[R] Rcpp and mclapply

[sophie_brugieres at mpipsykl.mpg.de]

Dear Jeff,

I had suspected something along those lines initially, however, as I had
stated, this only happens sometimes ...

Here is a partial top showing two of my calls:

17404 sophie    20   0 12.4g  11g 1996 R  100  4.6   5:50.11 R
17405 sophie    20   0 12.4g  11g 2016 R  100  4.6   5:49.86 R
17408 sophie    20   0 12.4g  11g 2016 R  100  4.6   5:49.13 R
17411 sophie    20   0 12.4g  11g 2016 R  100  4.6   5:48.39 R
17412 sophie    20   0 12.4g  11g 2016 R  100  4.6   5:48.14 R
 1461 sophie    20   0 12.7g  11g 2016 R    2  4.7  25:19.27 R
 1465 sophie    20   0 12.7g  11g 2016 R    2  4.7  25:19.05 R
 1476 sophie    20   0 12.7g  11g 2016 R    2  4.7  25:18.73 R
 1486 sophie    20   0 12.7g  11g 2016 R    2  4.7  25:18.39 R
 1491 sophie    20   0 12.7g  11g 2016 R    2  4.7  25:18.24 R

the ones in the 1400 range come from one call, the ones with 17000 from
another ...

It seems to be linked to the console instance that the call is made from
in a way that I cannot grasp yet.

Still puzzled ...

    Sophie

> I would bet that you are doing something in C++ that shares some resource
> between the workers and blocks all but one worker at a time.
> ---------------------------------------------------------------------------
> Jeff Newmiller                        The     .....       .....  Go
> Live...
> DCN:<jdnewmil at dcn.davis.ca.us>        Basics: ##.#.       ##.#.  Live
> Go...
>                                       Live:   OO#.. Dead: OO#..  Playing
> Research Engineer (Solar/Batteries            O.O#.       #.O#.  with
> /Software/Embedded Controllers)               .OO#.       .OO#.
> rocks...1k
> ---------------------------------------------------------------------------
> Sent from my phone. Please excuse my brevity.
>
> sophie_brugieres at mpipsykl.mpg.de wrote:
>>Dear all,
>>
>>I have an R script that uses Rcpp, and I have been trying to
>>parallelize
>>it using mclapply (I tried with the multicore and the parallel library)
>>
>>Sometimes (not always, interestingly), the CPU use for each core drops,
>>usually so that the total over all cores reaches 100%, i.e., as fast as
>>if
>>using just one single core fully. I tried my code directly from within
>>emacs, and also using a shell command - it happens either way.
>>
>>I suspect there might be some interaction between Rcpp and the
>>multicore/parallel libraries. Did any R(cpp) user encounter such
>>symptoms?
>>
>>Sophie
>>
>>______________________________________________
>>R-help at r-project.org mailing list
>>https://stat.ethz.ch/mailman/listinfo/r-help
>>PLEASE do read the posting guide
>>http://www.R-project.org/posting-guide.html
>>and provide commented, minimal, self-contained, reproducible code.
>
>
>