[RkWard-devel] R memory

[Stefan Roediger]

Am Dienstag, 27. Februar 2007 00:35 schrieb Thomas Friedrichsmeier:
> On Monday 26 February 2007 23:13, Stefan Roediger wrote:
> > does any of you know a recent information about R memory management. I
> > added link from RNews in the "Trouble Shooting Section" but it's dated
> > from 2001/1. http://cran.r-project.org/doc/Rnews/Rnews_2001-1.pdf
> > I can't remember exactly but there was a change in 2005 or so. Does any
> > of you know?
>
> There's also http://www.stat.uiowa.edu/~luke/R/gengcnotes.html, but it's
> more about the internals, hence may have changed more, and also, I don't
> know, how old it is. Anyway, you're asking because of concern for limiting
> memory allocation?
>
No not really a concern here. It has just an informative character for the 
user therefore I wanted something up  to date. I did some "tests"  on my 
machine (wilcoxon test of course ;) ) and tried to crash R (successfully) but 
this took a while on my average machine (512 MB RAM, 900sth MB SWAP). Giving 
such information is just a kind try to make people less frustrated in 
case ... . That's why I also gave the corresponding links for the mem test in 
the help page.

> As far as I can see, the information on that is still correct. I've given
> mem.limits() a short test, and it seems to work fine. So we could/should
> provide an interface to that function to limit memory usage. However, I
> don't think it's an urgent issue for now.

True, this is not urgent (otherwise I would have put it on the ToDo list and 
so on). I also don't see it as an important issue.

> Users dealing with huge data or 
> excessively memory-hungry tests can be expected to read up on mem.limits()
> for a while longer (and before looking about a corresponding setting in the
> GUI, they'd need to be aware of the issue, anyway).

Yes, true. That's why I made a note and wanted to give some references which 
describe what happens, how can one influence it and so on. Not more not less.

>
> Or did you have something else / more specific in mind?

No.
Thanks for the input.

>
> Regards
> Thomas

Stefan