[Wrf-users] WRFV2.0.3.1 and IFORT 8.1 on an 32-bit i686 (not xeon)

Wed Feb 9 15:30:01 MST 2005

Hi Anthony, Mahally, and all,

   Like Anthony, I am not trying to compile on i686 Xeon but on an Intel
i686 PC (Pentium 4 chipset, with Fedora Core 2, Intel Ifort 8.1
compiler, WRFV2.0.3.1), except that I can't even compile a test case. I
downloaded a fresh version of the zipped tarfile WRF2.0.3.1 and have
made no special changes to the configure file. My NetCDF path is set,
etc.

When I ran the configuration script, I chose option 7 [Intel xeon i686
ia32 Xeon Linux, ifort compiler (single-threaded, no nesting)]. I get
the following output (see bottom), basically a bunch of severe internal
compiler errors. 

   As far as I can tell, the configuration script has no defaults for
i686 non-Xeon PC's, so I'm guessing that a bunch of the flags for the
Xeon cases are incompatible for the PC architecture.

   Anyone have any thoughts on this? Is WRF supported on Intel PC
platforms, or just Xeon workstations?

Jonathan Vigh

[vigh at euler WRFV2]$ ./compile em_b_wave

**** Compiling: WRF_EM_CORE .

make -i -r MODULE_DIRS="-I../dyn_em -I../dyn_nmm  -module ../main
-I../external/io_netcdf -I../external/io_int -I../external/esmf_time_f90
-I../external -I../frame -I../share -I../phys -I../inc" ext
make[1]: Entering directory `/home/vigh/WRF-NCAR/WRFV2'
--------------------------------------
( cd frame ; make -i -r externals )
make[2]: Entering directory `/home/vigh/WRF-NCAR/WRFV2/frame'
( cd ../external/io_netcdf ; \
          make NETCDFPATH=/usr/remote/intel/ CPP="/lib/cpp -traditional"
FC="ifort -FR -I. -w" ; \
          /bin/cp wrf_io_flags.h wrf_status_codes.h ../../inc )
make[3]: Entering directory
`/home/vigh/WRF-NCAR/WRFV2/external/io_netcdf'
make[3]: Nothing to be done for `all'.
make[3]: Leaving directory
`/home/vigh/WRF-NCAR/WRFV2/external/io_netcdf'
( cd ../external/io_int ; \
          make CC="icc" CPP="/lib/cpp -traditional" FC="ifort  -O2 -w
-FR -cm -I. -Vaxlib -convert big_endian -mp  -w" all )
make[3]: Entering directory `/home/vigh/WRF-NCAR/WRFV2/external/io_int'
/bin/cp ../../frame/module_internal_header_util.F .
/lib/cpp -traditional  -I.. module_internal_header_util.F >
module_internal_header_util.f
ifort  -O2 -w -FR -cm -I. -Vaxlib -convert big_endian -mp  -w  -I..
-I../../inc -c module_internal_header_util.f
/lib/cpp -traditional  -I.. io_int.F90 | m4 -Uinclude -Uindex -Ulen - >
io_int.fifort  -O2 -w -FR -cm -I. -Vaxlib -convert big_endian -mp  -w 
-I.. -I../../inc -c io_int.f
/bin/rm module_internal_header_util.o
/bin/rm -f libwrfio_int.a
ar cr libwrfio_int.a io_int.o
echo libwrfio_int.a
libwrfio_int.a
Diffwrf will be built later on in this compile. No need to rerun
compile.
Diffwrf will be built later on in this compile. No need to rerun
compile.
if [ -f ../../frame/pack_utils.o ] ; then \
                /lib/cpp -traditional  diffwrf.F > diffwrf.f     ; \
                ifort  -O2 -w -FR -cm -I. -Vaxlib -convert big_endian
-mp  -w -c  diffwrf.f     ; \
                ifort  -O2 -w -FR -cm -I. -Vaxlib -convert big_endian
-mp  -w -o diffwrf diffwrf.o io_int.o  \
                              ../../frame/pack_utils.o
../../frame/module_internal_header_util.o ../../frame/module_machine.o
../../frame/module_wrf_error.o ; fi
make[3]: Leaving directory `/home/vigh/WRF-NCAR/WRFV2/external/io_int'
( cd ../external/esmf_time_f90 ; \
  make FC="ifort  -O2 -w -FR -cm -I. -Vaxlib -convert big_endian -mp "
CPP="/lib/cpp -traditional -I../../inc -I.  -DEM_CORE=1 -DNMM_CORE=0
-DNMM_MAX_DIM=1250 -DCOAMPS_CORE=0 -DEXP_CORE=0 -DNONSTANDARD_SYSTEM
-DF90_STANDALONE -DCONFIG_BUF_LEN=8192 -DMAX_DOMAINS_F=21" )
make[3]: Entering directory
`/home/vigh/WRF-NCAR/WRFV2/external/esmf_time_f90'
/lib/cpp -traditional -I../../inc -I.  -DEM_CORE=1 -DNMM_CORE=0
-DNMM_MAX_DIM=1250 -DCOAMPS_CORE=0 -DEXP_CORE=0 -DNONSTANDARD_SYSTEM
-DF90_STANDALONE -DCONFIG_BUF_LEN=8192 -DMAX_DOMAINS_F=21 -C -P -I.
-DF90_STANDALONE ESMF_Base.F90 > ESMF_Base.f
ifort  -O2 -w -FR -cm -I. -Vaxlib -convert big_endian -mp  -c -g
ESMF_Base.f
fortcom: Severe: **Internal compiler error: segmentation violation
signal raised** Please report this error along with the circumstances in
which it occurred in a Software Problem Report.  Note: File and line
given may not be explicit cause of this error.
in file (null), line 0, column 0

