OpenMP - Automatic Parallelization Tools: An Empirical Comparative Evaluation
Today, multicore
design has become the trend of
enhancing the processors performance, and most
industries have been considering multicore
as the
future of development. Thus, a programmer or a
compiler explicitly parallelizes the software, which
is the key to enhance the performance on multicore
design. Nevertheless, currently, needs an indepth
knowledge of both software and hardware design to
develop parallel applications. Automatic
parallelization is one of the approaches aiming at a
better and easier use of parallel computers. In
recent years, several research autoparallelization
tools appeared. However, the automatic
parallelization is yet to become a widely adopted
industrial practice. This paper presents an empirical
comparison between three research tools, namely
CETUS, PLUTO and GASPARD. Indeed, we discuss the
success of these tools to automatically generate
OpenMP parallel codes from serial C codes and
compare them using known benchmark C workloads
and some evaluation metrics.
Keywords: Multi-core, automatic parallelization, OpenMP, metrics, parallel benchmarks.
Download Full-Text
ABOUT THE AUTHORS
Emna Kallel Laadhar
enis
Yassine Aoudni
enis
Emna Kallel Laadhar
enis
Yassine Aoudni
enis
