[FLASH-BUGS] IRIX64: runtime error

[Mike Zingale]

Hi Shuang, this is an earlier version of IRIX than any of us have tested
on (we use IRIX 6.5).  If you can find the version number of MPI or mpt on
that machine we'd like to know that.  Do you normally run 64 bit or n32
executables on this machine?  The default Makefile probably builds an n32
executable.

Also, try setting up the problem with -debug, and recompile.  This will
use the debug flags in the compiler.  Let us know if there are any
differences.

Another thing to try is setting

msgbuffer = .FALSE.

in your flash.par.  This will turn off some message packing in the mesh
routines (there is a lot of allocation and deallocation going on there).

The memory statistics in the flash.log looked fine.

Mike


On Fri, 8 Feb 2002, szhang wrote:

> Hi Mike & Andrew:
> 
> Thanks for your quick attention.
> 
> 1. Attatched the flash.log file Mike suggested, also amr_log and
> the core dumped, with 1 CPU. Compiler flags I am just using whatever
> provided in the IRIX64 prototype.
> 
> 2. Our system & compiler:
> IRIX64 versicolor 6.2 03131016 IP25
> MIPSpro 7.2.1 (I noticed the problem description on your
> distribution email for 7.3.1.1).
> 
> 3. NOTE FLASH2.0-pre works well on the same
> machine, from 1 to 4 processors. I guess some of the data
> structures you changed from pre-version might matter.
> 
> Best wishes.
> 
> -shuang
> 
> *********************************************
> Zhang, Shuang
> Laboratory for Visiometrics & Modeling
> http://www.caip.rutgers.edu/~zhang
> *********************************************
> ----- Original Message -----
> From: "Mike Zingale" <zingale at flash.uchicago.edu>
> To: "szhang" <zhang at caip.rutgers.edu>
> Cc: <flash-bugs at flash.uchicago.edu>
> Sent: Friday, February 08, 2002 4:06 PM
> Subject: Re: [FLASH-BUGS] IRIX64: runtime error
> 
> 
> > This error seems to suggest that we are reallocating an array that has not
> > been deallocated.  I've never seen this produced with FLASH before.  Can
> > you set
> >
> > memory_stat_freq = 1
> >
> > in your flash.par file and rerun?  This will dump memory statistics to the
> > flash.log file that may show a memory leak (or nothing).
> >
> > Mike
> >
> >
> > On Fri, 8 Feb 2002, Mike Zingale wrote:
> >
> > > Please sent the version of the compilers used (we know that there are
> > > issues with 7.3.1.1), the compilation flags used, the full version of
> the
> > > operating system, and MPI, and the problem name, and output files
> > > (stdout/err, amr_log, flash.log).  This version of FLASH 2.0 runs
> > > regularily on several SGI machines without difficutly.
> > >
> > > Mike
> > >
> > >
> > > On Fri, 8 Feb 2002, szhang wrote:
> > >
> > > > Hi:
> > > >
> > > > Without looking in to further detail:
> > > >
> > > > The new version of FLASH 2.0, was not able to run on IRIX64 with the
> following
> > > > error stdout, while FLASH 2.0-pre works good on the same machine.:
> > > >
> > > > mpirun -np 2 flash2 flash.par
> > > >
> > > > *****************************************
> > > >   CALLING get_tree
> > > >       66 8.80E-02 1.01E-03 1.00E+99 1.00E+99 1.00E+99 1.00E+99
> 1.00E+99 1.01E-03
> > > >       67 9.00E-02 1.01E-03 1.00E+99 1.00E+99 1.00E+99 1.00E+99
> 1.00E+99 1.01E-03
> > > >       68 9.20E-02 1.01E-03 1.00E+99 1.00E+99 1.00E+99 1.00E+99
> 1.00E+99 1.01E-03
> > > >       69 9.40E-02 1.01E-03 1.00E+99 1.00E+99 1.00E+99 1.00E+99
> 1.00E+99 1.01E-03
> > > >       70 9.60E-02 1.01E-03 1.00E+99 1.00E+99 1.00E+99 1.00E+99
> 1.00E+99 1.01E-03
> > > >       71 9.80E-02 1.00E-03 1.00E+99 1.00E+99 1.00E+99 1.00E+99
> 1.00E+99 1.00E-03
> > > >       72 1.00E-01 1.01E-03 1.00E+99 1.00E+99 1.00E+99 1.00E+99
> 1.00E+99 1.01E-03
> > > >
> > > > lib-4411 : UNRECOVERABLE library error
> > > >
> > > > IOT Trap
> > > >
> > > > lib-4411 : UNRECOVERABLE library error
> > > >
> > > > IOT Trap
> > > > MPI: MPI_COMM_WORLD rank 0 has terminated without calling
> MPI_Finalize()
> > > > MPI: aborting job
> > > >
> > > > *****************************************************
> > > >
> > > > Thanks.
> > > >
> > > >
> > > > *********************************************
> > > > Zhang, Shuang
> > > > Laboratory for Visiometrics & Modeling
> > > > http://www.caip.rutgers.edu/~zhang
> > > > *********************************************
> > > >
> > >
> >
>