[R] Parallel Computing in R

[Pascal Oettli]

Hi,

You already asked this here
(http://stackoverflow.com/questions/28357210/parallel-computing-in-r)
and got some comments.

Regards,
Pascal

On Fri, Feb 6, 2015 at 11:58 AM, Zhenchuan Wang <zwang10 at mtu.edu> wrote:
> I usually need to compute P-values as following:
>  1. generate one sample (usually it is a matrix)
>  2. apply several methods (I already wrote a subfunction for each method,
> and they are independent) to the generated sample to get pvalues.
>  3. compare the pvalues.
> Since each method mentioned above takes long time (always different length
> of time) to compute the pvalue, I am try to computing the pvalues parallel.
> I want to assign computation of each method to each cores (I have intel
> i7). Do you have any suggestion? I put the four subfunctions together as
>
>     main.fun=function(i,x,y,numper)
>     {
>     if (i==1) z=cca1(y,x,numper)
>     if (i==2) z=2
>     if (i==3) z=2
>     if (i==4) z=3
>     z
>     }
>
> Each i indicates ith subfunction.
> But I always get
>
>      task 1 failed - "Lapack routine dgesv: system is exactly singular:
> U[84,84] = 0"
>
> But when I only run `cca1` function (not using `foreach`), there is no
> error.
> The `foreach` is like this
>
>     pvalue=foreach(i=1:4,.combine=c,.packages=c("MASS","base")) %dopar%
> main.fun(i,x,y,500)
>
> The single computation is like this
>
>     pvalue=cca1(y,x,500)
>
> I also put following in the top lines of my program
>
>     library(foreach)
>     library(doSNOW)
>     library(MASS)
>     cl=makeCluster(4,type="SOCK")
>     registerDoSNOW(cl)
>
> **This looks like when I compute the `pvalue` separately not using
> `foreach`, there is no error. But when I combine the subfuntions togeter
> like `main.fun`, it has error.**
>
>         [[alternative HTML version deleted]]
>
> ______________________________________________
> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
> 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.