R help - May 2013 - Trouble in parallel computing with snow package

Dear all,

 I want to make a baysian optimisation of the parameters values of a model,
written in FORTRAN. I thus need a parallel modeling procedure.

I started to use the R package "snow"  but I went through  trouble and
I
hope that you will bring me light on this problem.

Briefly, I first make a function called "run" which will execute the
.exe
of my FORTRAN code (the argument *plac*, is the ID of the plot, I want the
code to run 20 plots simultaneously).
*
**run<-function(plac) {**
**
setwd(paste('L:/Joannes/CASTANEA/Casta_opti2/CASTANEA_OPTI',plac,'/projet_fortran_intel_castanea/',
sep="") )**
**  system('Release/projet_fortran_intel_castanea.exe')               
**
**}**
*
I then use the following code to run the FORTRAN code over 20 plots in the
same time. My computer has 16 processors.

*cl20 <- makeCluster(20, type="SOCK")**
**  clusterApplyLB(cl20,1:20, run)**
**stopCluster(cl20)**
*
But I finally realized that this code did not run all the 20 plots but only
part of them. The code return a "0" for the plots which has been
simulated
and a "127" if nothing is done. If I run my R code several time, the
number
of simulated plots is different (but<20) and the plots which are simulated
are not always the same. I need all the 20 plots to be run simulaneously
and I cannot find out what is the problem with my code.

Any idea about how to solve my problem?

Thank you very much,


Joannès Guillemot
 <joannes.guillemot@u-psud.fr>

	[[alternative HTML version deleted]]