[R] Question about R and multiple CPU's

[R. Michael Weylandt]

On Wed, Aug 8, 2012 at 6:18 PM, Eberle, Anthony <aeber at allstate.com> wrote:
> I have a question about multiple cores and CPU's for running R.  I've
> been running various tests on different types of hardware and operating
> systems (64 bit, 32 bit, Solaris, Linux, Windows, RV.10, .12, .15,
> .15.1.)  Generally speaking, it seems that for a single user and process
> that R prefers to have as much resources as possible; especially memory.
>
> I've looked at some of the r-sig groups and it seems most threads about
> multicore or CPU's have to deal with using packages like parallel or
> snow or something such as this and leaving it up to the user what can be
> parallelized and how which makes perfect sense to me.
>
> The question I have is about what the advantage of machines with
> multiple cores or CPU's.  To me, it seems that until R is parallelized
> (or if I am writing my own code that can run in a parallel fashion) that
> a single user and single (non-parallel) process would work just as fine
> on a single core, single CPU box as it would on (for example) a quad
> core with 24 threads?  Is my observation off or am I missing something?
> Of course, I know there are other "things" going on with most modern
> systems such as operating system and other processes that in practice
> affect this observation a bit.  In addition if I had a R environment
> where multiple people were logging in and running R queries that would
> likely gain from multiple CPU's and a multitasking OS I'm assuming.

To a first approximation, I believe that is correct if you are doing
nothing else on the computer. If you are, it's nice to have that
second core around :-)

Note however, that it's very easy to write parallelizable code with R
since every *apply() statement is trivially parallelizable. The more
difficult question is asking yourself whether the overhead of
parallelizing is worth it (and that's a much harder question)

Best,
Michael

>
> I'm just trying to get a feeling and validate what I think I am seeing.
> Thoughts and comments are appreciated.
>
>
> Thanks,
>
> Anthony
>
>         [[alternative HTML version deleted]]
>
> ______________________________________________
> 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.