Re: [FLASH-USERS] FLASH on SunOS

Hi Mike & Jonathan:

Thanks again for your concern.

1. Yes, we do run parallel jobs on this machine. Your doubt on the
parallel system is right since there is certain instabilities and mis-
behavior on this system, which I am urging myself to figure out with
our system administrators. But I was able to run another distributed
AMR package: GrACE successfully on this machine (with upto
128 CPUs).

The configurations of the mahine is available at:

http://www.caip.rutgers.edu/~e10k/

2. Here is my mpi submission command:

bsub -q hpct -n 4 -e e1 -o o1 ./flash2 ./flash.par

Which I also got some reference from

http://www.npaci.edu/HPC10000/

3. Similar problems (not exactly the same) were observed and
discussed at:
http://manila.mems.rice.edu/developer/newsItems/viewDepartment$XNS

I tried the solutions and some further combination of compiler flags,
and none of them works here.

But I do noticed that different flags, all able to compile, gives different
run time
errors: some terminate the program (-fast -O3 -xarch=v9a), some makes
the program halt-- running for ever without any output (-xarch=v9a).

Best

*********************************************
Zhang, Shuang
Laboratory for Visiometrics & Modeling
http://www.caip.rutgers.edu/~zhang
*********************************************
----- Original Message -----
From: "Mike Zingale" <zingale@flash.uchicago.edu>
To: "szhang" <zhang@caip.rutgers.edu>
Cc: <flash-users@flash.uchicago.edu>; "Norman Zabusky"
<nzabusky@caip.rutgers.edu>
Sent: Thursday, February 07, 2002 2:03 PM
Subject: Re: [FLASH-USERS] FLASH on SunOS

> Hi Shuang, a quick look over these files does not reveal the problem -- it
> seems that the code is crashing at the redistribution step after
> refinement. Have you successfully run any other parallel jobs on this
> machine with the same environment?
>
> We will try to find a Sun machine around here to test with.
>
> Mike
>
> On Thu, 7 Feb 2002, szhang wrote:
>
> > Thanks for your quick attention, Mike:
> >
> > Attatched the files you needed:
> >
> > e1 is the stderr
> > o1 is the stdout
> >
> > Is it one possibility that six adptive mesh levels might generate
> > too much load on message passing?
> >
> > Best.
> >
> > -Shuang
> >
> > ----- Original Message -----
> > From: "Mike Zingale" <zingale@flash.uchicago.edu>
> > To: "szhang" <zhang@caip.rutgers.edu>
> > Cc: <flash-users@flash.uchicago.edu>; "Norman Zabusky"
> > <nzabusky@caip.rutgers.edu>
> > Sent: Thursday, February 07, 2002 12:40 PM
> > Subject: Re: [FLASH-USERS] FLASH on SunOS
> >
> >
> > > Can you attach the full sod_2d_0deg_6level.log, amr_log, and stdout
for
> > > this run?
> > >
> > > On Thu, 7 Feb 2002, szhang wrote:
> > >
> > > > Hi:
> > > >
> > > > I am trying to port FLASH2.0 to our Sun Enterprise 10000,
> > > > after sucessfully installed & ran it on our IRIX64 SGI cluster.
> > > >
> > > > After compiled it, I can only run it with single processor on SUN.
The
> > multi-
> > > > procossor run will be either terminated or halted, even for
> > > > the case compiled with 64 bit achitecture (I tried different
compiler
> > > > flags combinations).
> > > >
> > > > Flash stopped at the point in sod_2d_0deg_6level.log:
> > > >
> > > > flash: initializing for sod problem.
> > > > [02-07-2002 11:33.04] <<< refined: tot_blocks = 5 >>>
> > > >
> > > > Our Sun Machine (SunOS teal.rutgers.edu 5.7 Generic_106541-18
> > > > sun4u sparc SUNW,Ultra-Enterprise-10000) is running a LSF queueing
> > > > system for mpi jobs (bsub), which compiled with tmf90.
> > > >
> > > > Do you have any suggestions for the solution? Thanks a
> > > > lot for your attention.
> > > >
> > > > *********************************************
> > > > Zhang, Shuang
> > > > Laboratory for Visiometrics & Modeling
> > > > http://www.caip.rutgers.edu/~zhang
> > > > *********************************************
> > > >
> > >
> >
>
Received on Thu Feb 7 13:54:31 2002