Compiling parallel-netcdf 0.8.9 on QSC (ASCI Q)

Tyce Mclarty mclarty3 at llnl.gov
Thu Aug 21 10:35:10 CDT 2003


Can't you use the -i8 fortran flag to force 64 bit integers, so it will 
pass the configure test?

 From f90 man page:

   -i8         Makes default integer and logical variables 8-bytes long (same
               as the -integer_size 64 option).  The default is -integer_size
               32.

Tyce

At 05:59 PM 8/20/2003 -0500, JB Gallagher wrote:
>Rob, Jianwei,
>
>Today I tried getting parallelnetcdf working with the Flash code on ASCI Q 
>(Alpha Compaq cluster running Tru64) I was able to get older versions 
>supplied to me by Jianwei to compile (becasue the configure's don't check 
>if size_t and int are the same size) but code linked with this library 
>fails in a similar way that hdf5_parallel i/o is failing on Q in Flash, I 
>get errors saying the comminucator is not valid:
>
>4 4: The communicator is invalid8
>4 0: Exiting with status 500 (MPI_ERR_COMM in class MPI_ERR_COMM)
>3 3: Exiting with status 500 (MPI_ERR_COMM in class MPI_ERR_COMM)
>
>But most important to you all the configure script from your newest 
>distribution (0.8.9) stops with the size_t int comparison:
>
>checking for size_t... yes
>checking whether size_t and int size match... no
>configure: error: This is a critical error; please report this size 
>mismatch to the developers!
>
>I'll be talking with system folks that administor the ASCI machines 
>tommorow, so I'll tell them what I've learned (as I indicated before even 
>or hdf5_parallel module doesn't work on this system) so far, but if you 
>have any questions that I should ask them based on what I said here let me 
>know before tommorow morning.
>
>thanks,
>
>--Brad
>
>--
>Brad Gallagher
>ASCI Flash Center
>jbgallag at flash.uchicago.edu
>




More information about the parallel-netcdf mailing list