[Met_help] Comment on V1.1 Makefile.ibm

Tue Jul 15 11:52:18 MDT 2008

Matt,

Since support for the IBM is new, I've only tried building with xlC and xlf.  If you are successful building with the GNU compilers on an IBM, please let me know.  In particular, I'd like to know if 
any additional or different compiler options were required.

I'll go ahead and change those default paths to say:
CXX          = /set/this/path/to/xlC
FC           = /set/this/path/to/xlf
NETCDF_BASE  = /set/this/path/to/netcdf
BUFR_BASE    = /set/this/path/to/bufrlib
GSL_BASE     = /set/this/path/to/gsl
F2C_BASE     = /set/this/path/to/libf2c

Good luck,
John

Sittel, Matthew C CTR USAF AFWA 2 WXG/WE wrote:
> Hi John,
> 
> Yeah, after I read a bit more, I realized that's what you meant and that
> it was to be changed.  What threw me off was that the other Makefiles
> didn't have your directory path in them, just the ibm specific one.  I
> know you need a placeholder but I wouldn't want my name on it! :)  
> 
> I'm not even sure if the IBM one is where I should start; we have both
> GNU and XLC here.  I've always used the former though since the latter
> threw too many errors when I compiled my verification software.  This
> was back when we were the only people in the building not coding in
> Fortran!
> 
> I'm sorry if I caused any confusion.  I'll keep working and hopefully
> can get it sorted out... I may need to talk to our SysAdmin folks
> though!
> 
> Matt
> 
> -----Original Message-----
> From: John Halley Gotway [mailto:johnhg at rap.ucar.edu] 
> Sent: Tuesday, July 15, 2008 12:40 PM
> To: Sittel, Matthew C CTR USAF AFWA 2 WXG/WE
> Cc: met_help at ucar.edu
> Subject: Re: [Met_help] Comment on V1.1 Makefile.ibm
> 
> Matt,
> 
> Yes, that's true!  During the configuration process, we have users
> configure those paths to point to the correct location on their
> machines.
> Online tutorial page:
> http://www.dtcenter.org/met/users/support/online_tutorial/compilation/co
> nfigure.php
> 
> Perhaps I should change the default path to something like:
> /change/this/path
> 
> Those directories are just what we used during testing.  Since you'll be
> compiling on an IBM, I'll be interested to hear how it goes.  I kept my
> notes from compiling MET up on the bluefire 
> supercomputer.  On bluefire, I compiled everything as 32-bit to avoid
> problems with BUFRLIB.  To do so, I set the OBJECT_MODE environment
> variable to 32.  But I'm not sure if that's a standard thing 
> on all IBMs.  Listed below are the notes from building MET and it's
> dependencies on bluefire.  They may or may not be of use to you.
> 
> Let us know if you have any questions.
> 
> Thanks,
> John Halley-Gotway
> johnhg at ucar.edu
> 
> # Build external libraries and tools:
> 
> # *** VERY IMPORTANT TO DO THIS *** Set the OBJECT_MODE to 32
> setenv OBJECT_MODE 32
> 
> # BUFRLIB (under a minute):
> tar -xvf BUFRLIB.tar
> xlc -c -DUNDERSCORE *.c
> xlf -c -qextname *.f
> ar crv libbufr.a *.o
> 
> # CWORDSH (a few seconds):
> tar -xvf cwordsh.tar
> xlf -qextname -o cwordsh.x cwordsh.f ../bufrlib/libbufr.a
> # Edit line for CWRD in the cwordsh file to point to the current
> directory.
> 
> # F2C (a few minutes):
> make hadd
> make
> 
> # GSL (several minutes):
> gunzip gsl-1.9.tar.gz
> tar -xvf gsl-1.9.tar
> setenv CC /usr/bin/xlc
> ./configure
> --prefix=/blhome/johnhg/MET_releases/external_libs/gsl/gsl-1.9
> make
> make install
> 
> # NETCDF (a few minutes):
> gunzip netcdf-3.6.2.tar.gz
> tar -xvf netcdf-3.6.2.tar
> setenv CC /usr/bin/xlc
> setenv CXX /usr/bin/xlC
> ./configure
> --prefix=/blhome/johnhg/MET_releases/external_libs/netcdf/netcdf-3.6.2
> make check
> make install
> 
> # Build MET (a few minutes):
> gunzip METv1.1beta8.tar.gz
> tar -xvf METv1.1beta8.tar
> cp Makefile_ibm Makefile
> 
> # Configure Makefile as follows:
> # NETCDF_BASE  =
> /blhome/johnhg/MET_releases/external_libs/netcdf/netcdf-3.6.2
> # BUFR_BASE    = /blhome/johnhg/MET_releases/external_libs/bufrlib
> # GSL_BASE     = /blhome/johnhg/MET_releases/external_libs/gsl/gsl-1.9
> # F2C_BASE     = /blhome/johnhg/MET_releases/external_libs/libf2c
> 
> gmake >& make_met.log&
> tail -f make_met.log
> 
> # Redo the Fortran blocking on the sample PREPBUFR files
> cd /blhome/johnhg/MET_releases/METv1.1beta9/data/sample_obs/prepbufr
> foreach file (`ls`)
> /blhome/johnhg/MET_releases/external_libs/cwordsh/cwordsh unblk $file
> temp
> rm $file
> /blhome/johnhg/MET_releases/external_libs/cwordsh/cwordsh block temp
> $file
> rm $temp
> end
> 
> # Run the MET test scripts
> cd /blhome/johnhg/MET_releases/METv1.1beta9/scripts
> ./test_all.sh >& test_all.log&
> tail -f *.log
> 
> 
> 
> 
> Sittel, Matthew C CTR USAF AFWA 2 WXG/WE wrote:
>> Hi folks,
>>
>>  
>>
>> In the Makefile.ibm file, I noticed 4 references to what appears to be
> a
>> personal directory:
>>
>>  
>>
>> NETCDF_BASE  =
>> /homebv/johnhg/MET_releases/external_libs/netcdf/netcdf-3.6.2
>>
>> BUFR_BASE    = /homebv/johnhg/MET_releases/external_libs/bufrlib
>>
>> GSL_BASE     = /homebv/johnhg/MET_releases/external_libs/gsl/gsl-1.9
>>
>> F2C_BASE     = /homebv/johnhg/MET_releases/external_libs/libf2c
>>
>>  
>>
>> I don't know if this will cause a problem, but I'm guessing it could
> be
>> an issue to those outside of the UCAR system (like me)!
>>
>>  
>>
>> Thanks,
>>
>> Matt
>>
>>  
>>
>> // SIGNED //
>>
>> Matthew C. Sittel
>>
>> UCAR Visiting Scientist
>>
>> (402) 294-3473  DSN: 271-3473
>>
>>
>>
>>
>>
> ------------------------------------------------------------------------
>> _______________________________________________
>> Met_help mailing list
>> Met_help at mailman.ucar.edu
>> http://mailman.ucar.edu/mailman/listinfo/met_help