[petsc-users] floating point exception… but only when >4 cores are used...

Sat Apr 28 19:36:29 CDT 2012

When I attach debugger on error on the local machine, I get a bunch of lines like this one:

warning: Could not find object file "/private/tmp/homebrew-gcc-4.6.2-HNPr/gcc-4.6.2/build/x86_64-apple-darwin11.3.0/libstdc++-v3/src/../libsupc++/.libs/libsupc++convenience.a(cp-demangle.o)" - no debug information available for "cp-demangle.c".

then ". done"  And then nothing.  It looks like the program exits before the debugger can attach.  after a while I get this:

/Users/spott/Documents/Code/EnergyBasisSchrodingerSolver/data/ebss-input/basis_rmax1.00e+02_rmin1.00e-06_dr1.00e-01/76151: No such file or directory
Unable to access task for process-id 76151: (os/kern) failure.

in the gdb window.  In the terminal window, I get

application called MPI_Abort(MPI_COMM_WORLD, 0) - process 3
[cli_3]: aborting job:

if I just "start_in_debugger"  I just don't' get the "MPI_Abort" thing, but everything else is the same.

any ideas?

-Andrew

On Apr 28, 2012, at 6:11 PM, Matthew Knepley wrote:

> On Sat, Apr 28, 2012 at 8:07 PM, Andrew Spott <andrew.spott at gmail.com> wrote:
> are there any tricks to doing this across ssh?
> 
> I've attempted it using the method given, but I can't get it to start in the debugger or to attach the debugger, the program just exits or hangs after telling me the error.
> 
> Is there a reason you cannot run this problem on your local machine with 4 processes?
> 
>    Matt
>  
> -Andrew
> 
> On Apr 28, 2012, at 4:45 PM, Matthew Knepley wrote:
> 
>> On Sat, Apr 28, 2012 at 6:39 PM, Andrew Spott <andrew.spott at gmail.com> wrote:
>> >-start_in-debugger noxterm -debugger_nodes 14
>> 
>> All my cores are on the same machine, is this supposed to start a debugger on processor 14? or computer 14?
>> 
>> Neither. This spawns a gdb process on the same node as the process with MPI rank 14. Then attaches gdb
>> to process 14.
>> 
>>     Matt
>>  
>> I don't think I have x11 setup properly for the compute nodes, so x11 isn't really an option.
>> 
>> Thanks for the help.
>> 
>> -Andrew
>> 
>> 
>> On Apr 27, 2012, at 7:26 PM, Satish Balay wrote:
>> 
>> > On Fri, 27 Apr 2012, Andrew Spott wrote:
>> >
>> >> I'm honestly stumped.
>> >>
>> >> I have some petsc code that essentially just populates a matrix in parallel, then puts it in a file.  All my code that uses floating point computations is checked for NaN's and infinities and it doesn't seem to show up.  However, when I run it on more than 4 cores, I get floating point exceptions that kill the program.  I tried turning off the exceptions from petsc, but the program still dies from them, just without the petsc error message.
>> >>
>> >> I honestly don't know where to go, I suppose I should attach a debugger, but I'm not sure how to do that for multi-processor code.
>> >
>> > assuming you have X11 setup properly from compute nodes you can run
>> > with the extra option '-start_in_debugger'
>> >
>> > If X11 is not properly setup - and you'd like to run gdb on one of the
>> > nodes [say node 14 where you see SEGV] - you can do:
>> >
>> > -start_in-debugger noxterm -debugger_nodes 14
>> >
>> > Or try valgrind
>> >
>> > mpiexec -n 16 valgrind --tool=memcheck -q ./executable
>> >
>> >
>> > For debugging - its best to install with --download-mpich [so that its
>> > valgrind clean] - and run all mpi stuff on a single machine - [usually
>> > X11 works well from a single machine.]
>> >
>> > Satish
>> >
>> >>
>> >> any ideas?  (long error message below):
>> >>
>> >> -Andrew
>> >>
>> >> [14]PETSC ERROR: ------------------------------------------------------------------------
>> >> [14]PETSC ERROR: Caught signal number 8 FPE: Floating Point Exception,probably divide by zero
>> >> [14]PETSC ERROR: Try option -start_in_debugger or -on_error_attach_debugger
>> >> [14]PETSC ERROR: or see http://www.mcs.anl.gov/petsc/petsc-as/documentation/faq.html#valgrind[14]PETSC ERROR: or try http://valgrind.org on GNU/linux and Apple Mac OS X to find memory corruption errors
>> >> [14]PETSC ERROR: likely location of problem given in stack below
>> >> [14]PETSC ERROR: ---------------------  Stack Frames ------------------------------------
>> >> [14]PETSC ERROR: Note: The EXACT line numbers in the stack are not available,
