[Wrf-users] FW: Incorrect Results

[Brent Shaw]

All,

Here is some information I received today from Portland Group regarding
the problem I reported.  Looks like it will be corrected with 6.1-3
release on Thursday.

Regards,
Brent 

-----Original Message-----
From: trs at pgroup.com [mailto:trs at pgroup.com] 
Sent: Monday, February 27, 2006 4:47 PM
To: Brent Shaw
Subject: Re: Incorrect Results

Brent,

    This problem is due to TPR 3740, a problem we discovered recently.
The 64-bit size function in f90 is defective. This is fixed in our 6.1-3
release on Thursday.

    Until then, 6.0 release has size working.

    Sorry about the inconvenience.

regards,
dave

>
> -------
> Message
> -------
>
> WRF version 2.1 "real.exe" program does not work properly when 
> compiled with pgf90 6.1.  It works fine with version 6.0.
>
> The code compiles and runs with 6.1, but seems to improperly read some

> of the fields from the input data files.  For example, I ran the 
> serially compiled version of real.exe on the NCAR test data for the 1 
> Jun 2001 case (can be obtained at 
> http://www.mmm.ucar.edu/wrf/src/data/jun01.tar.gz),
> using the WRFV2/test/em_real/namelist.input.jun01 file.  Although it 
> runs, it generates the following error message for every grid point:
>
> changing to land at point [i] [j]
> 1 2 3 4 5 6 7 8 ....
>
> And then a bunch of warnings about the surface temperatures, finally 
> aborting with the message FATAL CALLED Assinging constant soil 
> moisture, bad idea
>
> If I recompile the same source code with version 6.0, it runs to 
> completion without any error.
>
> I have tried using two different NetCDF libraries.  One is the version

> originally compiled with pgi 6.0, the second was a version recompiled 
> with
> 6.1 using the same optimizations as WRF.  I also tried compiling with 
> all of the fortran optimizations turned off in 6.1, but the problem
remains.
>
> I can send you the log files if you like.
>
> Best regards,
>
> Brent
>
>
>
>