ERROR: incompatible domain grid coordinates

[rajan_cr@hotmail_com]

We started with CCSM 3.0.1 and ran it for T42 gx1v3 - it runs for around 10 minutes and crashed showing some fault.
But none of the logs reported the actual error. After this we moved to CCSM 3.0.1.14 beta version that has been released in 2006
and we were able to see the actual cause of the error in the logs. It goes like this:
************************************************************************
(main) -------------------------------------------------------------------------
(main) check atm/lnd and ocn/ice model domains for consistency
(main) -------------------------------------------------------------------------
(cpl_domain_compare) domain #1 name: atm contract domain
(cpl_domain_compare) domain #2 name: lnd contract domain
(cpl_domain_compare) the domain size is = 8192
(cpl_domain_compare) maximum latitude difference = 0.000E+00
eps_grid = 0.100000000000000002E-01
ndiff = 0
(cpl_domain_compare) maximum longitude difference = 0.000E+00
longitude eps_grid = 0.100000000000000002E-01
longitude ndiff = 0
(cpl_domain_compare) maximum relative error of area (model) = 0.000E+00
(cpl_domain_compare) maximum relative error of area (map) = 0.000E+00
(cpl_domain_compare) domain #1 name: ocn contract domain
(cpl_domain_compare) domain #2 name: ice contract domain
(cpl_domain_compare) the domain size is = 122880
(cpl_domain_compare) maximum latitude difference = 1.009E+04
eps_grid = 0.100000000000000002E-01
ndiff = 74037
(cpl_domain_compare) maximum longitude difference = 1.800E+02
longitude eps_grid = 0.100000000000000002E-01
longitude ndiff = 148073
(cpl_domain_compare) ERROR: incompatible domain grid coordinates
(shr_sys_abort) ERROR: (cpl_domain_compare) incompatible domain grid coordinates
(shr_sys_abort) WARNING: calling shr_mpi_abort() and stopping
*****************************************************************************
As seen above, clearly there is no difference in files for atmosphere and land while there is a difference seen for ocean and ice models. (Same problem with T31 model too). Currently we are running this on Linux and IBM XL compiler. Any help is much appreciated.

 

[dbailey]

Out of the box there should not be a problem with the gx1v3 grid. Do you have all of the appropriate grid files for your machine? Note that the ice and ocean point to the same set of files. They just have different naming conventions.

Dave

 

[rajan_cr@hotmail_com]

In cpl_domain_compare() function there are statements like the following:

!--- enforce consistency ?? ---
enforce = .false.
if (present(enforce_mask) .and. enforce_mask) enforce = .true.
if (present(enforce_all ) .and. enforce_all ) enforce = .true.

The same code can be rewritten as mentioned below

-------------
if (present(enforce_mask) then
if (enforce_mask) enforce=.true.
--------------

The first method is not safe because the compiler is allowed to evaluate the header argument (optional argument) before the present function is evaluated. If the header argument is not in fact present an out of bounds memory reference could occur, which could cause a failure.


Please refer this link http://www.cs.rpi.edu/~szymansk/OOF90/bugs.html

Precisely this is the issue.. We modified the code as mentioned in the second method and the whole run went fine.