[Met_help] ascii2nc
John Halley Gotway
johnhg at ucar.edu
Thu Mar 25 09:08:13 MDT 2010
Jianting,
No, you do not need to use F2C with the PGI compilers. The PGI C and Fortran compilers are smart enough to talk to each other. That's why we leave the F2C lines in the default PGI Makefile blank.
Perhaps instead, we'll add comments to the Makefile for PGI explaining that F2C is not necessary.
Thanks,
John
Jianting Chu wrote:
> Dear John,
> Thank you. Yes, you are right, there are really the mismatch with PGI
> 10.0 and NetCDF4.0.1. But I have to use PGI 10.0 since WPS, ARW, WPP, et
> al, are compiled with this compiler. The people of that supercompter
> center are trying to deal with this trouble...
> Another question: Should I compile f2c? Because this library is not
> installed in that supercomputer center.
> Thank you very much!
>
> With best wishes,
> Jianting
>
>
> John Halley Gotway wrote:
>> Jianting,
>>
>> Just to make sure everything works, I went ahead and built MET on a
>> 64-bit machine using PGI version 8.0-5 and NetCDF version 4.0.1. I
>> see that you're using PGI version 10.0, but 8.0-5 is the most
>> recent one we have here. It built fine without any problems.
>>
>> Those undefined references you're seeing are still from the NetCDF
>> library. Try running the following command:
>> nm /nfs/home1/HLRS/IPM/ipmjtchu/netcdf4.0.1/lib/libnetcdf_c++.a |
>> grep "add_att"
>>
>> When I run that, I get the following output. If you're not seeing
>> something similar, then you still have issues with NetCDF. Are you
>> perhaps linking to a version of NetCDF that's built on a
>> different machine that may be running a 32-bit operating system? If
>> so, try building NetCDF on the same machine that you're using to build
>> MET.
>>
>> 00000000000056a9 t __add_att__5NcVarFPCcScEND
>> 00000000000059a2 t __add_att__5NcVarFPCcT1END
>> 0000000000005633 t __add_att__5NcVarFPCccEND
>> 0000000000005929 t __add_att__5NcVarFPCcdEND
>> 00000000000058a2 t __add_att__5NcVarFPCcfEND
>> 00000000000057a4 t __add_att__5NcVarFPCciEND
>> 0000000000005aa9 t __add_att__5NcVarFPCciPCScEND
>> 0000000000005d29 t __add_att__5NcVarFPCciPCdEND
>> 0000000000005ca9 t __add_att__5NcVarFPCciPCfEND
>> 0000000000005ba9 t __add_att__5NcVarFPCciPCiEND
>> 0000000000005c29 t __add_att__5NcVarFPCciPClEND
>> 0000000000005b29 t __add_att__5NcVarFPCciPCsEND
>> 0000000000005a22 t __add_att__5NcVarFPCciT1END
>> 0000000000005825 t __add_att__5NcVarFPCclEND
>> 0000000000005725 t __add_att__5NcVarFPCcsEND
>> 00000000000010bc t __add_att__6NcFileFPCcScEND
>> 000000000000121d t __add_att__6NcFileFPCcT1END
>> 000000000000108c t __add_att__6NcFileFPCccEND
>> 00000000000011e2 t __add_att__6NcFileFPCcdEND
>> 00000000000011a2 t __add_att__6NcFileFPCcfEND
>> 000000000000116b t __add_att__6NcFileFPCciEND
>> 00000000000012a3 t __add_att__6NcFileFPCciPCScEND
>> 00000000000013e3 t __add_att__6NcFileFPCciPCdEND
>> 00000000000013a3 t __add_att__6NcFileFPCciPCfEND
>> 0000000000001363 t __add_att__6NcFileFPCciPCiEND
>> 0000000000001323 t __add_att__6NcFileFPCciPClEND
>> 00000000000012e3 t __add_att__6NcFileFPCciPCsEND
>> 0000000000001263 t __add_att__6NcFileFPCciT1END
>> 000000000000112d t __add_att__6NcFileFPCclEND
>> 00000000000010ed t __add_att__6NcFileFPCcsEND
>> 0000000000005640 T add_att__5NcVarFPCcSc
>> 0000000000005930 T add_att__5NcVarFPCcT1
>> 00000000000055d0 T add_att__5NcVarFPCcc
>> 00000000000058b0 T add_att__5NcVarFPCcd
>> 0000000000005830 T add_att__5NcVarFPCcf
>> 0000000000005730 T add_att__5NcVarFPCci
>> 0000000000005a30 T add_att__5NcVarFPCciPCSc
