R help - Jun 2012 - multicore, mclapply memory problem

Dear list,

I am using mclapply(from multicore) to use multiple cores on one
server. If I do not use mclapply, just one core, my R problem will use
4G memory. When I use mclapply and set mc.cores =10, each core will
use 4G memory. So totally it will use 40G memory. Is it possible let
each core to share some memory on the same server?

Thank you.

On 29/06/2012 07:59, Fabrice Tourre wrote:
> Dear list,
>
> I am using mclapply(from multicore) to use multiple cores on one
> server. If I do not use mclapply, just one core, my R problem will use
> 4G memory. When I use mclapply and set mc.cores =10, each core will
> use 4G memory. So totally it will use 40G memory. Is it possible let
> each core to share some memory on the same server?
Well, actually it does this automatically ....

On a decent OS (despite the posting guide, you did not tell us yours), 
forking will share pages between the forked copies until the page is 
changed.  So in particular any large R objects created before forking 
will remain shared between the worker processes.

And BTW, package multicore was superseded by parallel last year.


-- 
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

Thank you for your reply.

I used debian system. I will try parallel. Thanks.

uname -a
Linux m1000e1-m610-6 2.6.26-2-amd64 #1 SMP Tue Mar 9 22:29:32 UTC 2010
x86_64 GNU/Linux

sessionInfo()
R version 2.14.2 (2012-02-29)
Platform: x86_64-unknown-linux-gnu (64-bit)

locale:
[1] C

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base

other attached packages:
[1] multicore_0.1-7