Hi Jonas,

I can replicate this error with X10 2.5.3, however, it appears to be fixed
in the current Git HEAD.  The error appears to be in Team.allreduce, and we
have made a number of changes and improvements to this code since X10
2.5.3.

Are you able to build X10 from the current source HEAD (i.e. by cloning the
X10 Git repository)?  If so, could you please confirm whether this fixes
the problem for you?

We are planning a new release of X10 - 2.5.4 - in a week or so, which will
include the fixes to Team.

Many thanks,
                                                                                
   
 Josh Milthorpe                                                                 
   
 Post Doctoral Researcher                                                       
   
 Machine Learning & Decision Analytics                                          
   
 IBM Research                                                                   
   
                                                                                
   
                                                                                
    
                                                                                
    
                                                                                
    
 Phone: 1-914-945-2209                                            1101 
Kitchawan Rd 
 E-mail: jjmil...@us.ibm.com                             Yorktown Heights, NY 
10598 
                                                                      United 
States 
                                                                                
    







From:   Jonas Posner <jonas.pos...@student.uni-kassel.de>
To:     Mailing list for users of the X10 programming language
            <x10-users@lists.sourceforge.net>
Date:   12/15/2015 02:45 AM
Subject:        Re: [X10-users] MPI execution problems with GLB UTSG



Hey,

unfortunately makes no difference.

Jonas

2015-12-15 0:36 GMT+01:00 David P Grove <gro...@us.ibm.com>:
  Jonas Posner <jonas.pos...@student.uni-kassel.de> wrote on 12/14/2015
  04:55:40 PM:
  >
  > I have some MPI execution problems on a cluster with the UTSG sample in
  GLB.
  > My versions are:
  > x10c++ version 2.5.3 (from git)
  > gcc (GCC) 4.9.3
  > java version "1.8.0_66", Java(TM) SE Runtime Environment (build 1.8.
  > 0_66-b17), Java HotSpot(TM) 64-Bit Server VM (build 25.66-b17, mixed
  mode)
  > mpirun (Open MPI) 1.8.8
  >
  > I build x10 with: ant dist -DX10RT_MPI=true
  > Compiling UTSG in x10.dist/samples/GLB/uts: x10c++ -x10rt mpi
  > UTSG.x10 -o UTSG -O -NO_CHECKS


  Hi,

  Can you try building without the -NO_CHECKS flag?

  The Segfault below smells like a likely null pointer exception would have
  been raised if checking was enabled (accessing address 0x8).

  --dave


  >
  > [hla0001:14174] *** Process received signal ***
  > [hla0001:14174] Signal: Segmentation fault (11)
  > [hla0001:14174] Signal code: Address not mapped (1)
  > [hla0001:14174] Failing at address: 0x8
  > [hla0001:14174] [ 0] /lib64/libpthread.so.0(+0xf850)[0x7f8552514850]
  > [hla0001:14174] [ 1]



  ------------------------------------------------------------------------------


  _______________________________________________
  X10-users mailing list
  X10-users@lists.sourceforge.net
  https://lists.sourceforge.net/lists/listinfo/x10-users

------------------------------------------------------------------------------

_______________________________________________
X10-users mailing list
X10-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/x10-users

------------------------------------------------------------------------------
_______________________________________________
X10-users mailing list
X10-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/x10-users

Reply via email to