[R] R + Linux
Prof Brian Ripley
ripley at stats.ox.ac.uk
Fri Jun 6 23:09:10 CEST 2008
This is not sound advice. For 1GB yes, perhaps 2GB. Beyond that the
extra freedom in the address space of a 64-bit system pays off.
The user address space of a 32-bit Linux system is (in the examples I have
seen) 3 to 3.5Gb. See ?"Memory-limits" for why that is restrictive.
There are some anomalies, depending on the CPU. On Intel Core 2 Duos
manipulating 64-bit pointers seems to be as efficient as 32-bit ones and
on some platforms (e.g. Mac OS 10.5.3) 64-bit is actually faster than
32-bit R. So very similar CPUs can give quite different performance
differences with 32- vs 64-bit R.
On Fri, 6 Jun 2008, Roland Rau wrote:
> Dear all,
> a related follow up -- with the hope for some feedback from the specialists.
> Is the following general advice justified:
> If one has not more than 4GB RAM and one wants to run primarily R on one's
> Linux machine, it is a good idea to install the 32bit version of the
> operating system.
> The reasons are:
> The machine has 4GB RAM which implies that the 32bit version can
> (theoretically) use the whole available memory address space. The advantage
> of addressing more memory using 64bit is in this instance of a 4GB computer
> lost. Furthermore, 64bit often runs slower than 32bit (see Section 8 of R
> Admin Manual) due to the larger pointer size.
> steven wilson wrote:
>> Dear all;
>> I'm planning to install Linux on my computer to run R (I'm bored of
>> W..XP). However, I haven't used Linux before and I would appreciate,
>> if possible, suggestions/comments about what could be the best option
>> install, say Fedora, Ubuntu or OpenSuse which to my impression are the
>> most popular ones (at least on the R-help lists). The computer is a PC
>> desktop with 4GB RAM and Intel Quad-Core Xeon processor and will be
>> used only to run R.
> R-help at r-project.org mailing list
> 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
More information about the R-help