compilation aborted for ESMF_Base.f (code 3)
make[3]: [ESMF_Base.o] Error 3 (ignored)
/lib/cpp -traditional -I../../inc -I.  -DEM_CORE=1 -DNMM_CORE=0
-DNMM_MAX_DIM=1250 -DCOAMPS_CORE=0 -DEXP_CORE=0 -DNONSTANDARD_SYSTEM
-DF90_STANDALONE -DCONFIG_BUF_LEN=8192 -DMAX_DOMAINS_F=21 -C -P -I.
-DF90_STANDALONE ESMF_BaseTime.F90 > ESMF_BaseTime.f
ifort  -O2 -w -FR -cm -I. -Vaxlib -convert big_endian -mp  -c -g
ESMF_BaseTime.f/lib/cpp -traditional -I../../inc -I.  -DEM_CORE=1
-DNMM_CORE=0 -DNMM_MAX_DIM=1250 -DCOAMPS_CORE=0 -DEXP_CORE=0
-DNONSTANDARD_SYSTEM -DF90_STANDALONE -DCONFIG_BUF_LEN=8192
-DMAX_DOMAINS_F=21 -C -P -I. -DF90_STANDALONE ESMF_Calendar.F90 >
ESMF_Calendar.f
ifort  -O2 -w -FR -cm -I. -Vaxlib -convert big_endian -mp  -c -g
ESMF_Calendar.ffortcom: Severe: **Internal compiler error: segmentation
violation signal raised** Please report this error along with the
circumstances in which it occurred in a Software Problem Report.  Note:
File and line given may not be explicit cause of this error.
in file (null), line 0, column 0

On Wed, 2005-02-09 at 01:45, Anthony Toigo wrote:
> Hi,
> 
> I've also been trying to compile WRFV2.0.3.1 with the ifort v8.1 compiler
> on an i686 processor and have been running into some problems myself.
> 
> Starting with a fresh copy of the tar file, at this point I am simply
> trying to get one of the ideal cases to compile and run as a test.  I have
> been working with "em_b_wave" for now.
> 
> I have a 32-bit Pentium 3 i686 Fedora Core 3 Linux machine that I am
> building and testing this on.  I have the ifort v8.1 compilers, and started
> with gcc (3.4.2-6.fc3) but just today acquired the icc v8.1 compiler as
> well.  The netCDF version I am using is v3.6.0, and was built with ifort
> and gcc.  When compiling, I chose option "7.  Intel xeon i686 ia32 Xeon
> Linux, ifort compiler (single-threaded, no nesting)" with no other
> modifications or flags (with the one exception of using gcc vs. icc, but
> now I have used both and that makes no difference).
> 
> The good news is that WRF compiles just fine.  And ideal.exe runs fine to
> completion with no errors.  The problem comes with running wrf.exe.  The
> code always crashes with a SEGV (segmentation violation) in solve_em.
> The exact line (according to the debuggers) is:
> 
> 1080  history_outname = grid%history_outname
> 
> This same error occurs in the exact same place no matter whether I use gcc
> or icc, so at least I've ruled that out.
> 
> Adding a debug print line, like
> 
> write(0,*) 'history_outname=',history_outname
> 
> before the Registry-generated assignment statements also causes the code to
> crash.  I can't understand why the character*256 variable, or using it in
> an assignment statement, or even just printing it, would cause a
> segmentation violation, but I would like to know.
> 
> Has anyone else experienced this problem?  If anyone has gotten WRF to
> compile on a 32-bit i686 machine with Fedora Core 3 or so and the ifort v8.1
> compilers, did you run into any problems?  Did you make any changes to the
> configure.defaults file to get it to compile and/or run, and if so, what
> changes did you make?  Any help anyone can offer would be appreciated.
> 
> Thanks,
> 
> Anthony Toigo
> Cornell University
> _______________________________________________
> Wrf-users mailing list
> Wrf-users at ucar.edu
> http://mailman.ucar.edu/mailman/listinfo/wrf-users
-- 
---------------------------------------------------------------------
Jonathan Vigh, Ph.D. Candidate              work phone: 970.491.8633                         
Department of Atmospheric Science           vigh at atmos.colostate.edu
Colorado State University    http://euler.atmos.colostate.edu/~vigh/
---------------------------------------------------------------------
~