[mpich-discuss] MPICH2-1.1 + PVFS-2.8.1 installation problem

[Shawn Kim]

On Wed, Jul 1, 2009 at 10:35 AM, Rob Latham <robl at mcs.anl.gov> wrote:

> On Tue, Jun 30, 2009 at 08:31:10PM -0400, Shawn Kim wrote:
> > Hi,
> >
> > I installed PVFS-2.8.1 and MPICH2-1.1 correctly (I believe).
> >
> > MPICH2 configuration options:
> >
> > $ ./configure --enable-romio --enable-threads=multiple --enable-debuginfo
> > --enable-g=dbg --prefix=$MYHOME/mpich2-install/
> > --with-pvfs2=$MYHOME/pvfs-2.8.1 --with-file-system=ufs+nfs+pvfs2 |& tee
> > c.txt
> >
> > I could compile mpich2-1.1 with the above option successfully,  but
> whenever
> > I try to run a sample program, errors occured.
> >
> > $mpirun -n 2 mpi-io-test
> > rank 1 in job 3  aum10.cse.psu.edu_41952   caused collective abort of all
> > ranks
> >   exit status of rank 1: killed by signal 9
> > rank 0 in job 3  aum10.cse.psu.edu_41952   caused collective abort of all
> > ranks
> >   exit status of rank 0: killed by signal 4
> >
> > Is there anybody who know this error?
>
> I know of no reason why pvfs-2.8.1 and mpich2-1.1 should fail to work.
> Did you get a core file from those signals?  Can you post a stack
> trace?
>

I am afraid not. Because of urgency I installed mpich2-1.0.8p1 and
pvfs-2.8.1 and both work fine. Also both configure options are fine. I
suspect that valgrind has a issue.


>
> > In addition, I couldn't compile with the following option. I mean I could
> > finish "./configure" successfully, but when make, error happened. The
> only
> > difference is "--enable-g=all" instead of "--enable-debuginfo
> --enable-g=dbg
> > "
> >
> > $ ./configure --prefix=$MYHOME/mpich2-install/
> > --with-pvfs2=$MYHOME/pvfs-2.8.1 --enable-romio --enable-threads=multiple
> > --with-file-system=nfs+ufs+pvfs2 --enable-g=all |& tee c.txt**
> >
> > Error message:
> > $ make
> > ~~~~
> >
> > ../../../../../lib/libmpich.a(trmem.o)(.text+0x24c): In function
> > `MPIU_trvalid_unsafe':
> >
> >
> /home/mdl/seokim/mpich2-1.1+pvfs-2.8.1/mpich2-1.1/src/util/mem/trmem.c:523:
> > undefined reference to `VALGRIND_MAKE_MEM_NOACCESS'
>
> I'll let the other MPICH2 folks weigh in on that one.  Typically those
> trmem warnings kick in when you call malloc/calloc/realloc/free
> directly instead of through the memory management wrappers.
>
> ==rob
>
> --
> Rob Latham
> Mathematics and Computer Science Division
> Argonne National Lab, IL USA
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/mpich-discuss/attachments/20090701/4b07fa11/attachment.htm>