This archive contains answers to questions sent to Unidata support through mid-2025. Note that the archive is no longer being updated. We provide the archive for reference; many of the answers presented here remain technically correct, even if somewhat outdated. For the most up-to-date information on the use of NSF Unidata software and data services, please consult the Software Documentation first.
--- Forwarded mail from address@hidden (Olaf Heudecker) Date: Thu, 5 Jun 1997 10:49:32 +0200 From: address@hidden (Olaf Heudecker) To: address@hidden Subject: netCDF3.3 on t3e Dear Mr. Davis thank you for your answer concerning the installation of netCDF3.3 on the Cray T3E. I am very interested in using netCDF3.3 on the T3E with MOM. I am new in the buissiness of parallelisation and my job is to get netCDF-output on the T3E MOM runs here. It seems that you and/or some people in gfld have the same problems and I'd be glad if we could keep each other informed. As a first contribution you might be interested in the following: I managed to get the netCDF3.3 running on our t3e (uname -va yields: ... ... 1.3.171 unicosmk CRAY T3E) and F90-Version is: (Cray CF90 Version 2.0.3.0). I had use the following settings: (lines additional to the INSTALL file beginn with a plus sign) 1) Don't use ncx_cray.c for this machine (TODO: figure out why. Hypothesis: different address structure.)(glenn) 2) Environment settings: CPPFLAGS=-DNDEBUG CC=c89 CFLAGS=-O3 FC=f90 FFLAGS='-g -i64 -dp -Wp-F' + nf_test runs successfully even if you allow optimisation: + FFLAGS='-O2 -O unroll2,unsplit2 -i64 -dp -Wp-F' CXX="" 3) Comment out the par_io_init() in ncgen/main.c (line 96) since T3E lacks this routine though being _CRAYMPP (glenn) 4) Compile by hand: (in the directory libsrc) c89 -c -O2 -I. -DNDEBUG putget.c + c89 -c -O2 -I. -DNDEBUG attr.c +5) In fortran/netcdf-cpp.inc manage to define + FORTRAN_HAS_NO_FLOAT and + FORTRAN_HAS_NO_SHORT (maybe this is not necessary with 6) but since the packet compiles rather long I didn't check) +6) In nf_test change any occurences of REAL to REAL (KIND=4) + For practical use: edit the .m4 macros! + This is up to now what I guess all programmers have to do for the + time being +7) Since (our) T3E has no get_arg don't forget to link a little + subroutine: SUBROUTINE GETARG(I,C) IMPLICIT NONE INTEGER I,ilen,ierror CHARACTER*80 C CALL PXFGETARG(I,C,ilen,ierror) END SUBROUTINE GETARG ----- This did it. The only problem is that REALs have to be used with the explicit REAL (KIND=4) declaration in f90. ---End of forwarded mail from address@hidden (Olaf Heudecker)