[MITgcm-support] genmake2 on cg01 (fwd)

[Stephanie Dutkiewicz]

Any ideas on this problem - I still haven't found a solution.

synopsis:
I can run c52 on cg01 on one processor. Everything is happy.
But if I compile with MPI
   ../../tools/genmake2 -mods ../code_nonpzd -of 
../../tools/build_options/linux_ia32_pgf77+mpi_cg01

it compiles. But the run blows up after about 20 timesteps.

steph

---------- Forwarded message ----------
Date: Tue, 18 Nov 2003 10:05:47 -0500 (EST)
From: Stephanie Dutkiewicz <stephd at ocean.mit.edu>
To: mitgcm-support at mitgcm.org
Subject: Re: [MITgcm-support] genmake2 on cg01


Hi -
Still problems with my run. It compiled and started to run and
then quickly disintegrated to NaNs. But before it completely
croaked it gave me a hint:
Possibly you have different setenv PARALLEL and nThreads?

Is there something new (since c48 which was the last version
I ran with MPI) that I need to do to run with multiple processors
(c48 required only CPP_EEOPTIONS turn on MPI, and SIZE.h - nPx, nPy)?

from
a very fustrated quinea pig


On Mon, 17 Nov 2003, Patrick Heimbach wrote:

> It seems related once again to this bloody ETIME, CLOC etc. business
> in timers.F
> 
> Could you the following:
> in tools/
> cvs update genmake2
> then in your build-dir.
> make makefile
> make depend
> and re-compile.
> It might be that the latest genmake2 fixes it,
> but not sure. Ed will know.
> 
> -p.
> 
> 
> 
> Stephanie Dutkiewicz wrote:
> > Patrick -
> > Thanks - that produced a Makefile more like I the one I was expecting,
> > but now I have and error when compiling:
> > ------------------------------
> > /usr/local/pgi/linux86/bin/pgftn timers.f -x 124 0x1400 -x 122 0x40 -x 119 
> > 0x10000 -x 123 0x1000 -x 119 0xa00000 -x 127 4 -x 6 0x100 -x 51 0x20 -x 
> > 124 1 -astype 0 -stdinc 
> > /usr/local/pgi/linux86/include:/usr/local/include:/usr/i386-redhat-linux/include:/usr/lib/gcc-lib/i386-redhat-linux/egcs-2.91.66/include:/usr/include 
> > -ieee 1 -opt 2 -idir /usr/local/pkg/mpi/mpi-1.2.4..8a-gm-1.5/pgi/include 
> > -x 125 2 -x 124 8 -x 124 32 -nodclchk -extend -x 80 0x300 -y 80 0x1000 -x 
> > 9 1 -x 32 0x7d000 -vect 32 -x 16 12 -y 34 48 -y 19 8 -y 35 0 -asm 
> > /tmp/pgf77A4Kt2u 
> >   0 inform,   0 warnings,   0 severes, 0 fatal for timer_index
> >   0 inform,   0 warnings,   0 severes, 0 fatal for timer_control
> > PGFTN-S-0038-Symbol, etime, has not been explicitly declared (timers.f: 
> > 1518)
> >   0 inform,   0 warnings,   1 severes, 0 fatal for timer_get_time
> >   0 inform,   0 warnings,   0 severes, 0 fatal for timer_printall
> >   0 inform,   0 warnings,   0 severes, 0 fatal for timer_start
> >   0 inform,   0 warnings,   0 severes, 0 fatal for timer_stop
> > PGFTN/x86 Linux/x86 3.2-4: compilation completed with severe errors
> > pgf77: /usr/local/pgi/linux86/bin/pgftn completed with exit code 1
> > Unlinking 
> > Unlinking 
> > Unlinking /tmp/pgf77A4Kt2u
> > ----------------------------------------------------
> > 
> > On Mon, 17 Nov 2003, Patrick Heimbach wrote:
> > 
> > 
> >>In case you're running with mpi, try
> >>../../tools/genmake2 -mods ../code -of ../../tools/build_options/linux_ia32_pgf77+mpi_cg01
> >>
> >>The file linux_ia32_pgf77+mpi_cg01
> >>should have all the right settings.
> >>
> >>-p.
> >>
> >>Stephanie Dutkiewicz wrote:
> >>
> >>>Hi -
> >>>I am having problems getting my model to run on cg01.
> >>>With the old genmake I compiled as such:
> >>>../../tools/genmake  -platform=cg01+pgi -mods=../code
> >>>
> >>>with genmake2 the above doesn't work. I tried to hardwire 
> >>>the INCLUDES/FC/LINK/FFLAG/FOPTIM
> >>>so that it was the same in the old Makefile. Now the model
> >>>runs, but is spitting out NaN's (even though on another
> >>>machine with all same fields it is fine). Is there maybe an
> >>>endian problem?
> >>>
> >>>Any suggestions?
> >>>
> >>>steph
> >>>
> >>>
> >>>_______________________________________________
> >>>MITgcm-support mailing list
> >>>MITgcm-support at mitgcm.org
> >>>http://dev.mitgcm.org/mailman/listinfo/mitgcm-support
> >>
> >>
> >>
> > 
> > _______________________________________________
> > MITgcm-support mailing list
> > MITgcm-support at mitgcm.org
> > http://dev.mitgcm.org/mailman/listinfo/mitgcm-support
> 
> 
>