>> 0000000000005cb0 T add_att__5NcVarFPCciPCd
>> 0000000000005c30 T add_att__5NcVarFPCciPCf
>> 0000000000005b30 T add_att__5NcVarFPCciPCi
>> 0000000000005bb0 T add_att__5NcVarFPCciPCl
>> 0000000000005ab0 T add_att__5NcVarFPCciPCs
>> 00000000000059b0 T add_att__5NcVarFPCciT1
>> 00000000000057b0 T add_att__5NcVarFPCcl
>> 00000000000056b0 T add_att__5NcVarFPCcs
>> 0000000000001090 T add_att__6NcFileFPCcSc
>> 00000000000011f0 T add_att__6NcFileFPCcT1
>> 0000000000001060 T add_att__6NcFileFPCcc
>> 00000000000011b0 T add_att__6NcFileFPCcd
>> 0000000000001170 T add_att__6NcFileFPCcf
>> 0000000000001140 T add_att__6NcFileFPCci
>> 0000000000001270 T add_att__6NcFileFPCciPCSc
>> 00000000000013b0 T add_att__6NcFileFPCciPCd
>> 0000000000001370 T add_att__6NcFileFPCciPCf
>> 0000000000001330 T add_att__6NcFileFPCciPCi
>> 00000000000012f0 T add_att__6NcFileFPCciPCl
>> 00000000000012b0 T add_att__6NcFileFPCciPCs
>> 0000000000001230 T add_att__6NcFileFPCciT1
>> 0000000000001100 T add_att__6NcFileFPCcl
>> 00000000000010c0 T add_att__6NcFileFPCcs
>>
>>
>> Jianting Chu wrote:
>>
>>> Dear John,
>>> Thank you for your suggestion. Anyway, it still does not work, and
>>> the error is the same.
>>>
>>> Best wishes,
>>> Jianting
>>>
>>> John Halley Gotway wrote:
>>>
>>>> Jianting,
>>>>
>>>> I've found in the past that sometimes when I compile NetCDF, the "make
>>>> check" does NOT work. Just go ahead and do a "make" followed by "make
>>>> install" and then try recompiling MET. Let me know if
>>>> that works.
>>>>
>>>> Thanks,
>>>> John
>>>>
>>>> Jianting Chu wrote:
>>>>
>>>>
>>>>> Dear John,
>>>>> Thank you for your detailed explaination. But the NetCDF can not be
>>>>> built this time...
>>>>> My step is the follow:
>>>>>
>>>>> 1) make clean
>>>>> 2) export CC=/opt/compiler/pgi/10.0/linux86-64/10.0/bin/pgcc
>>>>> 3) export CXX=/opt/compiler/pgi/10.0/linux86-64/10.0/bin/pgCC
>>>>> 4) export FC=''
>>>>> 5) export F90=''
>>>>> 6) ./configure --prefix=/nfs/home1/HLRS/IPM/ipmjtchu/netcdf4.0.1
>>>>> --disable-netcdf-4 7) make check install (because compile can not be
>>>>> used) In fact, I once used last two steps, and passed.
>>>>> I attached the make.log file from step 7. Would you please
>>>>> checking it?
>>>>>
>>>>> Best wishes,
>>>>> Jianting
>>>>>
>>>>>
>>>>>
>>>>> John Halley Gotway wrote:
>>>>>
>>>>>> Jianting,
>>>>>>
>>>>>> Thanks for sending the output of make. Yes, it looks like there is a
>>>>>> problem compiling with NetCDF. I see that you're compiling MET using
>>>>>> the PGI compilers. And I'm wondering if you compiled the
>>>>>> NetCDF library with the PGI compilers as well. If you just used the
>>>>>> default configure/compile procedure for NetCDF, it probably did NOT
>>>>>> use PGI.
>>>>>>
>>>>>> If you think that may be the problem, please try recompiling your
>>>>>> NetCDF library in /nfs/home1/HLRS/IPM/ipmjtchu/netcdf4.0.1 using the
>>>>>> following commands:
>>>>>> make clean
>>>>>> setenv CC /opt/compiler/pgi/10.0/linux86-64/10.0/bin/pgcc
>>>>>> setenv CXX /opt/compiler/pgi/10.0/linux86-64/10.0/bin/pgCC
>>>>>> setenv FC ''
>>>>>> setenv F90 ''
>>>>>> ./configure --prefix="path/where/you/want/to/install/netcdf"
>>>>>> ./compile
>>>>>>
>>>>>> These steps are described here:
>>>>>>
>>>>>> http://www.dtcenter.org/met/users/support/online_tutorial/METv2.0/compilation/req_libs.php
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>> This will disable the compilation of the NetCDF/Fortran interface
>>>>>> that
>>>>>> MET doesn't use. Once you've re-compiled NetCDF, try recompiling MET
>>>>>> and see if it goes better.
