Array size mismatch in atmosphere.F90

[thomas-robinson]

When I run AM4 with the gnu compiler on gaea (prod), I encounter an error with an invalid reference to memory. When running in debug mode, the following error occurs:

Fortran runtime error: Array bound mismatch for dimension 4 of array 'qtend' (4/26)

This happens on line 437 of driver/GFDL/atmosphere.F90

   if ( any((/ id_qdt_dyn, id_qldt_dyn, id_qidt_dyn, id_qadt_dyn /) > 0) .or. &
        query_cmip_diag_id(ID_tnhusa) ) qtend(:, :, :, :) = Atm(mygrid)%q (isc:iec, jsc:jec, :, :)

Are qtend and Atm(mygrid)%q supposed to be the same size?

[bensonr]

This is a poor programming choice as the behavior will be compiler dependent. Traditionally there has been nothing in the standard about implied loops and whether the extents are right hand defined or left hand defined. Regardless, the qtend array is explicity defined to have a length of 4, because is is supposed to contain only the water vapor + 3 condensate species as defined for AM3/AM4.

Does this need to be fixed for 2020.02 or can it wait for another release?

[lharris4]

Hi, Tom, Rusty. S-J *always* encouraged explicit loops when doing an array assignment unless the assignment is to a single value. Lucas

…

On Mon, Apr 27, 2020 at 2:17 PM bensonr ***@***.***> wrote: This is a poor programming choice as the behavior will be compiler dependent. Traditionally there has been nothing in the standard about implied loops and whether the extents are right hand defined or left hand defined. Regardless, the qtend array is explicity defined to have a length of 4, because is is supposed to contain only the water vapor + 3 condensate species as defined for AM3/AM4. Does this need to be fixed for 2020.02 or can it wait for another release? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#34 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AMUQRVBVE5MZYEVMZN3AXQTROXDZXANCNFSM4MSAMBQA> .

[bensonr]

As this logic is purely for a diagnostic added by the physics team and only affects the GFDL driver, I will fix this change for the GNU compiler. Look for a coming PR for you to approve.

[thomas-robinson]

This can wait, but it is needed for running with gfortran