Skip to content

problems with memory allocation

3 messages · Thomas Lumley, Manuel, Ian Thurlbeck

#
On Thu, 21 Oct 1999, Manuel wrote:

            
I have no problem allocating --vsize 250M using R0.65.1 on either Debian
GNU/Linux or Solaris 2.7.  In fact, I can allocate --vsize 1000M under
Solaris, which is substantially larger than physical memory
  wompom% ~/Rarchive/R --vsize 1000M
then
  R> gc()
              free     total
  Ncells    128269    250000
  Vcells 131024950 131072000



Thomas Lumley 
Assistant Professor, Biostatistics
University of Washington, Seattle


-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
#
I hope that someone  has had a similar  trouble and  will be able to
help us :

 We , have installed the R package in a Digital Workstation  with 500Mb
of
RAM memory, running under Unix operating system. The package works  fine

but when we try to start the program with more than 120Mb, (vsize
- --120M) the
workstation refuses to  allocate this memory. The message that we get
is:
Fatal error: Could not allocate memory for vector heap.

Someone told us that the solution was an appropiate ulimit call, but
when we do ulimit -a we get only a number  1048576. We figure out that
this number
can be the data segment size.
When we do
ulimit -d unlimited
ulimit -s unlimited
ulimit -m unlimited
ulimit -v unlimited
we get the following mesage:
Requested ulimit exceed hard limit. We think that this mean that we have

no limit to
the amount of memory that can be allocated.

We have installed the same version of the program under linux (Redhat
6.0) and we
were also unable to allocate more than 120 M.

I would be very grateful if someone could let us some new advise in
order to solve the
problem.

Note: I don`t know if the R package is able to allocate more than 120M.
We need about 250 M of memory  because currently we are dealing with
problems in hight dimension.
Manuel.




-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
#
On Thu, 21 Oct 1999, you wrote:
Manuel

you could try re-setting (read: raise into the stratosphere) the hard
limits in the kernel:

1. Create 2 files, vm.stanza, proc.stanza:
-----vm.stanza ------
vm:
       ubc-maxpercent=80
       ubc-borrowpercent = 10
       vm-maxwire = 104857600
       vm-maxvas = 2134217728
       vm-vpagemax = 16000
       vm-syswiredpercent = 90  

-----end ------

-----proc.stanza--------
proc:
       per-proc-stack-size = 222097152
       max-per-proc-stack-size = 222097152
       per-proc-data-size = 1134217728
       max-per-proc-data-size = 2134217728
       max-per-proc-address-space = 2134217728
       per-proc-address-space = 2134217728 
-----end-------------

2. Load these into the /etc/sysconfigtab file and the in-core memory copy with:

# sysconfigdb -a -f vm.stanza vm
# sysconfigdb -a -f proc.stanza proc
# reboot

If you already have a vm: or proc: section in /etc/sysconfigtab then replace
the -a flag with -u, see the sysconfigdb man page.

Make sure you have /sbin/swapdefault pointing to your /dev/rz0b swap
partition, otherwise strange things may happen :)

Hope this helps

Ian
--
Ian Thurlbeck                http://www.stams.strath.ac.uk/
Statistics and Modelling Science, University of Strathclyde
Livingstone Tower, 26 Richmond Street, Glasgow, UK,  G1 1XH
Tel: +44 (0)141 548 3667           Fax: +44 (0)141 552 2079
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._