Subject: Important message regarding CCM3.2 on new CRAY f90 compilers
From: Erik Kluzek (erik@cgd.ucar.edu)
Date: Fri Feb 13 1998 - 13:13:14 MST
Important message regarding release 3.2 of the NCAR Community Climate
Model.
We have found a problem in ccm3.2 on the NCAR CRAY computers due to
changes made in the f90 compiler over the last several months. We did
check CCM extensively when we moved from cf77 to f90. However, the f90
compiler is periodically updated with "minor" patches and we are not
informed about these modifications. We do not know when this problem
first appeared, however, it was after March 1997.
The compiler incorrectly deals with a vector loop within the lsm code
resulting in an incorrect calculation. The symptom is that energy is
no longer conserved within the lsm. This effect is not normally
evident since energy and mass conservation is not checked within the
lsm by default. The lsm namelist variable, conchk, must be set to true
for energy and mass conservation to be verified.
If a user wants to determine if a run performed on the NCAR CRAY
computers had this problem, the user should repeat the first several
days of the original run with the original code. If the difference
between the two results is within the rounding error growth (see
Rosinski and Williamson, SIAM J. Sci. Comput, Vol18., No.2, pp552-564,
March 1997) then the compiler error existed in the original run.
Since CCM3.2 is publicly released we have submitted a patch and
a new tar file out on the web. If you down-load CCM3.2 now from either
the ftp site or the web -- you will get the version with the patched code. The
patch sets conchk to to true and the model will now stop if either mass or energy is
not conserved within the lsm.
To obtain the patch see http://www.cgd.ucar.edu:80/cms/ccm3/news.
Respectfully
Erik Kluzek (CCM Core Group)
Erik Kluzek, (CGD at NCAR)
National Center for Atmospheric Research
Boulder CO, (off) (303)497-1326 (fax) (303)497-1324
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
This archive was generated by hypermail 2b27 : Thu Jun 01 2000 - 09:06:05 MDT