Subject: Re: CCM3.6.6. with Sun Fortran 90 V1.2
From: Jim Rosinski (rosinski@bearmtn-e0.cgd.ucar.edu)
Date: Tue Nov 24 1998 - 09:30:16 MST
On Tue, 24 Nov 1998, BI XUNQIANG wrote:
> If your SUN workstation (or server) are using Sun Fortran 90 V1.2, you
> need
> do a little more work except following the step which Kluzek suggested.
> The reason is that Sun Fortran 90 V1.2 take real*16 as real*8 (in
> another word, there is no real*16 or complex*32 to f90 compiler), so
> you need to convert both the gauaw.F90 and bsslzr.F90 to FORTRAN 77
> file gauaw.F and bsslzr.F, then reload the object file.
Unfortunately you have to do lots more than this, unless I'm missing
something. Sun's f90 compiler does not have an auto-promotion flag for
floating point precision (-r8 on many compilers). Thus, all variables
declared "real" end up as 32-bit numbers. If nothing else, the netcdf
library calls will fail, or at least result in garbage due to a mismatch of
internal data types. Even if the model runs, a "32-bit climate" is something
which requires careful evaluation.
Regards,
Jim Rosinski
CCM Core Group
This archive was generated by hypermail 2b27 : Thu Jun 01 2000 - 09:07:42 MDT