>>>>>>
>>>>>> Thanks,
>>>>>> John
>>>>>>
>>>>>> Jianting Chu wrote:
>>>>>>
>>>>>>
>>>>>>> Dear John,
>>>>>>> Yes, you are right. The prolem maybe caused by the
>>>>>>> installment. We
>>>>>>> are using METv2.0, but do not use the lasted patches.tar.gz
>>>>>>> file.
>>>>>>> Since we need to use MET Tool in one supercomputer, I need to
>>>>>>> build
>>>>>>> the 4 outer libraries. I have built 3 except for f2c.
>>>>>>> I tested to make MET Tool, and it seems something is wrong with
>>>>>>> NetCDF.
>>>>>>> I attached my Makefile and make.log. Would you please taking a
>>>>>>> look?
>>>>>>> Thank you very much!
>>>>>>>
>>>>>>>
>>>>>>> Best wishes,
>>>>>>> Jianting
>>>>>>>
>>>>>>>
>>>>>>> John Halley Gotway wrote:
>>>>>>>
>>>>>>>> Jianting,
>>>>>>>>
>>>>>>>> Thanks for sending that data file. I ran it through ascii2nc here
>>>>>>>> and
>>>>>>>> found that it did retrieve 2,123,040 observations! The header
>>>>>>>> of the
>>>>>>>> NetCDF file is listed below.
>>>>>>>>
>>>>>>>> netcdf wess_198901 {
>>>>>>>> dimensions:
>>>>>>>> mxstr = 16 ;
>>>>>>>> hdr_arr_len = 3 ;
>>>>>>>> obs_arr_len = 5 ;
>>>>>>>> nhdr = 212304 ;
>>>>>>>> nobs = UNLIMITED ; // (2123040 currently)
>>>>>>>>
>>>>>>>> It occurs to me that perhaps you're not using the most recent
>>>>>>>> version
>>>>>>>> of MET. Before we look any further at this, could you please
>>>>>>>> download
>>>>>>>> the latest set of patches for METv2.0, recompile MET, and
>>>>>>>> then try rerunning?
>>>>>>>>
>>>>>>>> Please go to:
>>>>>>>> http://www.dtcenter.org/met/users/support/known_issues/METv2.0/index.php
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>> Read the first section on that page, download the file
>>>>>>>> "METv2.0_patches_20091223.tar.gz" and follow the instructions for
>>>>>>>> applying those patches.
>>>>>>>>
>>>>>>>> Please give that a try and let me know if you continue to
>>>>>>>> experience
>>>>>>>> problems.
>>>>>>>>
>>>>>>>> Thanks,
>>>>>>>> John
>>>>>>>>
>>>>>>>> Jianting Chu wrote:
>>>>>>>>
>>>>>>>>
>>>>>>>>> Dear Sir,
>>>>>>>>> I found a problem with ascii2nc.
>>>>>>>>> I can use this tool to transfer data format, anyway, the
>>>>>>>>> head of
>>>>>>>>> transformed NetCDF is like this:
>>>>>>>>> -------------------------------------------
>>>>>>>>> netcdf wess_198901 {
>>>>>>>>> dimensions:
>>>>>>>>> mxstr = 16 ;
>>>>>>>>> hdr_arr_len = 3 ;
>>>>>>>>> obs_arr_len = 5 ;
>>>>>>>>> nhdr = 212304 ;
>>>>>>>>> nobs = UNLIMITED ; // (0 currently)
>>>>>>>>> --------------------------------------------
>>>>>>>>> That means no observation data have been transformed. In fact,
>>>>>>>>> this
>>>>>>>>> file can not be used in ./point_stat.
>>>>>>>>> I attached the corresponding ASCII file. Would you please
>>>>>>>>> taking a
>>>>>>>>> look at it?
>>>>>>>>> Thank you very much!
>>>>>>>>>
>>>>>>>>> With best wishes,
>>>>>>>>> Jianting
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> ------------------------------------------------------------------------
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> _______________________________________________
>>>>>>>>> Met_help mailing list
>>>>>>>>> Met_help at mailman.ucar.edu
>>>>>>>>> http://mailman.ucar.edu/mailman/listinfo/met_help
>>>>>>>>>
More information about the Met_help
mailing list