[R-sig-hpc] mpi_comm_spawn error with Rmpi and snow on SGI Altix

Gad Abraham gabraham at csse.unimelb.edu.au
Mon Feb 16 00:24:52 CET 2009


Ei-ji Nakama wrote:
> Hi,
> 
> Sorry, Im poor English.
> 
> 2009/2/11 Gad Abraham <gabraham at csse.unimelb.edu.au>:
>> qsub is from Sun Grid Engine
> 
> Oops.
> 
>>>> #!/bin/sh
>>>> #$ -cwd
>>>> #$ -pe mpi 2
>>>> mpirun -up 2 -np 1 /home/gabraham/bin/R --vanilla < testsnow.R
> 
> | testsnow.R:cl <- makeCluster(4, "MPI")
> 
> necessary CPU resource.
> master 1 + slave 4 = 5
> 
> #!/bin/sh
> #$ -cwd
> #$ -pe mpi 5
> mpirun -up 5 -np 1 /home/gabraham/bin/R --vanilla < testsnow.R

Yep, it works now:

Without snow:
 > system.time({
+    r <- lapply(1:4, fun)
+ })
    user  system elapsed
487.580   0.064 488.578
 >
 > cat("With snow:\n")
With snow:
 > system.time({
+    r <- parLapply(cl, 1:4, fun)
+ })
    user  system elapsed
122.260   0.000 122.503
 >

Thanks for your help!
Gad


-- 
Gad Abraham
Dept. CSSE and NICTA
The University of Melbourne
Parkville 3010, Victoria, Australia
email: gabraham at csse.unimelb.edu.au
web: http://www.csse.unimelb.edu.au/~gabraham



More information about the R-sig-hpc mailing list