parallel-netcdf on a linux cluster

[John Tannahill]

I believe that is the case:

To install pnetcdf, I did:
   setenv MPICC mpiicc
   setenv MPIF77 mpiifc

Makefile for test code:
#

INCDIRS = -I/g/g17/jrt/pnetcdf_mcr/parallel-netcdf-0.9.0/include
LNKDIRS = -L/g/g17/jrt/pnetcdf_mcr/parallel-netcdf-0.9.0/lib

slvwrt:  slvwrt.o
         mpiifc -O3 -xW -tpp7 -Zp16 -ip $(LNKDIRS) -o slvwrt slvwrt.o 
-lpnetcdf -lm -lfmpi -lmpi

slvwrt.o:  slvwrt.F
         mpiifc -O3 -xW -tpp7 -Zp16 -ip -cm -w90 -w95 -extend_source 
$(INCDIRS) -c slvwrt.F

clean:
         /bin/rm -f slvwrt
         /bin/rm -f slvwrt.o
         /bin/rm -f slvwrt.nc



Robert Latham wrote:
> On Thu, Sep 11, 2003 at 10:28:46AM -0700, John Tannahill wrote:
> 
>>I am trying to run my test codes on a linux cluster here at LLNL
>>(mcr).  I have run into a problem.  It appears that the linker
>>(mpiifc) wants one underscore at the end of the nfmpi function
>>names and pnetcdf has two (used nm)?  Does this make sense and is
>>there a quick fix?
> 
> 
> we can only detect the fortran symbol underscore convention at compile
> time, so unfortunately, your app has to be built with the same
> fortran compiler used to build pnetcdf. 
> 
> if that is the case, then  let us know and we'll try to figure out
> what's going on.
> 
> ==rob
> 


-- 
============================
John R. Tannahill
Lawrence Livermore Nat. Lab.
P.O. Box 808, M/S L-103
Livermore, CA  94551
925-423-3514
Fax:  925-423-4908
============================