Re: [R] R benchmarking program

From: Prof Brian Ripley <ripley_at_stats.ox.ac.uk>
Date: Tue, 13 May 2008 17:16:37 +0100 (BST)

The best benchmark is a calculation you really want to do. For most people matrix algebra is an insignificant proportion of what they do in R. A few need complex matrix arithmetic (where compilers and BLAS differ a lot in speed). So there is no universal benchmark.

Also, the reference BLAS supplied with R is usually more accurate than the optimized BLAS, and many users will prefer accuracy to speed. There are packages that will not complete their tests with an optimized BLAS.

On Tue, 13 May 2008, Baker D.J. wrote:

> Hi All,
>
> I've just rebuild the latest R with the Goto BLAS on our new Intel quad
> core machines.

I do wonder why -- as the manual says, the easiest and most efficient way to do this is to link the Goto BLAS to R_HOME/lib/libRblas.so (assuming this is Linux). No rebuilding is needed.

> I did a few basic matrix calculations, and I was very
> impressed by the performance I saw. I wonder if anyone has a more
> rigorous benchmarking program for R. I downloaded a old R
> test/benchmarking program (see below), and this didn't work with the
> current R, and so I wondered if anyone could please direct me to a more
> recent program that does a good all round test of R.
>
> Regards -- David.
>
>
> # R Benchmark 2.3 (21 April 2004)
> # Warning: changes are not carefully checked yet!
> # version 2.3 adapted to R 1.9.0
> # Many thanks to Douglas Bates (bates_at_stat.wisc.edu) for improvements!
> # version 2.2 adapted to R 1.8.0
> # version 2.1 adapted to R 1.7.0
> # version 2, scaled to get 1 +/- 0.1 sec with R 1.6.2
> # using the standard ATLAS library (Rblas.dll)
> # on a Pentium IV 1.6 Ghz with 1 Gb Ram on Win XP pro
>
> # revised and optimized for R v. 1.5.x, 8 June 2002
> # Requires additionnal libraries: Matrix, SuppDists
> # Author : Philippe Grosjean
> # eMail : phgrosjean_at_sciviews.org
> # Web : http://www.sciviews.org
> # License: GPL 2 or above at your convenience (see: http://www.gnu.org)
> #
> # Several tests are adapted from the Splus Benchmark Test V. 2
> # by Stephan Steinhaus (stst_at_informatik.uni-frankfurt.de)
> # Reference for Escoufier's equivalents vectors (test III.5):
> # Escoufier Y., 1970. Echantillonnage dans une population de variables
> # aleatoires réles. Publ. Inst. Statis. Univ. Paris 19 Fasc 4, 1-47.
>
> ...
> Error in getClass(Class, where = topenv(parent.frame())) :
> "geMatrix" is not a defined class
> Calls: new -> getClass
> Execution halted
>
> ______________________________________________
> R-help_at_r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>

-- 
Brian D. Ripley,                  ripley_at_stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595


______________________________________________ R-help_at_r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.

Received on Tue 13 May 2008 - 17:11:35 GMT

Archive maintained by Robert King, hosted by the discipline of statistics at the University of Newcastle, Australia.
Archive generated by hypermail 2.2.0, at Tue 13 May 2008 - 17:30:40 GMT.

Mailing list information is available at https://stat.ethz.ch/mailman/listinfo/r-help. Please read the posting guide before posting to the list.

list of date sections of archive