[R-sig-hpc] undesired forking from either doMC, foreach or multicore
Niels Richard Hansen
Niels.R.Hansen+lists at math.ku.dk
Sat Jan 15 00:55:24 CET 2011
Hi
For what it's worth, I got a bunch of
Break on
__THE_PROCESS_HAS_FORKED_AND_YOU_CANNOT_USE_THIS_COREFOUNDATION_FUNCTIONALITY___YOU_MUST_EXEC__()
to debug
when using multicore (with or without foreach) on my dual core MacBook.
I was running R in a terminal as required for safe use of multicore,
but as far as I can tell, the problem occurred whenever I had used
the quartz device for plotting in the same session.
No plots or plots using the X11 device have so far not resulted in
the problem. See also this post:
https://stat.ethz.ch/pipermail/r-sig-mac/2009-August/006426.html
Best, Niels
On 08/01/11 13.32, Steve Lianoglou wrote:
> Hi,
>
> On Sat, Jan 8, 2011 at 3:21 PM, Michael Braun<braunm at mit.edu> wrote:
>> Thanks for the suggestion. I am not calling additional parallelized routines from the threaded functions. However, I am using the threaded version of the Intel MKL BLAS, and there are matrix manipulations in the threaded functions. Even though MKL is supposed to be thread-safe, would it make sense that this is what is causing the problem?
>
> For what it's worth, I'm running R on a very similar system (only 8
> cores, 2008 vintage, alas), but I *do* run into this problem
> occasionally.
>
> I don't use (don't have) MKL BLAS, and I'm pretty sure I'm not calling
> "additional parallelized routines from the threaded functions" either.
>
> I don't know how to create a meaningful test case to reproduce this
> either, as it seems to happen not so frequently (recently, anyway). My
> solution is the same as yours, which is to kill all "zombie" R
> processes and just try again.
>
> -steve
>
--
Niels Richard Hansen Web: www.math.ku.dk/~richard
Associate Professor Email: Niels.R.Hansen at math.ku.dk
Department of Mathematical Sciences nielsrichardhansen at gmail.com
University of Copenhagen Skype: nielsrichardhansen.dk
Universitetsparken 5 Phone: +1 510 502 8161
2100 Copenhagen Ø
Denmark
More information about the R-sig-hpc
mailing list