>> >> [14]PETSC ERROR:       INSTEAD the line number of the start of the function
>> >> [14]PETSC ERROR:       is given.
>> >> [14]PETSC ERROR: --------------------- Error Message ------------------------------------
>> >> [14]PETSC ERROR: Signal received!
>> >> [14]PETSC ERROR: ------------------------------------------------------------------------
>> >> [14]PE[15]PETSC ERROR: ------------------------------------------------------------------------
>> >> [15]PETSC ERROR: Caught signal number 8 FPE: Floating Point Exception,probably divide by zero
>> >> [15]PETSC ERROR: Try option -start_in_debugger or -on_error_attach_debugger
>> >> [15]PETSC ERROR: or see http://www.mcs.anl.gov/petsc/petsc-as/documentation/faq.html#valgrind[15]PETSC ERROR: or try http://valgrind.org on GNU/linux and Apple Mac OS X to find memory corruption errors
>> >> [15]PETSC ERROR: likely location of problem given in stack below
>> >> [15]PETSC ERROR: ---------------------  Stack Frames ------------------------------------
>> >> [15]PETSC ERROR: Note: The EXACT line numbers in the stack are not available,
>> >> [15]PETSC ERROR:       INSTEAD the line number of the start of the function
>> >> [15]PETSC ERROR:       is given.
>> >> [15]PETSC ERROR: --------------------- Error Message ------------------------------------
>> >> [15]PETSC ERROR: Signal received!
>> >> [15]PETSC ERROR: ------------------------------------------------------------------------
>> >> [15]PETSC ERROR: Petsc Release Version 3.2.0, Patch 6, Wed Jan 11 09:28:45 CST 2012
>> >> [14]PETSC ERROR: See docs/changes/index.html for recent updates.
>> >> [14]PETSC ERROR: See docs/faq.html for hints about trouble shooting.
>> >> [14]PETSC ERROR: See docs/index.html for manual pages.
>> >> [14]PETSC ERROR: ------------------------------------------------------------------------
>> >> [14]PETSC ERROR: /home/becker/ansp6066/local/bin/finddme on a linux-gnu named photon9.colorado.edu by ansp6066 Fri Apr 27 18:01:55 2012
>> >> [14]PETSC ERROR: Libraries linked from /home/becker/ansp6066/local/petsc-3.2-p6/lib
>> >> [14]PETSC ERROR: Configure run at Mon Feb 27 11:17:14 2012
>> >> [14]PETSC ERROR: Configure options --prefix=/home/becker/ansp6066/local/petsc-3.2-p6 --with-c++-support --with-fortran --with-mpi-dir=/usr/local/mpich2 --with-shared-libraries=0 --with-scalar-type=complex --with-blas-lapack-libs=/central/intel/mkl/lib/em64t/libmkl_core.a --with-clanguage=cxx
>> >> [14]PETSC ERROR: ------------------------------------------------------------------------
>> >> [14]TSC ERROR: Petsc Release Version 3.2.0, Patch 6, Wed Jan 11 09:28:45 CST 2012
>> >> [15]PETSC ERROR: See docs/changes/index.html for recent updates.
>> >> [15]PETSC ERROR: See docs/faq.html for hints about trouble shooting.
>> >> [15]PETSC ERROR: See docs/index.html for manual pages.
>> >> [15]PETSC ERROR: ------------------------------------------------------------------------
>> >> [15]PETSC ERROR: /home/becker/ansp6066/local/bin/finddme on a linux-gnu named photon9.colorado.edu by ansp6066 Fri Apr 27 18:01:55 2012
>> >> [15]PETSC ERROR: Libraries linked from /home/becker/ansp6066/local/petsc-3.2-p6/lib
>> >> [15]PETSC ERROR: Configure run at Mon Feb 27 11:17:14 2012
>> >> [15]PETSC ERROR: Configure options --prefix=/home/becker/ansp6066/local/petsc-3.2-p6 --with-c++-support --with-fortran --with-mpi-dir=/usr/local/mpich2 --with-shared-libraries=0 --with-scalar-type=complex --with-blas-lapack-libs=/central/intel/mkl/lib/em64t/libmkl_core.a --with-clanguage=cxx
>> >> [15]PETSC ERROR: ------------------------------------------------------------------------
>> >> [15]PETSC ERROR: User provided function() line 0 in unknown directory unknown file
>> >> application called MPI_Abort(MPI_COMM_WORLD, 59) - process 14PETSC ERROR: User provided function() line 0 in unknown directory unknown file
>> >> application called MPI_Abort(MPI_COMM_WORLD, 59) - process 15[0]0:Return code = 0, signaled with Interrupt
>> >
>> 
>> 
>> 
>> 
>> -- 
>> What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead.
>> -- Norbert Wiener
> 
> 
> 
> 
> -- 
> What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead.
> -- Norbert Wiener

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20120428/88752663/attachment.htm>