Hello,

A quick look in Open MPI source code seems to say that it's manipulating
XML topologies in these lines.

Please go into your hwloc-1.9 build directory, and run "tests/xmlbuffer"
(you will may have to build it with run "make xmlbuffer -C tests").

If it works, try running "make check".

Also, can you check whether hwloc is using libxml2 or not? (just check
whether topology-xml-libxml.lo is built under src, or look for
libxml2.so in "ldd libhwloc.so").

thanks
Brice



Le 09/07/2014 14:06, Nick Papior Andersen a écrit :
> Dear users,
>
> I think this is some kind of bug, but I would like to post here to
> hear if this is true.
> I have only tested this using the fortran compiler and fortran version
> of openmpi/hwloc.
>
> My setup:
> intel compiler:
>   composer_xe_2013.3.163
> ifort --version:
>   13.1.1 20130313
>
> I am compiling using these flags:
> icc/ifort -m64 -fPIC -O2 -xHost -prec-div -prec-sqrt -opt-prefetch
>
> My test code is very basic (compiled using mpif90 -o hw hw.f90):
> ##### hw.f90
> program hello_world
> use mpi
> integer :: rank, size, ierror, tag, status(MPI_STATUS_SIZE)
>
> call MPI_INIT(ierror)
> call MPI_COMM_SIZE(MPI_COMM_WORLD, size, ierror)
> call MPI_COMM_RANK(MPI_COMM_WORLD, rank, ierror)
> print*, 'node', rank, ' of ',size
> call MPI_FINALIZE(ierror)
>
> end program hello_world
> ##### hw.f90
>
> I compiled these versions:
>
> A) hwloc-1.8.1 and openmpi-1.6.5
> B) hwloc-1.9 and openmpi-1.6.5
> C) hwloc-1.8.1 and openmpi-1.8
> D) hwloc-1.9 and openmpi-1.8
>
> Besides the --prefix flag I did this:
> For all compilations of hwloc I compiled using these configure settings:
> --disable-{libnuma,opencl,cuda,nvml,gv} --enable-static
> For all compilations of openmpi I compiled using these configure settings:
> --with-hwloc=<placement of equivalent hwloc installation>
>
> A and C both terminate perfectly. No errors.
> B and D both terminate with errors. I have attached the error message
> for both instances.
>
> It seems related to this:
>   ../../orte/util/nidmap.c at line 106
> which does something with a hwloc topography.
>
> I have also tried option D with the GNU compiler: 
> gfortran/gcc --version
>   4.7.2-5
> In this case I don't get any errors. Hence my suspicion lies entirely
> on intel compiler and hwloc-1.9. :)
>
> Note, I can run some of the utilities installed by hwloc without
> problems. However, I am in no way experienced in debugging hwloc so I
> might be doing something wrong here?
> Also, I am not sure whether this is related to openmpi or hwloc,
> however, as I only found the error dependent on the hwloc version I
> suspected this might be the correct place.
>
> Please advice me if you need additional information.
> PS: I could not find anything on the bug-tracker.
>
> -- 
> Kind regards Nick
>
>
> _______________________________________________
> hwloc-users mailing list
> hwloc-us...@open-mpi.org
> Subscription: http://www.open-mpi.org/mailman/listinfo.cgi/hwloc-users
> Searchable archives: 
> http://www.open-mpi.org/community/lists/hwloc-users/2014/07/index.php

Reply via email to