[R] memory usage on Windows (was no subject)

[ripley@stats.ox.ac.uk]

On Wed, 19 Mar 2003, kjetil brinchmann halvorsen wrote:

> On 19 Mar 2003 at 8:57, Michael D. Ward wrote:
> 
> Please use a subject line!

Yes, please do follow your own advice!

> I have a machine with 1.5Gb of physical memory, and have for a long 
> time used --max-mem-size 2G, 
> without any problems. Just yesterday that wasn't enough, so I tried 
> to increase to --max-mem-size 4G, which was rejected by R. 
> I lowered to 3G, and the program seemt to be running fine, but 
> eventually I had to kill the process (hanging). 
> 
> This is windows XP professional. So, at least on XP, there is 
> certainly not a hard coded limit of 1G. But why the limit somewhere 
> between 3 and 4G?
> 
> Have others had problems with R processes eventually hanging on XP, 
> if using more than 2G og memory?

XP is said to have a 2Gb limit for the user memory address space.  So does
Windows 2000 Professional, but Server is said to have 3Gb.  That's for all
processes, not per process.  So I would expect a 2Gb or 3Gb limit, from 
the OS, not R.

As the documentation says (in the CHANGES file), up to 1700Mb has been
used successfully on 2000/XP.

> 
> Kjetil Halvorsen
> 
> > Has anyone allocated more than one Gig of memory for R under Windows? When I try the diagnostic tells me I am decreasing memory and the memory size value is a very large negative number.
> > 
> > I suspect the problem may be a signed integer that overflows. Is 1 G a hard
> > limit for memory allocation in R under Windows?

-- 
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