[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

20031231: pgf90 5.1 & netcdf



Barnard,

> From: Bernard Legras <address@hidden>
> Organization: ?
> Keywords: netCDF PGI Fortran90 5.1 Mandrake 9.2 Linux

The above message contained the following:

> netcdf runs nicely with PGI pgf90 version 5.0 but it fails the tests of 
> the f90 interface with version 5.1. Here is the result. This is done on a 
> 32 bit Linux box running mandrake 9.2, no extra option to the compiler. 
> Is there a fix or could we use the
> netcdf library safely if we do not use the f90 interface?

Ignoring an interface that doesn't pass its "make test" is always
a possibility (except for the fundamental netCDF-3 C interface, of
course).

> I will do tests under 64-bit Linux and pgf90 on an AMD Opteron. I got
> a factor 3 speedup of my code with respect to 32-bit version and would 
> like to link netcdf library for the analysis tools.
> 
> Happy New Year 
> 
> Bernard Legras
> 
> 
> /d9a/contrib/data/netcdf/netcdf-3.5.0/src/f90
> Making `test' in directory /d9a/contrib/data/netcdf/netcdf-3.5.0/src/f90
> 
> make[2]: Entering directory 
> `/d9a/contrib/data/netcdf/netcdf-3.5.0/src/f90'
> example_good.cdl example.cdl sont différents: octet 881, ligne 37
> *** Failure ***
> *** example_good.cdl    2000-04-05 21:33:14.000000000 +0200
> --- example.cdl 2003-12-31 13:07:11.000000000 +0100
> ***************
> *** 34,43 ****
>     953, 954, 955,
>     956, 957, 958,
>     959, 960, 961,
> !   962, 963, 964,
> !   965, 966, 967,
> !   968, 969, 970,
> !   971, 972, 973 ;
> 
>    lat = -90, -87.5, -85, -82.5 ;
> 
> --- 34,43 ----
>     953, 954, 955,
>     956, 957, 958,
>     959, 960, 961,
> !   950, 951, 952,
> !   953, 954, 955,
> !   956, 957, 958,
> !   959, 960, 961 ;
> 
>    lat = -90, -87.5, -85, -82.5 ;
> 
> make[2]: *** [test] Erreur 1

I've seen this problem before with version 5.1 of the "pgf90" compiler.
(See

    http://www.unidata.ucar.edu/cgi-bin/msgout?/glimpse/netcdf/5426

for more information.)  I suspect that that version of the Portland
Group's Fortran-90 compiler has a bug in it.

Unfortunately, resolution of the problem will have to wait until we
obtain version 5.1 of the pgf90 compiler.  In the meantime, I recommend
using version 5.0 of the compiler, instead of version 5.1.

> -- 
> @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @
> @ Bernard Legras
> @ Laboratoire de Meteorologie Dynamique
> @ Ecole Normale Superieure
> @ 24 rue Lhomond
> @ F-75231 PARIS Cedex 05
> @ tel : (1) 44 32 22 28   fax : (1) 43 36 83 92
> @ E-mail : address@hidden
> @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @

Regards,
Steve Emmerson