. 4, . http://gdmltest.u-ga.fr/item/cai2315/