Applied High Performance Computing Using R

Stefan Theußl

Publikation: AbschlussarbeitDiplomarbeit

Abstract

In the 1990s the Beowulf project smoothed to way for massively parallel
computing as access to parallel computing power became aordable for
research institutions and the industry. But the massive breakthrough of
parallel computing has still not occurred. This is because two things were
missing: low cost parallel computers and simple to use parallel programming
models. However, with the introduction of multicore processors for mainstream
computers and implicit parallel programming models like OpenMP
a fundamental change of the way developers design and build their software
applications is taken place|a change towards parallel computing.
This thesis gives an overview of the eld of high performance computing
with a special focus on parallel computing in connection with the R environment
for statistical computing and graphics. Furthermore, an introduction
to parallel computing using various extensions to R is given.
The major contribution of this thesis is the package called paRc, which
contains an interface to OpenMP and provides a benchmark environment to
compare various parallel programming models like MPI or PVM with each
other as well as with highly optimized (BLAS) libraries. The dot product
matrix multiplication was chosen as the benchmark task as it is a prime
example in parallel computing.
Eventually a case study in computational nance is presented in this
thesis. It deals with the pricing of derivatives (European call options) using
parallel Monte Carlo simulation.
Kur
OriginalspracheEnglisch
Gradverleihende Hochschule
  • Wirtschaftsuniversität Wien
PublikationsstatusVeröffentlicht - 1 Apr. 2007

Österreichische Systematik der Wissenschaftszweige (ÖFOS)

  • 102023 Supercomputing

Zitat