Potentially wrong grid size in shr_dmodel_readgrid for multi-column scm

[mwarusz]

I am not familiar with this code at all, but this line looks suspect to me and it caused overflows in some of my DP-SCREAM runs:

E3SM/share/streams/shr_dmodel_mod.F90

Line 386 in f50d263

gsize = abs(scm_nx*scm_nx*nzgo)

Shouldn't it be scm_nx * scm_ny * nzgo ?

[bogensch]

Thanks for reporting. So that I can reproduce the error could you please provide some more information for a setup where you see this error. Such as machine used, case, number of elements in the x and y directions and as well domain size. Providing the exact script you used would be best.

[ambrad]

Excellent find, @mwarusz! I'm investigating another issue and have come across this same line as a problem. @bogensch it's possible that the GNU issues you see on PM are in fact due to this line. I have a reliable reproducer on Chrysalis with Intel running and GNU failing, and I isolated the seg fault to an operation that relies on this line being correct. I.e., with the fix, the case now runs.

[bogensch]

Yes thanks @mwarusz and @ambrad, I will test this fix first chance I get. I must admit, I read this issue incorrectly the first time and I see now the error is pretty obvious.

[bogensch]

I can report that putting this fix in solves all the issues I was having when attempting large domain sizes with DPxx. thanks for reporting @mwarusz! I will issue a PR for this fix early next week.