Quoting Juergen Rose <r...@rz.uni-potsdam.de>:
Hi,
I was just able to install the patched Rmpi on the second system with
openmpi-1.5.3. What can we that Rmpi_0.5-9a.tar.gz becomes a standard
CRAN package?
Regards Juergen
Now there is a new problem. Testing openmpi-1.5.3-r1 and
Rmpi_0.5-9a.tar.gz with the attached R script I got a Segmentation
fault or the program hangs in the vicinity of closeCluster() and
mpi.quit(), which does not happen with openmpi-1.4.3 and
Rmpi_0.5-9.tar.gz:
rose@orca:/home/rose/Txt/src/Test/R/Parallel(5)$ qlist -Iv openmpi
sys-cluster/openmpi-1.5.3-r1
rose@orca:/home/rose/Txt/src/Test/R/Parallel(6)$ mpirun -np 4 -wdir .
R --slave --vanilla -f mpi_test_parallel_mini.R
[1] "Res=1.000 usrtime= 0.0 elatime= 0.0"
[1] "Res=4.000 usrtime= 0.0 elatime= 0.0"
[1] "Res=9.000 usrtime= 0.0 elatime= 0.0"
[1] "Res=16.000 usrtime= 0.0 elatime= 0.0"
[1] " Res=7.500 userTime= 0.0 elapTime= 0.0"
[1] "before closeCluster()"
[1] "after closeCluster()"
[orca:29255] *** Process received signal ***
[orca:29255] Signal: Segmentation fault (11)
[orca:29255] Signal code: Address not mapped (1)
[orca:29255] Failing at address: 0x7f8869c0e460
Speicherzugriffsfehler
rose@caiman:/home/rose/Txt/src/Test/R/Parallel(47)$ qlist -Iv openmpi
sys-cluster/openmpi-1.4.3
rose@caiman:/home/rose/Txt/src/Test/R/Parallel(48)$ mpirun -np 4
-wdir . R --slave --vanilla -f mpi_test_parallel_mini.R
[1] "Res=1.000 usrtime= 0.0 elatime= 0.0"
[1] "Res=4.000 usrtime= 0.0 elatime= 0.0"
[1] "Res=9.000 usrtime= 0.0 elatime= 0.0"
[1] "Res=16.000 usrtime= 0.0 elatime= 0.0"
[1] " Res=7.500 userTime= 0.0 elapTime= 0.0"
[1] "before closeCluster()"
[1] "after closeCluster()"
It seems, that the issue is not a special Rmpi problem, compare
https://bugs.gentoo.org/show_bug.cgi?id=373923.
______________________________________________
R-help@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.
